»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg camelia perl6: ... | irclog: irc.perl6.org | UTF-8 is our friend!
Set by sorear on 25 June 2013.
00:32 benabik joined 00:41 jaldhar_ joined 00:44 abnorman left 00:47 fridim_ left 00:57 jaldhar_ left 01:00 jaldhar_ joined 01:02 bruges joined 01:11 ssutch left
raiph jnthn++ # O(MG) 01:33
01:37 abnorman joined 01:48 FROGGS_ joined 01:51 FROGGS left 02:07 crab2313 joined 02:29 abnorman left 02:36 konundra left 02:56 konundra joined 03:05 jaldhar_ left
sorear good * #perl6 03:17
diakopter sorear: hi
sorear so which one of you had the bright idea of trying to load the JVM interop engine at SETTING INIT TIME?
diakopter doesn't know enough to cringe or giggle 03:18
sorear diakopter: I haven't completely read the backlog yet, but it looks like someone thought rakudo didn't load slowly enough, so they unconditionally initialized a major subsystem in the setting init 03:19
03:19 jaldhar_ joined
sorear s/major/expensive/ 03:19
diakopter oh.
sorear also it leaks memory when initialized, which is probably breaking the eval server
and the testsuite 03:20
sorear goes into full backlog mode
03:52 preflex_ joined, ChanServ sets mode: +v preflex_ 03:53 preflex left, preflex_ is now known as preflex 04:04 crab2313 is now known as i_am_old_w_
sorear done, anyway 04:17
to fix the jvm problems we need to modify Threading.pm so that it doesn't try to instantiate a full scheduler immediately 04:18
in every process
we could probably add initialized flags and do it that way
but a large part of me is going "why are you making the setting 200 lines bigger? use a module!" 04:19
cognominal sorear, what do you think of gist.github.com/cognominal/5989429 04:35
04:36 i_am_old_w_ left
sorear I'm having trouble parsing it 04:37
cognominal sorear: what do you have trouble with? 04:39
sorear that's a really dubious grammar
here's why it doesn't work:
C<aa> 04:40
you start to parse in, you now are looking at 'aa>' and try to parse <chunk>+?
start by calling chunk zero times. '>' can't match
now try one time. it gobbles aa>. '>' can't match '' 04:41
so try two times. <chunk> itself fails.
so the whole <chunk>+? fails.
+? is almost never what anyone wants
jnthn wrote a regex single stepper for parrakudo, you don't have to ask me to single-step your regexes :) 04:43
cognominal when you try one time it should gobble 'aa', not 'aa>'
sorear that requires read-your-mind functionality, which is not enabled in this version of perl 6 04:44
cognominal I will try jnthn's stepper 04:45
sorear [ <!before 'C<' > . ]+ treats '>' as any other character
cognominal ho, now I see 04:46
thx
sorear
.oO( RYM not supported in free version of P6 )
cognominal that's a big problem :( :) 04:47
I will learn to use jnthn's stepper 04:48
sorear that was sort of a joke
I'm not sure perl6-debug has seen much real use yet 04:49
lue RYM? 04:52
cognominal jokes often offers big wisdom. With the Perl power come many way to go astray. I suspect that's why many people avoid Perl. I know better because with a simpler language you end up with bigger code. I usually trust more Perl implementers code than mine.
RYM is Read Your Mind
lue figures I'd guess it the instant I ask :)
sorear sadfaces 04:55
someone did something to cause an O(0.5s) regression in jakudo startup time
diakopter wonders what jakudo could mean in Japanese 04:56
sorear nevermind, forgot I had a cpu-hungry background process 04:59
it's back to hugging 6.000s like it should be
05:07 crab2313 joined 05:08 crab2313 left 05:25 benabik left 05:26 benabik joined 05:27 mtk joined 05:36 SamuraiJack joined 05:38 crab2313 joined 05:46 kaare__ joined 05:58 itz joined 06:00 itz_ left 06:02 cooper joined, cooper left, cooper joined
PerlJam anyone awake? 06:08
Teratogen yes
but probably not very useful =) 06:09
PerlJam I want to use classify, but modify the values en passant and I don't quite see how to do it. 06:11
i.e., if I have my @values = <+foo +bar -baz>, I'd like to colect those into two lists <foo bar> and <baz>. 06:13
sorear hmm 06:14
r: my @values = <+foo +bar -baz>; my %h; %h.push(@values.map: { .substr(0,1) => .substr(1) }); say %h.perl 06:15
camelia rakudo 00cd34: OUTPUT«No such method 'substr' for invocant of type 'Any'␤ in block at /tmp/gtYXrEq2Mo:1␤␤»
sorear r: my @values = <+foo +bar -baz>; my %h; %h.push(@values.map: { $^a.substr(0,1) => $a.substr(1) }); say %h.perl 06:16
camelia rakudo 00cd34: OUTPUT«("+" => ["foo", "bar"], "-" => "baz").hash␤»
PerlJam oh, I can do it "manually". I was wishing for a way to do it with classify
sorear yes, it's possible to install a screw with a hammer 06:17
TimToady sounds a bit XY to me too
PerlJam Maybe so. 06:18
TimToady have you tried modifying $_? 06:19
sorear that probably shouldn't work, even if it does :) 06:20
r: my @values = <+foo +bar -baz>; say @values.classify(<-> $_ { s/.//; $0 }).perl 06:21
camelia rakudo 00cd34: OUTPUT«use of uninitialized value of type Any in string context in block at /tmp/PSGybg8EKJ:1␤␤use of uninitialized value of type Any in string context in block at /tmp/PSGybg8EKJ:1␤␤use of uninitialized value of type Any in string context in block at /tmp/PSGybg8EK…
sorear r: my @values = <+foo +bar -baz>; say @values.classify(<-> $_ { $_ = "foo"; "bar" }).perl
camelia rakudo 00cd34: OUTPUT«("bar" => ["foo", "foo", "foo"]).hash␤»
PerlJam huh
sorear r: my @values = <+foo +bar -baz>; say @values.classify(<-> $_ { substr($_,0,1,'') }).perl 06:22
camelia rakudo 00cd34: OUTPUT«Too many positional parameters passed; got 4 but expected between 2 and 3␤ in sub substr at src/gen/CORE.setting:2846␤ in block at /tmp/HW4zwvewGF:1␤␤»
sorear r: my @values = <+foo +bar -baz>; say @values.classify(<-> $_ { substr-rw($_,0,1,'') }).perl
camelia rakudo 00cd34: OUTPUT«Too many positional parameters passed; got 4 but expected between 1 and 3␤ in sub substr-rw at src/gen/CORE.setting:5452␤ in block at /tmp/pp3cpGSR2w:1␤␤»
sorear isn't there supposed to be a 4-arg verison of substr?
n: my @values = <+foo +bar -baz>; say @values.classify(<-> $_ { substr($_,0,1,'') }).perl
camelia niecza v24-86-g39ab531: OUTPUT«{"b" => ["bar", "baz"], "f" => ["foo"]}␤»
sorear it works in niecza. sort of :D 06:23
n: sub what(\l, \r) { my $o = l; l = r; $o }; my @values = <+foo +bar -baz>; say @values.classify(<-> $_ { what(substr-rw($_,0,1),'') }).perl
camelia niecza v24-86-g39ab531: OUTPUT«{"+" => ["foo", "bar"], "-" => ["baz"]}␤»
sorear what() should probably have a name and become a builtin 06:24
if it's not already
n: my $str = 'foo'; say substr($foo,0,1,'') 06:26
camelia niecza v24-86-g39ab531: OUTPUT«===SORRY!===␤␤Variable $foo is not predeclared at /tmp/4Si2k_5hkG line 1:␤------> my $str = 'foo'; say substr(⏏$foo,0,1,'')␤␤Potential difficulties:␤ $str is declared but not used at /tmp/4Si2k_5hkG line 1:␤------> [3…
sorear n: my $str = 'foo'; say substr($str,0,1,''); say $str;
camelia niecza v24-86-g39ab531: OUTPUT«o␤oo␤»
sorear clearly wrong
06:27 daniel-s_ left
TimToady nr: my @values = <+foo +bar -baz>; say @values.classify({ s/(<[+\-]>)//; ~$0 }).perl 06:27
camelia rakudo 00cd34: OUTPUT«use of uninitialized value of type Any in string context in block at /tmp/etehveaOuB:1␤␤use of uninitialized value of type Any in string context in block at /tmp/etehveaOuB:1␤␤use of uninitialized value of type Any in string context in block at /tmp/etehveaOu…
..niecza v24-86-g39ab531: OUTPUT«{"+" => ["foo", "bar"], "-" => ["baz"]}␤»
sorear TimToady: I tried that first :[ 06:28
TimToady it works in niecza++ :)
PerlJam yeah, I tried that already too (rakudo--)
moritz \o 06:29
TimToady o/ 06:30
sorear o/ moritz 06:31
06:32 crab2313 left 06:50 raiph left 07:04 daniel-s_ joined 07:10 PacoAir joined 07:12 Rotwang1 joined 07:18 PacoAir left 07:20 PacoAir joined, FROGGS_ left 07:22 FROGGS joined 07:30 Ayiko joined 07:31 drbean joined
dalek p: c154ef8 | sorear++ | src/vm/jvm/runtime/org/perl6/nqp/ (5 files):
Remove a bunch of FOR_STAGE0 stuff.
 07:46
sorear I have two failing sprintf tests. I hope that's expected
FROGGS sorear: I think so, yes 07:47
sorear .tell jnthn Your problems are because you're initializing the JVM interop module in the setting blocks, which means it initializes on every run. Because the JVM interop leaks memory (this can be fixed) and is slow to start (this probably can't), I'd really rather we didn't start it every time the setting loads 07:58
yoleaux sorear: I'll pass your message to jnthn.
sorear .tell jnthn We should either have the threads stuff do initialize-on-first-use, or evict it to a module which is loaded only when needed 07:59
yoleaux sorear: I'll pass your message to jnthn.
sorear sleep&
08:00 PacoAir left 08:09 mtk left, SamuraiJack left 08:12 mtk joined 08:16 Rotwang1 left
moritz I'll migrate the IRC logs to a new DB schema in a few minutes, and do a major software upgrade 08:21
there'll be a short downtime
08:21 daniel-s_ left, dmol1 left 08:22 spider-mario joined 08:23 domidumont joined, _ilbot left 08:26 dmol joined 08:27 SamuraiJack joined 08:29 mtk left 08:33 mtk joined 08:39 _ilbot joined, _ilbot left 08:42 _ilbot joined
moritz logger is back. 08:43
08:46 SamuraiJack left 08:47 dmol left 08:50 domidumont left, domidumont joined 08:54 estrabd joined 09:07 berekuk joined
moritz d) and is slow to start (this probably can't), I'd really rather we didn't start it every time the setting loads 09:09
09:58 <+yoleaux> sorear: I'll pass your message to jnthn.
sorry, Ronja discovered copy&paste :-)
nwc10 Guten tag, Ronja 09:14
kanst du lesen?
:-)
moritz she can't, being 2.5 years old :-) 09:15
nwc10 yes, I kind of guessed that. E is almost 4 and doesn't even reliably know all letters yet 09:16
one of E's friends *had* cracked reading by about 2.75, but she is a bit special
moritz wow, that's scary 09:17
when I was young it was an accomplishment to read when you came into school
nwc10 not sure whether it was more than just simple words
moritz still scary
nwc10 but still, yes, scary
moritz anyway, the 'today' page from the IRC logs now has a button which loads new lines via ajax 09:21
nwc10 nice 09:22
moritz if anybody wants to improve the UI (nicer layout, display a spinning wheel, whatever), that would be great
I'm not nearly as good with UI as I'd like to be :-/
(but it has worked in the past a couple of times that I came up with something useful but really ugly, and somebody else contributed a nice design; happend with the IRC logs, perl6.org, the 'features' matrix) 09:24
09:44 pmurias joined 09:55 dmol joined 09:59 berekuk left 10:05 berekuk joined 10:10 berekuk left 10:14 PacoAir joined 10:15 mstefanko left
dalek p: f888397 | (Pawel Murias)++ | t/nqp/59-nqpop.t:
Test nqp::istrue on lists.
 10:19
p: d1b71f3 | (Pawel Murias)++ | t/nqp/25-class.t:
Expand test 25 to check that methods have access to surrounding nonstatic lexicals.
10:23 mstefanko joined
jnthn .tell sorear well, how was I to know the interop engine was too inefficient to bind a few types at startup :/ 10:33
yoleaux 07:58Z <sorear> jnthn: Your problems are because you're initializing the JVM interop module in the setting blocks, which means it initializes on every run. Because the JVM interop leaks memory (this can be fixed) and is slow to start (this probably can't), I'd really rather we didn't start it every time the setting loads
jnthn: I'll pass your message to sorear.
07:59Z <sorear> jnthn: We should either have the threads stuff do initialize-on-first-use, or evict it to a module which is loaded only when needed
dalek kudo/nom: 5cab663 | jonathan++ | src/vm/jvm/core/Threading.pm:
Delay initialization of interop until later.

It doesn't take well to being initialized once per spectest run in the eval server due to a memory leak. This may introduce a race condition, but it's better than hosed spectests.		 10:37
jnthn .tell sorear I've done initialize on first use, hopefully it helps.
yoleaux jnthn: I'll pass your message to sorear.
jnthn & 10:38
dalek kudo-js: dd354d7 | (Pawel Murias)++ | / (4 files):
Implement setinvocationspec.
 10:51
kudo-js: 2b4ac4d | (Pawel Murias)++ | src/QAST/Compiler/JavaScript.nqp:
Fix typo in comment.
kudo-js: 7c0b2f8 | (Pawel Murias)++ | / (3 files):
Update code refs once the outer lexpad for them is ready.
kudo-js: 33e6d00 | (Pawel Murias)++ | runtime.js:
Empty lists boolify to false and others to true.
pmurias jnthn: should I merge fix-set-invocation-spec? 10:52
10:54 berekuk joined 12:02 daniel-s_ joined 12:03 spider-mario left 12:06 spider-mario joined
timotimo it's pretty good that you're contributing all those tests :) 12:19
12:40 Rotwang1 joined 12:41 ztt joined 12:57 ztt left, berekuk left 13:04 domidumont left 13:14 frdmn left 13:15 frdmn joined 13:18 berekuk joined
timotimo can i use niecza as a base to figuring out how nodal hyper should work? 13:32
sorear good * #pelr6 13:35
yoleaux 10:33Z <jnthn> sorear: well, how was I to know the interop engine was too inefficient to bind a few types at startup :/
10:37Z <jnthn> sorear: I've done initialize on first use, hopefully it helps.
timotimo tadzik: in your comment, do you mean "why .Str.path rather than .path" or "why not IO.path.new(...) instead of ...Str.path"? 13:41
tadzik the latter 13:42
well, not even .new
just adding .path seems to work
timotimo i think it was because Cool didn't have a .path method that worked i believe 13:43
i fixed that some time ago in rakudo
er, wait, that explains the former, not the latter 13:44
i think i thought it was prettier to have $thing.path rather than IO.path($thing)
tadzik I mean, now panda uses IO.copy 13:45
in your path it's IO.str.copy
* patch
why not IO.path.copy?
timotimo oh? hold on
tadzik I feel a bit confused about Str.copy being file copying, frankly
timotimo hold on, i don't do that 13:46
tadzik hmm
timotimo isn't .path a constructor to create paths?
tadzik I think so
timotimo r: "foo/bar".path.perl.say
camelia rakudo 5cab66: OUTPUT«IO::Path.new(path => "foo/bar")␤»
tadzik yep
oh, wait 13:47
I'm confused again :)
.Str.path
timotimo well, i think .Str.path can be replaced with .path in all those places?
except, wasn't there some strange Find::Result class?
tadzik ...I'm confused again
Find::Result is obsolete now that we have proper Path objects
I'll remove it and get back to you
timotimo i think the reason behind making the .Str was because Find::Result had a "path" method that returned a string
tadzik yeah 13:48
timotimo so i had to use the .Str method instead and turn the result fo that into an actual path
tadzik I see
yeah, that'll explain it
timotimo after removing Find::Result, you can throw away my pull request and just replace .IO.copy with .path.copy globally or something
so, if i see that correctly, the only spot where "is nodal" really changes anything is for unary pre- or postfix things, like .[] or .{} 13:53
13:54 dmol left 13:59 jlaire joined
timotimo turns out even the "is nodal" marker trait isn't declared yet 13:59
13:59 dmol joined
tadzik wow, File::Find becomes so much simplier with IO::Path 14:00
everything Just Works
labster++
timotimo \o/
sorear o/ 14:01
tadzik any idea how do I make submodules point to the latest revision, The Right Way? 14:02
timotimo cd submodule && git fetch && git checkout origin/master && cd .. && git add submodule - i *think*
rom1504 no not add submodule, commit 14:03
tadzik something happened :)
timotimo ah, o
thanks rom1504
rom1504 oh I thought this was #git :D 14:04
tadzik :D 14:05
14:06 birdwindupbird joined, crab2313_ joined
timotimo is going to test his very first, probably wrong, prototype for "is nodal" unary hypers 14:07
huh. 14:13
timotimo typo'd "nodal" to "nodl" and wondered why there were exceptions >_< 14:16
14:18 crab2313_ left
timotimo apparently it's not enough or actually wrong to put "is nodal" on the proto method :| 14:23
14:26 Rotwang1 left
timotimo huh! 14:28
it's parsing, then World gets called to throw a typed exception and then it bombs 14:29
14:30 ecocode joined 14:45 Rotwang1 joined
dalek nda: aa8e228 | tadzik++ | ext/File__Tools:
Bump File::Tools version
 14:47
nda: 7b2d7b2 | tadzik++ | / (5 files):
Chase File::Find changes and IO::Handle.copy deprecations
tadzik I suppose I broke All The Things on the way
after a release of 2013.07 I'm going to tag panda with 2013.07 as well, and advice people to keep the two in sync 14:48
<3 'rakudobrew update all' 14:50
14:52 raiph joined 14:55 ecocode left 15:03 konundra left
pmurias tadzik: we have a rakudobrew? 15:05
tadzik pmurias: yeah, sort of :) 15:06
github.com/tadzik/rakudobrew
pmurias tadzik: what would be super awesome would be a way to build (and test) rakudo on a specified nqp branch
tadzik if you could add rakudo-js to it, that'd be great :) 15:07
pmurias: as noted in the readme, the only requests I accept are pull requests :P
pmurias there's only nqp-js at this point, rakudo-js is still vapourware
tadzik I see
well, I know, but that'd still be cool :)
the moar the merrier 15:08
15:09 mst joined 15:12 rindolf joined 15:22 domidumont joined
moritz
.oO( moarvm the merriervm )
 15:28
timotimo now i know why my "is modal" stuff doesn't end up in the hyperop
because making the hyperop apparently creates a block for that? 15:29
huf that v looks like it feels roman and wants to be a u
timotimo oh, actually ... it's a sub from further above apparently? METAOP_HYPER_CALL thet is used for >>.[], that would certainly make sense and explain things
15:30 FROGGS left
timotimo no, that's wrong, hyper_call is >>(...) isn't it? 15:30
15:34 FROGGS joined, xilo joined
timotimo i ... really don't know how to fix Mu::dispatch:<hyper> to pass on wether or not the op is nodal, because it depends on the object that gets passed in *later* >_< 15:36
that can't be right
yeah, how is that even supposed to work. 15:37
ah, but of course
(<a b>, <c d>>>[0] should look up wether or not [] is nodal in relation to the object that's to the left of the >>, so (...) 15:38
and *that* can certainly work
but if the logic for that is in dispatch:<hyper> only, that can't be enough; otherwise if there's a hypothetical blorp method that is not nodal in List, but it is nodal in Fish, it would have to descend into a List and then not-iterate over Fish (which is, incidentally, Iterable) 15:41
15:48 rindolf left
TimToady I think we can safely assume that the topmost method represents all of them in this regard, and if you're duck-typing harder than that, it's just erroneous 15:51
otoh, if the algorithm itself is recursively hyper, it can make the decision again, I suppose 15:53
15:53 Rotwang1 left
TimToady doesn't really have enough caffeine in him yet to think... 15:53
16:00 berekuk left 16:02 rindolf joined 16:03 SamuraiJack joined 16:08 Rotwang1 joined 16:21 Rotwang1 left
timotimo early prototype ready to get tested 16:22
only need to come up with a whole bunch of tests
what other operations are supposed to be nodal? 16:23
what about something like .elems?
TimToady yes, anything that presumes a structure that would otherwise be recursed into 16:24
.elems on leaves would be rather uninteresting 16:25
timotimo i seem to still be doing "it" wrong
TimToady just as subscripting on leaves tends to be rather too interesting :)
16:25 lelf joined
timotimo should the "is nodal" be placed on the proto method or on all candidates or on both? 16:26
TimToady generally the proto would be the &foo that gets passed in to the higher-order function, I'd think
arguably any trait on the proto maybe oughta be reflected in the multis though 16:27
timotimo okay; but that's what i have and i'm not getting IS_NODAL here; could it be because i'm checking in the method dispatch:<hyper> of the Mu class? and it says Mu \SELF: ...? 16:28
TimToady dunno, I've never actually looked at how rakudo does HOP
timotimo OK
TimToady it's possible it's assuming something for optimization that makes it harder in the general case... 16:29
you'd need a rakudo expert for that, though
TimToady is just the chief handwaver here... 16:30
pmurias jnthn: ping?
timotimo :)
jnthn: ping from me, too (putting traits on protos vs multi candidates)
16:33 domidumont left, lelf left
pmurias jnthn: should the class really be in the list of it's parents? as returned my .HOW.parents? 16:33
16:33 lelf joined
moritz r: say Int.^parents 16:34
camelia rakudo 5cab66: OUTPUT«␤»
moritz r: say Int.^parents.perl
camelia rakudo 5cab66: OUTPUT«()␤»
moritz huh.
pmurias moritz: in nqp 16:36
dalek osystem: 7e6294d | duff++ | META.list:
Add Questhub to META.list
16:37 benabik left
pmurias nqp: my $parent := NQPMu.HOW.parents(NQPMu)[0];say($parent.HOW.name($parent)) 16:37
camelia nqp: OUTPUT«NQPMu␤»
pmurias moritz: ^^
[Coke] wonders if it is now safe to run the daily spectest run
16:37 benabik joined
timotimo for a custom-built operator that "is nodal", the hyper operator will not flatten nor descend 16:39
[Coke] hurls regexcrossword.com/
timotimo oh that's nicer than the one that went around the net half a year ago 16:40
16:40 konundra joined
moritz pmurias: ok 16:40
maybe it's easier for mro calculation that way?
moritz has no idea 16:41
timotimo In order to save your progress you have to login with Facebook. <- ... wat :(
moritz github login would be the far more natural choice, IMHO 16:42
timotimo javascript localstorage even more so.
oh, i know why i'm failing. i'm calling Parcel's postcircumfix:<[ ]> all the time >_> 16:44
and i neglected to mark that nodal
benabik Hm. A* down, .* .* across. That _should_ match anything I fill out, including nothing, yes? Or are the patterns all surrounded by ^$ ?
timotimo but parcel calls self.flat; does that make sense?
benabik: apparently they are surrounded by ^$
is (<a b>, <c d>)>>[0] supposed to give <a c>? because if that's the case, then the flattening aspect of Parcel's [ ] must be given up; that's gotta explode somewhere else. 16:46
er, actually ... no 16:47
does nqp::p6listitems(obj.tree.eager) perhaps not work the way i think it does? does it flatten anyway? 16:48
16:50 grondilu joined
moritz the annoying part of regexcrosswords is that the punctuation is so small that I have trouble reading it when vertical 16:51
TimToady perhaps the underlying problem is that things like .flat are defined operationally when in fact they should be define HOPishly
timotimo "higher order programmingishly"? 16:52
TimToady yes, so all you have to do is look at the 'is nodal' of the flattening meta-operator 16:53
seems to me that flattening implies nodal 16:54
and you don't even have to look at what it is you're flattening
pmurias moritz: it's calculated from the mro, by the use of the identity function ;)
TimToady but if the flat is buried operationally in the other thing, rather than being a higher-order function, you can't just look a the flat op, but have to distribute that info elsewhere 16:55
benabik It would be better if the vertical clues were rotated 45° instead of being fully vertical
timotimo benabik: my thoughts exactly 16:56
timotimo doesn't like running into walls like in the flat case now 16:58
that doesn't seem quite like something i myself could try fixing just like that
17:00 berekuk joined
TimToady well, I wouldn't worry too much about Parcels, if you can get the other bits to work 17:03
timotimo right. it seems like when i write [<a b c>, <d e f>]>>[0] it will see that a nodal hyper is needed and knows not to descend
and then it calls [0] on a, b, c, d, e and f 17:04
TimToady in the normal case, it probably doesn't matter whether the flat is hidden or not, since the thing that flat is applied to will also be nodal
dalek pan style="color: #395be5">perl6-examples: 32f57e6 | (Lucien Grondin)++ | parsers/Newick.pm:
style
timotimo i think it has to do with: $items := nqp::p6listitems(obj.flat.eager) (which was there before), which i changed to depend on $descend and have $items := nqp::p6listitems(obj.eager) instead
oh, eager flattens! 17:05
and .tree.eager and .lol.eager all flatten, too 17:06
oh
[<...>, <...>] is actually syntax for something else isn't it >_<
that was pretty darn dumb.
r: say [<a b>.item, <c d>.item] 17:07
camelia rakudo 5cab66: OUTPUT«a b c d␤»
timotimo r: say [<a b>.item, <c d>.item].perl
camelia rakudo 5cab66: OUTPUT«[$("a", "b"), $("c", "d")]␤» 17:08
timotimo that's what it should have been
yays, it actually works! 17:10
17:12 zby_home_ joined 17:13 census_ joined
timotimo i may end up making a pullrequest this evening, maybe tomorrow. 17:16
commute&
oh, next tram in 30 minutes, guess i'll code some more 17:17
17:19 birdwindupbird left
timotimo TimToady: should the specs include a list of nodal methods and post/prefix operators? i'll probably build a bunch of spectests that ensure each of those that I (we) think should be nodal actually is 17:24
17:25 mstefanko left 17:30 ajr joined, ajr is now known as Guest16306, Guest16306 is now known as ajr_ 17:31 census_ left 17:36 jaldhar_ left, jaldhar_ joined
benabik I find the crossword to be slightly easier to read after using the element inspector to set the clues to be non-bold monospace. 17:49
lelf I find it to be a lot easier with pen & paper 17:52
... maybe cause I dislike to tilt my head 17:54
benabik has been tilting his laptop.
lelf but... but.. then you have to tilt your hands 17:56
timotimo does it seem sensible to have a S03-operator/hyper-nodal.t? 18:04
dalek p: 83e79a3 | (Pawel Murias)++ | t/nqp/71-setboolspec.t:
Add a test for setboolspec.
 18:07
p: c39ed26 | (Pawel Murias)++ | t/nqp/ (2 files):
Test using the metamodel at runtime more. Expand test 65 and test 72 which tests using the metamodel for inspecting and applying roles at runtime.
18:09 lob joined, lob is now known as dodolob
timotimo hm. nodal things: elems, uniq... what about pick, roll? reverse? sort? values? keys? kv? pairs? 18:10
that all seems to be nodal in some way
dalek kudo-js: 67218aa | (Pawel Murias)++ | / (7 files):
Implement MODE_CALL_METHOD for setbootspec. nqp::rebless, fixed getting the method name, fixed mixing in roles and a couple of other small things.

All that to get QAST::Node.set_compile_time_value to works.		 18:11
timotimo .min and .max seem nodal, too 18:15
and .shift and .pop, too :|
i can't just go around and throw an "is nodal" on almost all of the methods of Any and List and Hash now can i? 18:16
and also: how about derived classes? a user-derived class would then have to plop an is nodal to their reimplementation? maybe that is sensible to ask of them, though.
sorear o/ 18:21
timotimo "as well as various methods that assume they are already working something Positional or Associative", that does sound like i can just go aronud slapping "is nodal" on everything and their dog
tadzik I keep reading that as "is nadal", which in Polish means "is still" 18:22
like "it's still not working :("
timotimo hehe
TimToady hang in there a sec 18:26
timotimo prepares for sage advice and wisdom about the nodality issue to trickle down from TimToady 18:28
dalek ecs: 948f48a | larry++ | S32-setting-library/Containers.pod:
Don't use Nil to mean () anymore
 18:30
ecs: 674a76b | larry++ | S03-operators.pod:
clarify what is nodal
timotimo hum. okay. 18:31
that clarifies things
TimToady that's the general principle of the thing, which is more extensible than a list
the first conjecture there would save a whole lot of 'is nodal' declarations 18:32
but we can't just declare Any to be Iterable...
timotimo that's true. but i'm not sure how to implement the check properly 18:34
the conjecture about "methods declared in a class that is iterable" 18:36
TimToady probably need jnthn++ to help with that; I see something like 'is Iterable' sets a dynamic variable in the compiler $*NODAL or some such, and maybe there's a 'use nodal' pragma that can also set it for the rest of the statements in a class 18:37
unless we want .ACCEPTS declared in Any to also be nodal
benabik `role Iterable is nodal` perhaps?
TimToady can argue that both ways 18:38
benabik: maybe, jnthn can probably suggest the best way to go about that
timotimo and then i'll have to go through the classes and roles that "is nodal" and change all the methods in it?
benabik (as opposed to just making Iterable magical in some way) 18:39
TimToady maybe we just do the 'is Iterable' thing, and mark every method in Any as nodal
timotimo "mark every method in any as nodal" by hand? ;)
(by hand == with a macro)
TimToady the point is not to put 'is nodal' all over 18:40
timotimo right.
TimToady you'd only need it in a class like Any that isn't Iterable 18:41
and only the proto's location should matter
timotimo textual location, yes?
yeah, of course
TimToady I'm assuming that protos are not autogenerated in a class higher than they ought to be here 18:42
but that they're generated in the highest class that actually has a multi
if that is a bad assumption, proto location won't work 18:43
but I think protos should be generated no higher in the hierarchy than they need to be
18:44 dodolob left 18:45 preflex left
dalek p: 48d3bde | (Solomon Foster)++ | src/vm/ (3 files):
Add pow_n operation.

This adds a test for raising 1.0 to a power, which allows 1.0 ** Inf to work.		 18:46
p: 260c3b5 | (Solomon Foster)++ | src/vm/ (3 files):
Add mod_n that fits NQP definition.
p: b625b30 | (Solomon Foster)++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/Ops.java:
Make div_I conform to p6 spec for infix<div>.
18:46 colomon joined
TimToady r: class Foo { multi method bar() { say "!!!" } }; Any.bar 18:46
camelia rakudo 5cab66: OUTPUT«No such method 'bar' for invocant of type 'Any'␤ in block at /tmp/RG5jktP1Q1:1␤␤»
sorear o/ colomon
TimToady looks like the assumption is good
colomon \o sorear!
18:47 preflex joined, ChanServ sets mode: +v preflex 18:48 kaare_ joined 18:51 kaare__ left
dalek kudo/nom: aa75b03 | (Solomon Foster)++ | src/core/Num.pm:
Remove JVM-specific hack.

Now that JVM's pow_n works correctly, the hack is no longer needed.		 18:52
kudo/nom: e62b1a3 | (Solomon Foster)++ | src/core/Num.pm:
Remove JVM-specific hack.

Now that JVM's mod_n works correctly, the hack is no longer needed.
kudo/nom: ee08521 | (Solomon Foster)++ | src/core/Real.pm:
Note spec issue on infix:<mod>.
pmurias sorear: where should the opcode be documented? 18:53
dalek ast: 705a673 | (Solomon Foster)++ | S03-operators/arith.t:
Unfudge for rakudo.jvm.
 18:54
sorear pmurias: if it needs special documentation, make a pod or markdown file in docs/. otherwise, we don't really have a place for general opcode docs, you should start one. 18:56
pmurias: note: I have no idea what opcode you're talking about. 18:57
pmurias sorry, should have said 'a opcode' 18:59
19:00 preflex left
pmurias I'm thinking of documenting them in docs/qast.markdown 19:00
what tool should I use for previewing markdown? 19:01
tadzik I use markdown_calibre
it comes with calibre, an ebook thingy
you can use Text::Markdown from panda ;) 19:02
labster tadzik: why does the new File::Find still use .IO? *hint hint*
19:02 preflex joined, ChanServ sets mode: +v preflex
tadzik labster: *hmm hmm* 19:10
labster: for, uh, clarity? :) 19:11
labster my pull request from YAPC:NA already removed them 19:12
19:13 ecocode joined
dalek rl6-roast-data: f9842f8 | coke++ | / (5 files):
today (automated commit)
 19:15
19:22 rindolf left
[Coke] slight loss on rakudo.jvm as rakudo.parrot's fudging issue is resolved; sorear++ or jnthn++ got the spec tests running again, yay. 19:26
19:28 colomon left
timotimo r: class Foo { }; say Foo.^find_method('elems') === Any.^find_method('elems') 19:33
camelia rakudo ee0852: OUTPUT«True␤»
sorear yay 19:35
r: role Foo { }; say Foo.^find_method('elems') === Foo.^find_method('elems')
camelia rakudo ee0852: OUTPUT«True␤»
benabik r: role Foo {}; role Bar { sub elems { 0 } }; say Foo.^find_method('elems') === Bar.^find_method('elems') 19:37
camelia rakudo ee0852: OUTPUT«True␤»
benabik Hmmmmm....
Oh. sub.
r: role Foo {}; role Bar { method elems { 0 } }; say Foo.^find_method('elems') === Bar.^find_method('elems')
camelia rakudo ee0852: OUTPUT«False␤»
benabik \o/
labster I finally have tuits to look at the io-spec-unix error. Hm. It looks like java returns cwd with a slash appended to the end. 19:38
GlitchMr I was reading commits log. I found github.com/rakudo/rakudo/commit/a2...0fca7d6cfd 19:40
diakopter pmurias: google chrome plugin renders .markdown files okay
GlitchMr But I was confused when (1.0 ** 3) properly returned 1.0 (not 1).
I assume that it was already fixed.
benabik GlitchMr: It's been removed again: github.com/rakudo/rakudo/commit/aa...9cb2c66435
sorear r: role Foo { method elems { 0 } }; say Foo.^find_method('elems') === Foo.^find_method('elems') 19:41
camelia rakudo ee0852: OUTPUT«True␤»
GlitchMr Makes sense.
I never understood why it needed such hack anyway. Is ** operator that slow?
(I don't think 1 ** is common case anyway)
sorear GlitchMr: pow_n was returning bogus values for that case on jvm, so it had to be worked around 19:42
GlitchMr: see github.com/perl6/nqp/commit/48d3bdef27 19:43
grondilu List.combinations is NYI?
sorear unperl6y values, anyway
grondilu n: .say for <a b c>.combinations(2)
camelia niecza v24-86-g39ab531: OUTPUT«Unhandled exception: Unable to resolve method combinations in type Parcel␤ at /tmp/PkmLgNW09p line 1 (mainline @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4579 (ANON @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4580 (module-CORE @ 576) ␤ …
grondilu r: .say for <a b c>.combinations(2)
camelia rakudo ee0852: OUTPUT«No such method 'combinations' for invocant of type 'Parcel'␤ in block at /tmp/R34S7snKyo:1␤␤»
dalek kudo-js: 6ce9794 | (Pawel Murias)++ | / (4 files):
Implement nqp::findmethod. Test QAST::Node.{flat,named}.
 19:44
kudo-js: 9af914f | (Pawel Murias)++ | t/qast-node.t:
Test QAST::Node.unique.
 19:49
19:56 SamuraiJack left 20:00 Khisanth joined
pmurias diakopter: Markdown Preview? 20:03
diakopter pmurias: maybe 20:04
I don't remember
20:08 Stygia joined
pmurias stackedit 20:08
jnthn back-ish 20:09
diakopter pmurias: neat
not quite github dialect, but close enough I'm sure
tadzik labster: oh, indeed. 20:11
20:11 berekuk left
tadzik I now see that there were 2 pull requests doing this :) 20:11
I blame github interface
pmurias jnthn: have you seen my .parents question? 20:12
jnthn pmurias: Not yet... 20:13
jnthn just looks in the ir clog 20:14
20:14 berekuk joined
jnthn r: class A { }; class B is A { }; say B.^parents 20:14
camelia rakudo ee0852: OUTPUT«(A)␤»
jnthn r: class A { }; class B is A { }; say B.^parents(:all)
camelia rakudo ee0852: OUTPUT«(A) (Any) (Mu)␤»
jnthn pmurias: Those answers are correct
pmurias jnthn: in nqp
jnthn oh.
pmurias nqp: my $parent := NQPMu.HOW.parents(NQPMu)[0];say($parent.HOW.name($parent)) 20:15
camelia nqp: OUTPUT«NQPMu␤»
jnthn Well, parents should not include the class itself, no
.HOW.mro should.
pmurias I'll fix that then
writing test for nqp-js appears to be succesfull at find nqp bugs ;) 20:16
* finding
sorear nqp: say(nqp::elems(NQPMu.HOW.parents(NQPMu)))
camelia nqp: OUTPUT«1␤»
timotimo jnthn: can you suggest how to efficiently and correctly as well as simply implement "is the method given by $name implemented in a class that is marked 'Iterable'?" and/or "does the proto for this method come from Any?" 20:17
jnthn Is the first one not just "it has a method of that name and it istype Iterable"? 20:18
pmurias jnthn: to avoid rakudo breakage what should I check before merging fix-help and fix-set-invocation-spec? 20:19
20:19 konundra left
jnthn pmurias: Run a Rakudo spectest, making sure (if Rakudo has dependnet changes) to bump NQP_REVISION 20:19
timotimo jnthn: no, if the method comes from a superclass that on't work 20:20
jnthn oh.
20:21 kaare_ left
jnthn No, there's not a good way to do that. That sounds horribly like the duck typing thing I didn't figure out how to make fast ;) 20:21
[Coke] so, I have a copy of the eval server running, started by hand. if I run replace my perl6 with the eval client, and invoke ./perl6 -e 'say 3', I get a new line but no 3. ?? 20:23
jnthn timotimo: How hot path is it?
sorear [Coke]: any output on the terminal where you started the eval server? (there shouldn't be, but...) 20:25
[Coke] sorear: ope. 20:26
*n
sorear how did you "replace my perl6 with the eval client"?
[Coke] github.com/coke/perl6-roast-data/b...jvm.sh#L31 20:28
(just did those 3 lines in a fresh checkout)
and started the eval server with: ./perl6-eval-server -bind-stdin -cookie TESTCOOKIE -app perl6.jar
sorear ah 20:31
@ARGV = ("-e", "say 3")
"perl eval-client.pl TESTCOOKIE run @ARGV" = "perl eval-client.pl TESTCOOKIE run -e say 3"
arnsholt Is NQP on JVM failing two sprintf tests expected?
sorear so it's like you ran ./perl6 -e say 3 normally
the program is say(), @*ARGS = ("3") 20:32
arnsholt Git checkout from this morning plus some of my changes
sorear 06:33 < sorear> I have two failing sprintf tests. I hope that's expected
FROGGS arnsholt: yes
sorear 06:33 < FROGGS> sorear: I think so, yes
arnsholt Right, so my NativeCall hacking didn't magically break sprintf. Good =) 20:33
FROGGS arnsholt: relieved? :o)
arnsholt Moderately, let's say =D
FROGGS hehe 20:34
[Coke] sorear: so, ./perl6 -e say 3 actually works. 20:36
[Coke] just reverts ./perl6 to something he can test with. 20:37
sorear [Coke]: your problem is a lack of quoting in the perl file. 20:39
interpolating @ARGV into an exec(STRING) command means you lose the distinction between strings 20:40
20:40 konundra joined 20:41 ecocode` joined 20:42 ecocode` left, ecocode left, ecocode` joined
[Coke] Yes. didn't feel like fighting with it. 20:47
arnsholt Heh. I like how the commit dates on my branch aren't monotonically increasing anymore, after a bit of rebasing 20:50
[Coke] sorear: thanks for diagnosing, though. will save me time later when I feel like fixing it. 20:52
argh. lost my fix to t/fudgeandrun...
FROGGS :/ we really need a gitfs 20:54
20:55 ajr_ left
dalek kudo/nom: bce1048 | coke++ | t/fudgeandrun:
allow script to work on parrot & jvm
 21:03
ast: 15c110c | coke++ | S06-multi/type-based.t:
fudge for rakudo.jvm
 21:10
pmurias jnthn: any ideas/hints what could cause "No code ref at index 2" when compiling pastie.org/8138308 with pastie.org/8138304
jnthn we still have spew in nqp? :) 21:13
pmurias github.com/perl6/nqp/blob/master/s...IO.nqp#L38 21:15
jnthn hah
Not much idea aside from something to do with the clone callback stuff perhaps...
Givne it's a role 21:16
pmurias I'll saturate nqp with say statements and look some more ;)
21:18 cooper left 21:20 zby_home_ left
jnthn pmurias: It's almost certainly something to do with closure handling 21:21
pmurias it's some wird interaction between backends, as not emitting javascript got rid of this 21:25
what is the compiling SC stack? 21:26
21:27 snoopy left
jnthn When we start compiling someting we push the SC onto the stack 21:28
it gets popped in QAST::Compiler, or QAST::CompilerJAST, etc.
pmurias that was the source of the error 21:30
so if I compile something on two backends when should this get popped?
timotimo jnthn: it'd be either in hyper(\op, \obj) or in dispatch:<hyper>, probably only in the latter
so it's a kind of warm path
21:31 berekuk left
timotimo the nodal-or-not decision could possibly be compile-time evaluated. 21:33
huh. no probably not 21:35
pmurias jnthn: the SC stack is used just for determining the current SC, for repossesion purposes?
jnthn pmurias: yes; it should be popped at the point of SC serialization, I think, but I don't know the exactly timing once you reach the QAST -> backend stage matters terribly 21:36
pmurias jnthn: and it's used for what I think? 21:37
jnthn repo, yes 21:38
we scwbObject or so in Ops.java on the JVM backend, for example 21:39
timotimo: OK, will have to ponder it when I'm not drinking a tripel... :)
timotimo aah, don't worry about it; i'm having some rum right now ;) 21:40
alternatively, it may make more sense to just make something that'll automatically set "is nodal" on everything that's defined in a class 21:41
less searching that way; and it'll mostly be at compile-time
(and it would require no more changes in the hyper code i have)
i should probably push to a branch and put it up for a little bit of review 21:42
grondilu rn: my @a = <foo bar>; say join ":", $@a 21:46
camelia rakudo bce104, niecza v24-86-g39ab531: OUTPUT«foo bar␤»
grondilu maybe the fact that we don't actually need parenthesis should be mentioned in S07:100-111 21:47
synopsebot Link: perlcabal.org/syn/S07.html#line_100
grondilu hates parenthesis and he's happy whenever he can not use them 21:48
FROGGS nqp@parrot HEAD is broken? 21:51
error:imcc:syntax error, unexpected NREG, expecting '(' ('$N5001')
lelf r: say 'ababa' ~~ m:ex/aba/ 21:53
camelia rakudo bce104: OUTPUT«False␤»
timotimo is there a shorter thing than "git checkout -b things_i_have_done && git push -f . origin/nom:nom" for when i commited my changes on nom but want them on a new branch? 21:55
benabik (on nom) git branch new ; git reset @{u} ; git check new ? 21:57
Maybe not shorter. :-/
git config alias.oops '!sh -c "git branch $1 && git reset @{u} && git checkout $1"' ? 21:58
timotimo what is @{u}? 22:00
benabik Upstream
timotimo ah, cool 22:01
github.com/timo/rakudo/compare/nodal - this is what i've done so far and it apparently works 22:02
i ought to make some more test cases, though.
22:03 berekuk joined
grondilu reads 'Perl 6 changes - 2013W28' and have no clue what 'Promise' is :/ 22:06
grondilu found nothing relelvant in the spec
jnthn grondilu: I'll write some stuff about that when I get chance 22:11
22:15 spider-mario left
TimToady "As Nil is (), it shouldn’t affect code returning Nil." This is incorrect. Nil is not supposed to return (). 22:20
FROGGS this is the commit which breaks nqp@parrot: 260c3b5f00c50ca0595ac3a1392b13faf9c386bc 22:21
timotimo so the _n is wrong? 22:26
FROGGS looks like 22:27
22:27 berekuk left
benabik grondilu: en.wikipedia.org/wiki/Promise_(programming) ? 22:27
22:30 berekuk joined
grondilu benabik: yep I eventually figured it out. I didn't know about this. 22:35
not related at all: is there any plan to have fmt support large integers? I mean, for instance to print a big integer in binary: say (2^80).fmt("%b") 22:37
r: say .fmt("%040b") given 2**39
camelia rakudo bce104: OUTPUT«0000000000000000000000000000000000000000␤»
grondilu r: say .fmt("%032b") given 2**31 22:38
camelia rakudo bce104: OUTPUT«10000000000000000000000000000000␤»
grondilu see?
FROGGS r: say sprintf("%#b", 2**31)
camelia rakudo bce104: OUTPUT«0b10000000000000000000000000000000␤»
FROGGS nqp-jvm: say(sprintf("%#b", 2**31))
camelia nqp-jvm: OUTPUT«Lexical '&sprintf' not found␤ in <anon>␤ in <anon>␤ in eval␤ in evalfiles␤ in command_eval␤ in command_line␤ in MAIN␤ in <anon>␤ in <anon>»
FROGGS nqp-jvm: say(nqp::sprintf("%#b", 2**31)) 22:39
camelia nqp-jvm: OUTPUT«No registered operation handler for 'sprintf'␤ in compile_op␤ in as_jast␤ in as_jast␤ in <anon>␤ in process_args␤ in <anon>␤ in <anon>␤ in compile_op␤ in as_jast␤ in as_jast␤ in <anon>␤ in compile_all_the_stmts␤ in as_jast␤ in as_jast␤ in <anon>␤ in compi…
grondilu r: printf("%#b", 2**32)
camelia rakudo bce104: OUTPUT«0b0»
FROGGS huh
grondilu: I thought nqp@jvm does already do that, but it seems not 22:40
once this new sprintf is capable of all needed features, it will land on nqp@parrot too
grondilu notice that in order to test it, you should use 2**32, not 2**31 (on a 32-bit machine) 22:41
22:44 abnorman joined
FROGGS the machine doesnt when using bigints 22:44
22:45 raiph left
TimToady r: say (2**39).base(2) 22:45
camelia rakudo bce104: OUTPUT«1000000000000000000000000000000000000000␤»
lue I have a 64-bit machine, and 2**32 wfm (but not, predictably, 2**64)
TimToady r: say (2**65).base(2)
camelia rakudo bce104: OUTPUT«100000000000000000000000000000000000000000000000000000000000000000␤»
TimToady how hard can it be?
22:46 abnorman left 22:47 abnorman joined
grondilu ok 22:49
r: my $x = 2**100; say $x +& $x 22:50
camelia rakudo bce104: OUTPUT«1267650600228229401496703205376␤»
grondilu r: my $x = 2**100; say ($x +& $x).base(2)
camelia rakudo bce104: OUTPUT«10000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000␤»
grondilu r: my $x = 2**65; say ($x +& ($x + 1)).base(2)
camelia rakudo bce104: OUTPUT«100000000000000000000000000000000000000000000000000000000000000000␤»
22:50 Stygia left
grondilu r: my $x = 2**65; say ($x +| ($x + 1)).base(2) 22:50
camelia rakudo bce104: OUTPUT«100000000000000000000000000000000000000000000000000000000000000001␤»
grondilu oh, so bitwise logic does work with big integers. Good. 22:51
r: say +^ 2**65 # just an other check 22:52
camelia rakudo bce104: OUTPUT«-36893488147419103233␤»
grondilu r: say (+^ 2**65).base(2) # just an other check
camelia rakudo bce104: OUTPUT«-100000000000000000000000000000000000000000000000000000000000000001␤»
grondilu hum
r: say (+^ 2**10).base(2) # just an other check
camelia rakudo bce104: OUTPUT«-10000000001␤»
grondilu is confused by negative integers in bitwise logic 22:53
benabik Bit inverting a signed x is same as -x-1 due to 2s compliment. 22:54
TimToady +^ is 1's complement but the representation always pretends to be 2's complement 22:55
r: say (+^ 0)
camelia rakudo bce104: OUTPUT«-1␤»
TimToady there's the degenerate case, flipping all the bits
well, it always flips all the bits
setting all the bits, in this case 22:56
pmurias jnthn: it wasn't the stack poping
TimToady the fact that there are in infinite number of them is okay, since we can represent an infinite number of 1 bits with a -
grondilu ok. I guess I won't use ^+ as it won't suit my need, then
pmurias jnthn: serialize_sc seems to have some global state, will try to chase it down tommorow
grondilu meant +^ 22:57
TimToady grondilu: it should work exactly the same when stored into a 2's complement native (in the absence of overflow, of course)
r: say (2**65).msb 22:59
camelia rakudo bce104: OUTPUT«65␤»
TimToady but if you're just looking for the magnitude of a binary number, .msb is what you want
grondilu no I want to do bitwise logic on large number (I'm working on rosalind.info/problems/cset/) 23:00
because there are large boolean arrays stored as 000001011100000111...
(as strings of 0 and 1, that is) 23:01
TimToady if you want to limit it to a certain number of bits, and not go negative, then you'll want to xor with a large positive 2n-1 number
grondilu or just do 2**MAX - x
I think that's what I'll do
TimToady but if you don't care about the number of bits actually used, or the sign of the result, unary +^ should work fine
it does, in fact do a 1's complement 23:02
grondilu but getting back the boolean array would be tricky, wouldn't it?
I mean, +^ would confuse me too much. But I guess I need to take some time to think clearly about these stuff 23:03
TimToady you might have to & with a sized 2n-1 at that point to make it positive, but all the intermediate stuff doesn't have to
jnthn pmurias: are you generating code for two backends in one process? 23:04
jnthn never did it that way :)
pmurias jnthn: yes
jnthn k :)
dalek kudo/nom: 6c80bf5 | (Solomon Foster)++ | src/core/Rational.pm:
Fix broken div uses.

Rational.pm was written to work with NQP Parrot's broken infix:<div>. Since NQP JVM now has a properly working infix:<div>, a bunch of these functions started failing. This corrects floor, ceiling, and Int.
pmurias jnthn: a prop is being attached to code refs to mark them as cloned 23:05
jnthn pmurias: yes
pmurias: oh, hm
pmurias: I think that ops abstract that though 23:06
23:06 colomon joined
colomon turns out that getting infix:<div> working properly on jakudo (it never has on parrakudo) broke some of the basic Rat methods. So now I've fixed those, too. 23:07
pmurias abstract that? the problem is that they get cloned and put into the closure table of one backend, and on the other one are already marked
jnthn pmurias: ah
pmurias but it should be fairly easy to workaround that if that is the case
jnthn pmurias: thats entirely possible. I never really designed it with your use case in mind.
BS in general is a kind of marking thing. 23:08
pmurias BS? 23:09
FROGGS colomon: github.com/perl6/nqp/commit/260c3b...nt-3628793
jnthn: ------^ any idea about that?
jnthn pmurias: bound serialization
FROGGS anybody else has problems building nqp@parrot HEAD? 23:10
colomon I must have modified the wrong file somehow. :(
jnthn FROGGS: that looks like sokmebody confusssed a nqp:: and pir::
oh wow my typing rocks
FROGGS I can read it :o) 23:11
colomon didn't even notice
jnthn pmurias: um, bounded, not bound 23:12
FROGGS gnight, see ya in a few hours
jnthn 'night, FROGGS
23:12 FROGGS left
colomon FROGGS++ 23:12
23:17 dmol left
dalek ecs: 8f30d2d | larry++ | S03-operators.pod:
nodal ok on normal methods too, not just protos
 23:17
p: a60837b | (Solomon Foster)++ | src/vm/parrot/QAST/Operations.nqp:
Undo changes that didn't belong in NQP Parrot.
 23:18
colomon With luck that undoes the damage I did to Parrot.
o/ # back to cabin
23:18 colomon left
pmurias jnthn: I could either move to the stable sc way of doing things or we could cache the serialization somehow 23:20
TimToady r: say [].can('elems').perl 23:22
camelia rakudo 6c80bf: OUTPUT«(method elems(List : Mu *%_) { ... }, method elems(Iterable : Mu *%_) { ... }, method elems(: Mu *%_) { ... })␤»
TimToady r: say [].can('elems')[*-1].perl
camelia rakudo 6c80bf: OUTPUT«method elems(: Mu *%_) { ... }␤»
TimToady r: say [].can('elems')[*-1].WHO
camelia rakudo 6c80bf: OUTPUT«().hash␤»
TimToady r: say [].can('elems')[*-1].WHAT
camelia rakudo 6c80bf: OUTPUT«(Method)␤»
TimToady r: say [].can('elems')[*-1].WHICH 23:23
camelia rakudo 6c80bf: OUTPUT«Method|-172260614␤»
TimToady any way to tell which class that method is declared in?
jnthn TimToady: well, can always introspect the signature's first arg's type
TimToady uh, I though we didn't require the self to do typing 23:24
in fact, Any doesn't show up above, though List and Iterable do 23:25
just trying to figure out how to best to look up nodal in those cases we can intuit it 23:26
jnthn The type of the self is what makes multi-dispatch disambiguate in a subclass
Yeah...I need to read le spec more carefully and ponder it :)
TimToady well, except most of the methods in Any aren't multis
jnthn Sure, but the type is still set on the invocant 23:27
It's user-selectable though so it can lie
Though iirc S12 is like, "don't do that" :) 23:28
TimToady nodnod 23:29
jnthn I mean, any kind of custom dispatch is possible given we keep a hash of methods at each level (walk MRO and se .^method_table) but I guess we want something more efficient. 23:30
*see
If we have a typed array, are we meant to consider nodality of the individual things and respect subtyping, ooc? 23:31
.oO( is that even a sensible question... :) )
TimToady well, put 'is nodal' on every &foo that might be used in a hyper is probably best, given it requires no additional run-time lookup in the hyper
jnthn Oh, of course, it's on hte operator...
TimToady so probably the compiler needs to just have some way of setting that on all the appropriate &foo's
jnthn k
TimToady probably just a $?NODAL (or compile-time $*NODAL) that declarators can examine is an okay hack, and we just force it on when a class has 'is Iterable' in the declaration, or we see a 'use nodal;' 23:33
or somethin'
it doesn't seem like it needs to be an extensible mechanism, so a compiler hack is probably okay in this case
23:34 census_ joined
TimToady unless we can think of something elegant that doesn't violate S12 :) 23:34
in any event, there should be no run-time component to determining whether a function is nodal, other than looking up its trait directly 23:35
"has 'is Iterable' in the declaration" is misstated. shoulda said "derived from Iterable" 23:45
maybe a compose-time thing, to handle methods coming from roles correctly 23:46
pmurias jnthn: do you think it would break anything if I made serialization caching on parrot? 23:47
23:47 ecocode` left
jnthn pmurias: What do you mean by serialization caching? 23:48
pmurias if serialize_sc is called a second time on the same sc, it should return the same thing
jnthn ah 23:49
probably ok, I mean, the serialization string should come out the same both times
Prolly don't want it to be a "default" though 23:50
pmurias the serialization string might change if the object in QAST::WVal change
jnthn yeah, that doens' tnormally happen though 23:52
TimToady
.oO(constant folding, immutability, value identity...)
jnthn TimToady: I menat once it reaches the backend 23:57

Powered by Cro and the Raku® Programming Language.

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

Thank you!