»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.
retupmoca #perl6! I decided to write something in perl 6, so I attempted a port of Email::Simple at github.com/retupmoca/p6-Email-Simple 00:30
so
is there a defined way to get this into the modules list? 00:31
[Coke] retupmoca++ 00:36
sorear I think you have to talk to tadzik 00:42
TimToady plus whatever is on modules.perl6.org/ 00:43
unfortunately our most modular people are asleep in some other hemisphere... 00:45
retupmoca the only thing I see on modules.perl6.org is #perl6, and the guide says to poke a maintainer
TimToady there's a link to a guide there
sorear "a maintainer" means tadzik these days. 00:46
TimToady 3rd paragraph
sorear 0146 Warsaw time atm
TimToady which point to ttjjss.wordpress.com/2010/08/09/so-...-6-module/
retupmoca yeah, I saw that
was following it, but all it says is to poke a maintainer 00:47
who seems to be asleep
TimToady there's probably a serkit git repo that has to be tweaked 00:48
and yes, the docs appear to be a bit haphazard on that subject so far :)
retupmoca github.com/perl6/ecosystem maybe 00:49
hrm
I need a META.info
TimToady do you have a github id? 00:52
retupmoca yeah 00:52
retupmoca
TimToady hmm, github is being slow, but I think I can add you 00:54
retupmoca otherwise I could do a fork / pull request 00:55
TimToady github has been suffering from DoS attacks lately, so maybe that's why it's being sluggish 00:56
retupmoca yeah, it's been sporadically slow
TimToady okay, I think I did it 00:57
you can now destroy^Wcontribute to any of the Perl 6 projects, welcome aboard
retupmoca github says read+write access
yay!
TimToady++
TimToady as we say, have the appropriate amount of fun, which may from time to time be negative, but hopefully mostly positive 00:59
dalek osystem: ffd609f | (Andrew Egeler)++ | META.list:
Add Email::Simple
 01:00
TimToady retupmoca++ 01:02
retupmoca ok, so it should show up next time modules.perl6.org updates then 01:03
TimToady \o/
retupmoca hrm, now that I have this access, I'll have to write more modules 01:04
TimToady or tests, or docs, or webpages, or... :)
retupmoca yes indeed
TimToady we could also use more porting of modules back and forth between rakudo and niecza 01:05
diakopter mmm .meth 01:06
TimToady careful, I don't want to end up on some proscribed list for buying decongestants :) 01:07
I have a trick sinus that doesn't like to go up in airplanes 01:09
well, actually, it's fine going up, but coming back down sometimes feels like an icepick in the forehead
TimToady is looking forward to flying on the more-pressurized 787's to see if that helps 01:10
but maybe I shouldn't mention airplanes either... :) 01:11
colomon ==> Testing Email::Simple 03:14
t/basic.t .. ===SORRY!===
Can only provide exactly one initial value to a mixin at /Users/colomon/.panda-niecza/src/Email__Simple/lib/Email/Simple/Header.pm6 line 90:
------> return @values but @values[0]⏏;
I don't think niecza likes the @values but @values[0] construct. 03:20
sorear @values but @values[0] eh... 03:26
o/ colomon 03:27
colomon \o
does that make sense, even?
sorear I'm not sure 03:28
it's not entirely clear what the semantics of the implied @values.clone would be
(the error message, OTOH, is clearly bogus) 03:29
colomon oh, I thought it was complaining there were too many values (ie more than one) in @values 03:32
colomon bedtime 03:47
TimToady o/ 03:48
retupmoca ok, Email::Simple doesn't use @values but @values[0] anymore 04:34
if someone wants to test it on niecza
I would, but I don't have it installed yet
kcw I'm not sure if I've found a bug with gather/take or I'm doing something wrong... 04:44
> my @spam = gather loop { take 'eggs'; last }
Method 'eager' not found for invocant of class 'Integer' 04:45
in rakudo, that is
TimToady perl6: my @spam = gather loop { take 'eggs'; last }; say @spam 04:46
p6eval rakudo bd5adb: OUTPUT«Method 'eager' not found for invocant of class 'Integer'␤ in block <anon> at /tmp/B6jpxyUQ0v:1␤␤»
..pugs b927740, niecza v14-42-g1d02da9: OUTPUT«eggs␤»
TimToady I'd say you're right
TimToady the eager likely comes from the assignment, but I've no idea where the Integer comes from--looks like a Parrot type 04:47
kcw k. I'll submit it to perlbug. Glad to see I'm not just misunderstanding something; I've been trying to figure out why this doesn't work for a while. 04:48
TimToady nom: my @spam = gather while 1 { take 'eggs'; last }; say @spam 04:49
p6eval nom bd5adb: OUTPUT«eggs␤»
TimToady hmm
nom: my @spam = gather loop (;1;) { take 'eggs'; last }; say @spam
p6eval nom bd5adb: OUTPUT«eggs␤»
TimToady seems like there's a hint
nom: my @spam = gather loop (;;) { take 'eggs'; last }; say @spam 04:50
p6eval nom bd5adb: OUTPUT«Method 'eager' not found for invocant of class 'Integer'␤ in block <anon> at /tmp/7pxw9NGc3l:1␤␤»
TimToady I'd say the default condition is a Parrot type rather than a Perl type
sorear btw, kcw, welcome 04:51
is nbrown one of us? 04:52
Teratogen I'm working on a book 08:20
"Parsing HTML with Regular Expressions for Fun and Profit"
sorear there's nothing wrong with being a fan of Gabriel. 08:23
troll failed. :D
Teratogen AND IIIIIEIIIIIIIEIIIIII WILL ALWAYS LOVE YOOOUUUEUUUUUEUUUU 08:27
gah
can't get that song out of my head 08:28
masak me neither. 09:12
good #perl6, morning.
sorear masak morning. good. 09:14
.oO( forwhy is masak mo(u)rning?)
 09:15
jnthn becomes vaguely concious-ish 09:17
sorear o/ jnthn
jnthn o/ 09:18
jnthn sees somebody found a code-gen bug
.oO( the masakbot is slow to see the word "bug" today... )
 09:21
masak heh :)
I haven't backlogged yet. I'll get to it.
retupmoca: welcome! retupmoca++ 09:23
nom: my @spam = gather loop { take 'eggs'; last }; say @spam 09:26
p6eval nom bd5adb: OUTPUT«Method 'eager' not found for invocant of class 'Integer'␤ in block <anon> at /tmp/QoVvAM5Irq:1␤␤»
masak submits rakudobug
oh, submitted already by kcw++ 09:28
jnthn oh!
moritz masak: any feedback for the bt printer? 09:33
one thing I noticed is that it's fairly useless for debugging the setting 09:47
maybe I can change that without making the normal user experience worse 09:48
jnthn moritz: I guess for that we can use --ll-exception or add a --debug-setting flag or something
--setting-exception
sorear fwiw, this sort of thing is a lot of the reason I haven't touched niecza's bt printer yet 09:49
moritz jnthn: well, I think it's a fairly artificial thing to exclude the setting
sorear I don't think I can improve it without making it unusable for all the uses I haven't thought of
jnthn moritz: True
moritz jnthn: currently I exclude the setting when searching for the first "interesting" call frame 09:50
jnthn moritz: Maybe we should include it but mark various of the list guts as excluded
moritz jnthn: but if we are consequent with usage of 'is hidden_from_backtrace', then I can drop that condition
jnthn moritz: That trait doesn't bother me. 09:51
moritz sorear: well, you can just sit back and watch us experiment, and copy whatever workable thing we come up with
sorear moritz: that's what I'm doing :D 09:52
moritz jnthn: if things work out the way I am imagining right now, we don't need to special-case the list guts at all to keep them out of the vast majority of backtraces 09:53
jnthn moritz: OK :)
masak moritz: haven't been toying around with it yet; been too busy. 09:54
moritz \o/ seems to work 10:01
dalek kudo/nom: d43594e | moritz++ | src/core/Backtrace.pm:
stop special-casing the setting in the bt printer
 10:29
Teratogen now THAT, my friends, was a fricking netsplit! 11:28
masak it was. 11:31
JimmyZ \o, dyncall build works now 13:11
masak \o/
wow, END phasers in Perl 5 run both after a 'die' and after an 'exit'. 13:13
JimmyZ wants to update dyncall to v0.7, and commit some fixs to nqp 13:16
jnthn JimmyZ: +1 13:17
jnthn (on the 0.7 upgrade) 13:18
I just hadn't got there yet
oh but...
I think we have some additional patches on top of what 0.7 has
JimmyZ jnthn: but that also needs to make some changes to dyncall makefile
jnthn JimmyZ: Have you submitted those changes upstream to the dyncall folks?
moritz you should upgrade dyncall by applying the diff between the old and the new version 13:19
not by overwriting stuff
JimmyZ jnthn: well, dyncall.org doesn't seem to be open enough, I should email the patch to authors ? 13:20
jnthn JimmyZ: Yes, they've responded to emails from me before.
They seem quite friendly and happy to accept patches/input :)
JimmyZ it's odd that 'as' can't compile .S file 13:21
moritz gist.github.com/1826725 # my experiments with ObjHash
binding works, but assignment doesn't. Seems I've misunderstood $!whence. 13:22
jnthn moritz: Two things. 13:30
moritz: whence runs before the actual storage is done. So $v won't have its value there
moritz: Second, you need to bind it into %!keys and %!values 13:31
moritz: Otherwise you'll re-containerize.
moritz: gist.github.com/1826773 works for me
moritz jnthn: cool, thanks 13:32
ah right, that's the difference between Proxy and whence 13:33
jnthn moritz: Also one of efficiency. 13:35
(the whence mechanism is *way* faster)
masak whereas the Proxy one is more flexible, I guess? 13:36
jnthn masak: Yes.
masak: They're kinda apples and oranges in a way.
masak: You could build the stuff we do with $!whence on top of Proxy.
masak .oO( apples are way faster, whereas oranges can be peeled in countless ways )
jnthn masak: But it'd look quite different. 13:37
:P
masak nom: sub fp($x, &c) { c($x) === $x ?? $x !! fp(c($x), &c) }; for ^10 { repeat while fp(my $s, *.subst("[]", "")) { $s = "[]".comb.roll(10).join }; say $s } 13:46
p6eval nom bd5adb: OUTPUT«[[]][][][]␤[[]][[]][]␤[][[]][][]␤[][][[[]]]␤[[[[[]]]]]␤[][][[[]]]␤[[[[[]]]]]␤[[[]]][][]␤[[[[]]][]]␤[][][[]][]␤» 13:47
masak <3 Perl 6
nom: sub fp($x, &c) { c($x) === $x ?? $x !! fp(c($x), &c) }; sub shuffle { @^list.pick(*) }; for ^10 { repeat while fp(my $s, *.subst("[]", "")) { $s = "[[[[[]]]]]".comb.&shuffle.join }; say $s } 13:49
p6eval nom bd5adb: OUTPUT«[][[[[]]]]␤[[]][[[]]]␤[[]][[[]]]␤[[][][]][]␤[[][][]][]␤[][[[][]]]␤[[[][[]]]]␤[][][[]][]␤[][[]][][]␤[[][[]]][]␤»
masak Whirlpool. domainlanguage.com/ddd/whirlpool/ 14:00
masak timotimo: the goal is to generate any such fixed-length balanced string with equal probability. 14:31
masak timotimo: it's been up for discussion before. TimToady++ w/ family figgered out how to do it using an auxiliary table. 14:31
timotimo i see 14:32
in that case i believe your solution works pretty well 14:33
masak my version above "fails" in that it has to reject a certain percentage of candidates. the table approach always succeeds but is longer.
I can't help thinking that there is a short, simple solution out there.
timotimo mhm
[Coke] .u *face* 14:52
phenny ValueError: Broken regexp: '*FACE*' (file "/home/sbp/Dropbox/inamidst.com/phenny/modules/codepoints.py", line 58, in codepoint_extended)
[Coke] ZOMG, PHENNY IS MADE OF PYTHON!
jnthn .u fa.* 14:53
phenny U+0618 ARABIC SMALL FATHA (◌ؘ)
U+063D ARABIC LETTER FARSI YEH WITH INVERTED V (ؽ)
U+063E ARABIC LETTER FARSI YEH WITH TWO DOTS ABOVE (ؾ) [...]
masak .u (((((((((((.*)*)*)*)*)*)*)*)*)*)*)* 14:58
phenny masak: Sorry, your input is too long!
masak :)
timotimo heh 15:00
moritz masak: also it's pretty pointless; only regexes that fail to match actually backtrac 15:07
k
masak troo 15:08
I'm not too good at being malicious.
jnthn phenny: .u (\w* \d* \s* XX)+
That one's more...oops. :)
dammit, I didn't mean to actually send it! 15:09
moritz perlgeek.de/blog-en/perl-tips/in-se...egexp.html
jnthn That one is straight out of my C# regexes course...
Well. Regexes for C# programmers.
masak jnthn: I think it's harmless if you write 'phenny: ' as you did. 15:11
benabik phenny: .u t
.u t
phenny U+0074 LATIN SMALL LETTER T (t)
masak right.
jnthn masak: er, I didn't cyrillic it this time :P
But I'm happy phenny is still alove. :)
*alive
masak phenny: en ru "I'm doing science and I'm still alive."? 15:12
phenny masak: "Я делаю науку , и я все еще жив." (en to ru, translate.google.com)
jnthn oh, Russian...thanks for the reminder... 15:14
jnthn afk :)
masak hee hee 15:15
perl6: say "Happy Valentine's \x2665" 15:17
p6eval pugs b927740: OUTPUT«Happy Valentine's â¥␤»
..rakudo bd5adb, niecza v14-42-g1d02da9: OUTPUT«Happy Valentine's ♥␤»
masak there's something the matter with the way Pugs output gets presented. 15:20
moritz aye 15:21
moritz I promise to investigate whem pugs is again actively developed 15:23
masak deal :)
setting up an environment on feather where Pugs can be actively developed will be good for everyone. 15:24
moritz feather has fairly recent ghc and cabal-install packages 15:26
not sure what else we need
I also have no ideas which pugs repo is the most recent 15:27
masak github.com/audreyt/Pugs.hs , most likely. 15:28
activity 11 days ago. 15:29
masak clones that one into his home directory 15:30
oh, but the hacking file says to use github.com/perl6/Pugs.hs
which has an identical HEAD. 15:31
moritz so it's already synchronized?
masak guessing au-chan does stuff in her repo, then pushes to the perl6 one. 15:34
so perl6/Pugs.hs is the "upstreamest" one.
moritz
.oO( the ups and downs in the life of a stream ... )
 15:36
masak .oO( The Upstremiest Repo That Ever Lived ) 15:38
Upstreamiest*
moritz ok, successfully built a Pugs.hs from source 15:42
on feather, that is 15:43
moritz and that Unicode problem is also present on the command line 15:43
so it's not an evalbot thing
[Coke] moritz: does every user have their own cabal ... downloads? 15:46
timotimo there is no cabal! 15:46
[Coke] guesses so, as he's now installing packages with cabal. 15:48
moritz [Coke]: yes 15:50
TimToady std: Happy ♥'s Day 15:51
p6eval std 52f3895: OUTPUT«===SORRY!===␤Confused at /tmp/i1su4d3F_Y line 1:␤------> Happy ⏏♥'s Day␤ expecting any of:␤ bracketed infix␤ infix or meta-infix␤ statement modifier loop␤Undeclared name:␤ 'Happy' used at line 1␤Parse failed␤FAILED 00:01
..109m␤»…
masak \♥/ 15:53
[Coke] gist.github.com/1827728 - pugs build failure on feather 15:55
colomon Conversation just had here: Jen: "You should thank TimToady not to give you an STD." colomon: "std is standard, dear!" Jen: "STDs are *not* standard!" 15:56
moritz :-)
masak [Coke]: that's odd. what's the GHC version? 16:00
[Coke] looks like it might be 7.4.1 16:01
"whatever is installed on feather". ;)
masak yeah, getting 7.4.1 on feather too. 16:02
and the last thing that happened to Pugs.hs was that it got updated for 7.4.1 compatibility. 16:03
which indicates that it *should* work on 7.4.1 :)
moritz and it does, here 16:04
daxim ooh, opensuse packagers are slacking 16:05
moritz [Coke]: what commit is your Pugs.hs repo on?
[Coke] 08f0e5052cfccf839f852c883a913298697d146e
moritz ok, same as mine 16:06
[Coke] mine's in ~coke/sandbox/Pugs if that helps.
masak moritz: it worksforyou on feather?
moritz masak: it compiled. 16:07
masak then the more interesting question is how your setup and [Coke]'s differ.
moritz and the resulting 'pugs' binary seems to be able to evaluate basic expressions
masak wohoo! 16:08
moritz what I did is run 'cabal update', and then 'cabal install <packages>' for all the packages that 'make' complained about
masak sounds easy enough.
[Coke] moritz: did you specify the > < modifiers also?
masak I'll try it here too.
[Coke] (one of them had a < requirement.)
moritz [Coke]: nope 16:09
[Coke] cabal doesn't have a "list of installed" ?
benabik cabal list --installed 16:10
[Coke] feather.perl6.nl/~coke/cabal.txt
moritz same as mine 16:12
[Coke] O_o 16:14
moritz my karma is just better :-)
masak karma moritz 16:15
aloha moritz has karma of 2965.
masak karma [Coke]
aloha [Coke] has karma of 0.
masak that would explain it, yes.
benabik karma Coke
aloha Coke has karma of 575.
moritz [Coke]++ 16:16
karma [Coke]
aloha [Coke] has karma of 0.
moritz karma Coke
aloha Coke has karma of 575.
moritz wtf?
PerlJam Coke++
[Coke] That's fine.
PerlJam Pepsi--
:)
[Coke] I'm not [Coke] anyway. it's just a stupid wrapper I have to wear on freenode. 16:17
masak it's the adapter pattern. 16:18
masak I followed moritz' instructions. I'm getting [Coke]'s error. :( 16:21
karma masak
aloha masak has karma of 629.
benabik make IGNORE_KARMA=Yes 16:23
[Coke] I'm using [email@hidden.address] 16:34
masak too 16:38
moritz is using git://github.com/perl6/Pugs.hs.git 16:39
I can't imagine the difference matters though, since it's the same HEAD.
benabik Those should be the same repo (although the git:// URL is read-only)
masak right. 16:40
moritz I don't have any suspicious environment variables either 16:53
masak curiouser and curiouser, thought Alice. 16:54
masak decommutes 16:58
[Coke] tries to figure out how to call excel from visual basic from java on windows. 16:59
[Coke] gets a little dizzy. 16:59
[Coke] # 02/14/2012 - niecza at 100% 17:57
"niecza", 20055, 8, 751, 1573, 22387, 23233
"rakudo", 20055, 28, 609, 1863, 22555, 23243
wow. ;)
TimToady okay, here's a first whack at my rethink of acceptance types that I promised 18:09
gist.github.com/1828667
commuting &
moritz also it feels slightly wrong to base the presence of the REALLY method on the presence of the payload 18:37
it means you need to go rather meta jut for instantiating very basic stuff
isBEKaml masak: I got pugs setup on my machine with GHC 6.12.3, a few months back. :) 18:38
moritz also 18:39
moritz the proposal tries to address the matter of having too many acception/rejection types, but has just as many in the result 18:40
if you count OK/OK+ and Nil/Nil+ as separate types, which it seems you should, because they behave differently
well, Nil/Nil+ behave differently, OK/OK+ don't do so much 18:41
so 6 -> 5
isBEKaml perl6: my $x = 42 but False; say "blah" if $x; # nom ? 18:46
p6eval rakudo bd5adb: OUTPUT«blah␤»
..niecza v14-42-g1d02da9: ( no output )
..pugs b927740: OUTPUT«*** Cannot cast from VBool False to VCode (VCode)␤ at Prelude.pm line 541, column 5-16␤»
isBEKaml I was confused for a moment when I tried this on nom until I saw niecza. :) 18:47
BTW, is that right what rakudo does? known bug, I presume?
jnthn nom: my $x = 42 but False; say $x.WHAT; 18:48
p6eval nom bd5adb: OUTPUT«Int+{<anon>}()␤»
jnthn nom: my $x = 42 but False; say $x.^methods(:local); 18:48
p6eval nom bd5adb: OUTPUT«Bool␤» 18:48
jnthn nom: my $x = 42 but False; say $x.Bool 18:49
p6eval nom bd5adb: OUTPUT«False␤»
moritz I guess that's because Int has a separate boolification protocol?
jnthn moritz: OH
moritz++
benabik nom: my $x = 42 but False; say ?$x
p6eval nom bd5adb: OUTPUT«False␤»
isBEKaml Oh, I have to ?$x when doing it in bool context?
jnthn isBEKaml: No, we just need to know when to stop cheating. :)
moritz one more thing about the proposal... it feels wrong to $regex_match<foo> die in sink context if it has failure information 18:49
isBEKaml: it's a rakudobug 18:50
benabik isBEKaml: ?$x is another way to explicitly ask for the bool.
moritz std: :72<foo>
p6eval std 52f3895: OUTPUT«ok 00:01 107m␤»
jnthn ouch, I think I missed the thingy that's being discussed at the moment due to network failz...
jnthn hits the ir clogs
isBEKaml jnthn: Yeah, I was thinking: "Hey I know this is meant to be used as a bool when I said but False. ?$x is just overdoing it." :)
moritz jnthn: gist.github.com/1828667
jnthn isBEKaml: Yeah, compiling the "obvious patch" for the issue now to see if it works. 18:53
isBEKaml jnthn++ # great!
isBEKaml is a bit sad that he can't try nom/rakudo anymore on his machine as the build requirements are far out of reach for his poor old machine. 18:54
isBEKaml
.oO(talk about sleepy times and verbosity)
 18:55
benabik isBEKaml: That may get better when jnthn finishes his serialization work (*knocks on wood*) 18:56
jnthn It almost certainly will. 18:57
TimToady jnthn: the OKness proposal is basically trying to name your primitive cheating mechanism :)
jnthn TimToady: I did a o.O as soon as it started talking about roles being inside or out of Any. ;)
TimToady where they instantiate to, really 18:58
jnthn (Generally they only pretend to be in any of these so we can pass them around first class)
TimToady: OK
benabik
.oO( role OK is instantiates(Mu) {...} )
 18:59
jnthn Dang my patch fai...oh...that's because I didn't build the new nqp with op moritz++ added :)
jnthn Failure and Match will both go away as insufficiently general. 19:01
==> Match stays in the sense of it being the Match type object though?
It's just not what we return on a failed Match?
isBEKaml benabik,jnthn: that's great! # hopeful
jnthn A successful match now returns OK instantiated with match info, if any. 19:02
==> Are we talking about "top level" matches here? Or all the way down?
TimToady gist.github.com/1828667 has tweaks from some of your suggestions 19:03
jnthn To first approximation, OK/Nil force only truth, definedness, and .REALLY; 19:04
==> we're talking .defined level here, not .DEFINITE level?
TimToady well, I think they track in this case, offhand 19:05
but OKness is necessarily on a primitive level, so carries .DEFINITE as well as primitive boolean
jnthn .DEFINITE is *really* low level in 6model 19:06
TimToady only question is whether .defined delegates, and I don't think it does
jnthn Like, "checks a bit somewhere" level
Like, "if you screw this up we segfault" level.
TimToady I'm hoping with a good enough definition of OKness we can get away from juggling circularity saws, at least on booleans and defineds 19:07
jnthn It's really the "is this a type object" flag.
Well, we added .DEFINITE distinct from .defined to be a circularity saw.
TimToady is there a name for primitive boolean context? 19:08
jnthn TimToady: No. (more)
TimToady: I've considered that "primitive" thing as purely an optimization so far.
It's leaking out in the "but True" example you showed in just a straight "getting caught cheating" thing. 19:09
I can imagine a primitive boolean context style operation.
But my primary goal was optimization at the low level, rather than different semantics at the high level.
TimToady sure, just trying to clean up the success/failure semantics to make it easier, and perhaps make it more obvious when we're cheating on purpose 19:10
jnthn *nod*
TimToady KEEP/UNDO probably simplifies a bit under OKness too 19:11
jnthn Yeah, I'm happy to clear things up. Just need to get my head around the details. :)
TimToady I probably need some rows in my table for things like type objects, and maybe s'more columns
well, except type objects are largely outside of OKness, so they aren't treated as automatic success/failure by smartmatch 19:13
jnthn TimToady: Is there a particular reason OK and Nil are themselves roles?
TimToady: That'd imply you could mix them in to things.
jnthn If you do that, what are the semantics? 19:13
TimToady was originally thinking of instantiation rather than delegation, but I don't think you want to be mixing them into other things; more the other way around, since they're primitive 19:14
and "delegation" might not be high-level delegation either
jnthn Delegation tends to imply you have an instance. :) 19:15
TimToady "mixin" likely means something different than high-level mixin
jnthn (which holds the attribute to delegate to)
TimToady this is more just trying name the primitives we're already using anonymously, along with trying to straighten out the user-visible namespace 19:16
[Coke] woot. I think I'm going to YAPC! 19:17
TimToady niecza: say <a b c> »~~» /b/
p6eval niecza v14-42-g1d02da9: OUTPUT«Match() #<match from(0) to(1) text(b) pos([].list) named({}.hash)> Match()␤»
TimToady I'm trying to get that to come out: Nil OK[stuff] Nil
skids Slightly OT: Is rakudo's <{fail}> in regexp formally considered a cheat? It does what I consider to be the right thing and propagates something that explodes down to the root of the returned result, but that's despite the fact that it is tested for truthfulness during the regexp match, which should mark it as "handled" (?)
TimToady skids: the interaction of normal failure with backtracking is not yet quite nailed down; it might be that only specific kinds of failure induce specific kinds of backtracking 19:19
but all such failures would be some variant of Nil in the new model 19:20
jnthn Not to mention that the regex engine just does the cheapest thing it can do figure out if the match was a success :-)
*.does(OK) & none *.can('REALLY') 19:21
When would an OK not be able to REALLY?
TimToady meta-talk really
I'm talking about if the "delegate" slot is empty
jnthn ok
Ah, that's what the "payload" column means
TimToady and perhaps not even allocated
i.e. wouldn't need a delegate pointer if the 'delegate exists' bit was false 19:22
jnthn "Types outside of OKness must call .Bool to determine truthiness" 19:22
Actually the boolification cheating I have is to avoid having to call a method when we can very quickly check for zero, which is the case for, say, an Int or Num. 19:23
TimToady trying to nail down when you can just look at the bit, and when you have to ask the object nicely
we might find some way of bending that part 19:24
jnthn OK, but in some cases I've wanted to cheat for things that are more than "look at the bit" :)
skids jnthn: RE delegation, well there is the NYI "method foo () handles" which could return a class expecting calls to the class instead of an instance... 19:24
TimToady it's difficult to know when you're breaking Liskov on that though 19:25
jnthn TimToady: Yeah, that's what the "but True" bug just caught us doing :) 19:25
TimToady I don't know whether this is quite sane yet, but I think this is a path toward sanity 19:26
jnthn TimToady: Yes, there's much to like in here. And much to ponder.. 19:28
Match ~~ OK[Match] # ??
OK[Match] ~~ Match # ??
(using [...] in a handwavey sense here) 19:29
oh
The latter can't be true
H
Hmm
Well
Nil[Match] ~~ Match had best be false :)
TimToady Types outside of OKness must call .Bool to determine truthiness (though native types and known-non-derived types can generally optimize this away). 19:32
Along with Bool, OKness types may not be used literally in smartmatching, and will warn when such usage is parsed.
jnthn Yeah, known non-derived is what blew things here.
jnthn TimToady: My questions above were not in the smart matching sense, but in the type checking sense 19:33
TimToady that level may also require cheats
jnthn sub foo(Match $m) { }; foo("abc" ~~ /a/)
TimToady there is no Match anymore 19:34
TimToady there's a conjecture that you write that: sub foo(OKness $m) 19:34
jnthn So...there's no class Match { ... } ?
TimToady nor any class Failure
jnthn OK, what represents a Match object?
TimToady these turn notionally into annotated OK and Nil
OK with match info inside 19:35
benabik Match info stored in something other than a Match object?
TimToady such an OK is not much different than our current Match type
jnthn wonders if it's a different meta-object
TimToady benabik: basically it's just a rename of Match to OK
moritz how can I tell if an OK is a result of a regex match?
jnthn ok Match { has $!CURSOR; ... }
TimToady and OK by itself is now generalized success 19:36
jnthn Hm
No, not that... :)
TimToady: What do you think the definition of the thing that replaces Match will look like?
TimToady moritz: use case?
moritz TimToady: debugging
TimToady well, not many things will be returning OK other than match 19:37
most .ACCEPTS just return True or False
moritz can't we just continue to call it Match then?
and require defined-but-false Match to .gist-ify something that clearly indicates a match failure? 19:38
TimToady Match is not a good name for success, I think
moritz nom: say 'a' ~~ /b/
p6eval nom bd5adb: OUTPUT«#<failed match>␤»
moritz or maybe return a different type from a failed match
moritz is kinda fond of Match being returned from a match.
TimToady so I perceive, but maybe it's just sentimental attachment 19:39
moritz and maybe it's genuine
TimToady we're trying for clean primitives here, and most languages just give you something vs Nil
before it was just handwaving that makes Match work with smartmatching (when it does work) 19:40
moritz yes, I agree that something should be done in that area 19:41
[Coke] Anyone planning to go to YAPC::NA? they're looking for intro talks, wondering if an intro to perl6 would be good.
TimToady I really do think that people will not find it confusing for matches to return OK or Nil
eek, *do not 19:42
eek, nevermind
TimToady can't read
benabik OK just being able to store random information for the match seems stranger to me.
skids Yeah it's the part about OK being the type where things like .from and .to are specified that feels a bit icky. 19:44
TimToady well, of course it's stranger till you get used to it...historically most of my strange unifications have been...strange...
moritz TimToady: do you think of OK as actually delegating to its payload, or more like mixing stuff in? 19:45
TimToady there probably is a "match" object hiding inside of OK that, as the delegate, dispatches all those matchly methods
skids but cannot exist without an OK wrapping it?
TimToady moritz: it's really a primitive composition, and doesn't really describe itself well in terms of high-level constructs 19:46
skids: it can, but won't respond to smartmatching the way you'd think
jnthn TimToady: So your use of "role" here is just trying to convey the idea rather than say what you really think OK and Nil are?
TimToady nodnod 19:47
moritz TimToady: creating a new OO construct feels like a bit of a warning sign (not saying it's wrong, but...)
TimToady sorry for all the misleading metaphors
moritz: more trying to describe what's already there on some levels
or in some cases ought to be there as primitive as booleanness already is 19:48
jnthn TimToady: It almost feels like you want to apply the decorator pattern at the meta-object level. :)
jnthn TimToady: That is, we're really defining a class BUT with some set of presumptions that normal classes don't have. 19:49
TimToady yes, just like native types 19:50
and with similar restrictions on being productive
(as well as potential efficiency benefits)
TimToady OK might even delegate the HOW to the "real" object, since OKness is just some bits 19:53
jnthn TimToady: Yes, that's what I was getting at when I said decorator.
TimToady there might be no effective difference between "0 but OK" and "OK but 0"
jnthn That seems hard to imagine 19:54
TimToady it would have to be trapped and cheated on, if so
or we use other notation for it 19:55
jnthn Yes, because flattening composition normally implies that the meta-object being composed in doesn't play a role beyond the composition (maybe apart from appearing in the type check list of the target type) 19:56
TimToady if we hijack OK[] then we cheat in role space rather than infix:<bug>
er, *but :)
benabik Does but True and but False become playing with OK and Nil? 19:57
TimToady or we just have some primitive macros that flip bits
benabik: perhaps, if we cheat using "but"
TimToady officemates at the door... 19:58
lunch &
dalek kudo/nom: d4d4a21 | jnthn++ | src/Perl6/Metamodel/Mixins.pm:
Fix but True / but False.
 20:18
spine hi #perl6 20:54
masak OK/Nil seems to me like an odd pairing.
spine anyone got an idea why my local variables don't get gced on nom? 20:55
masak spine: hi!
spine perl6 -e 'sub a { my $b = "x" x 10_000; }; a() for 1..*;' eats up my memory at 40MB/s.
masak sounds like a bug.
jnthn nom doesn't know about sink context yet 20:56
So it's acting like map and building up a result list
It's not a GC bug, it's that we're actually holding on to them until the end of the for loop, when we should be throwing them away. 20:57
I'll bet if that sub is sub a { my $b = "x" x 10_000; 1; } then it doesn't leak. 20:58
spine jnthn: oh, I see
masak I like the unification of making fail/succeed semantic cousins.
as well as Nil.anything passing on Nil. 20:59
if Nil is now an item, we need a new idiom for the "empty for loop" use case.
jnthn spine: Yeah. It's a bug, but not about GC of local variables. 21:00
spine jnthn: so I'll just be careful to return something small from my functions for now :)
PerlJam jnthn: I agree with your assessment, but it still leaks with the return value 21:01
masak bare return still needs to return something that's false and undefined, IMO.
PerlJam masak: agreed. 21:02
(but I may just be agreeing because of my history with Perl 5)
jnthn PerlJam: er, yeah
I shoulda said "leaks a lot slower" rather than "doesn't leak"
Nil as the last thing may not leak.
But the real answer is "need sink" 21:03
PerlJam aye
masak PerlJam: yeah, me too. maybe it is as moritz++ says that 'return' now means, 'OK, we're done here, return successfully'. 21:04
PerlJam: maybe routines that exit with 'return' should be considered to have done a 'void return', and the value will explode if ever inspected by the caller :) 21:05
dalek p/bs: 1cb59b4 | jnthn++ | / (2 files):
Sketch out NQPRoutine and NQPSignature, which we'll migrate towards in place of a couple of PMCs. This will ease serialization, eliminate two PMCs and also eliminate our use of :instanceof. Hopefully...
p/bs: 52e0e73 | jnthn++ | / (6 files):
Move NQP's multi-dispatcher out of the 6model folder; it's not a core part of 6model.
p/bs: 37c6c88 | jnthn++ | / (2 files):
NQPRoutine needs to delegate invocation in order to serve as a code object.
dalek p/bs: 22467fe | jnthn++ | src/guts/multi_dispatch.h:
Sketch out C struct mappings for NQPRoutine and NQPSignature.
 21:05
masak that doesn't sound too Perlish, though.
PerlJam nope
almost makes me want another keyword ... "finish"
but maybe we should get rid of return and entirely and just use succeed/fail 21:07
s/and//
[Coke] how would succeed differ from return?
. o O (succeed Failure.new()) 21:08
PerlJam [Coke]: mostly in connotation. "succeed" says more about the result than "return" does
masak maybe that's the thing. 'return' is neutral. 'succeed' is a successful return. 'fail' is a return in shame. 21:12
I could certainly imagine coding like that :)
PerlJam me too
the only problem is that to "steal" succeed from when blocks like that probably means it will need a replacement :) 21:13
[Coke] meh.
masak does it. 21:14
s/\./?/
PerlJam (I'd vote for "done") 21:15
masak [Conjecture: fail scopes to topicalizer/sub the same as succeed]
sounds to me we'd get to keep the 'when' block semantics.
PerlJam I dunno. that's deeper thoughts than I can give right now.
masak and 'done' is taken by Test.pm :)
tk_ hi, so what is the different between rakudo star and rakudo on parrot 21:16
masak tk_: hi!
PerlJam tk_: rakudo star has a bunch of modules and things
masak tk_: both of those run on Parrot.
PerlJam tk_: rakudo star *is* "rakudo on parrot" but with more stuff 21:17
masak Rakudo proper is just the compiler. Rakudo Star bundles a bunch of nice stuff.
tk_ oh so rakudo star is snapshots of rakudo compiler + modules and stuff?
masak aye. 21:18
tk_ :)
appreciate it
masak rakudo: say "OH HAI, tk_!"
p6eval rakudo bd5adb: OUTPUT«OH HAI, tk_!␤»
masak star: say "This is Rakudo Star speaking :)"
p6eval star 2012.01: OUTPUT«This is Rakudo Star speaking :)␤»
masak see there, the rakudo gives a git hash (latest commit) as its version, but the star one gives a release version number. 21:19
tk_ :)
so cool
masak so it's wheels within wheels. one develops quickly, the other is more stable.
[Coke] star: use Nativecall;\
p6eval star 2012.01: OUTPUT«===SORRY!===␤Could not find Nativecall in any of: lib, /home/p6eval/.perl6/lib, /home/p6eval/rakudo-star-2012.01/install/lib/parrot/3.11.0/languages/perl6/lib, .␤»
masak yeah, the bots are cool. I forget sometimes how cool they are. :)
[Coke] star: use Nativecall;
p6eval star 2012.01: OUTPUT«===SORRY!===␤Could not find Nativecall in any of: lib, /home/p6eval/.perl6/lib, /home/p6eval/rakudo-star-2012.01/install/lib/parrot/3.11.0/languages/perl6/lib, .␤»
[Coke] star: use NativeCall; 21:20
PerlJam [Coke]: heh, I just did that in /msg
p6eval star 2012.01: ( no output )
[Coke] (ugh!)
rakudo: use NativeCall;
p6eval rakudo bd5adb: OUTPUT«===SORRY!===␤Could not find NativeCall in any of: lib, /home/p6eval/.perl6/lib, /home/p6eval/nom-inst1/lib/parrot/4.0.0-devel/languages/perl6/lib, .␤»
[Coke] (there's a concrete example.)
tk_ [Coke]: tyvm
was it difficult updating rakudo to work with parrot 4.0? 21:21
masak star: say "[[[[[]]]]]".comb.pick(*).join
p6eval star 2012.01: OUTPUT«][[][]][][␤»
masak tk_: don't think so.
tk_: communication these days flows fairly effortlessly between the Rakudo camp and the Parrot camp.
tk_ parrot seems like a very difficult project
PerlJam tk_: like anything else, the hard part is nailing down what "it" is. 21:22
tk_ yes, i admire the vision of the parrot foundation
masak Parrot is suffering a little bit from lack of a Very Satisfied Customer. 21:23
Rakudo is its main customer, but we've had to invent some of the stuff itself because (at the time, at least) Rakudo couldn't wait for Parrot to get them.
tk_ so was the stuff built on rakudo or contributed to parrot? 21:24
PerlJam tk_: both
tk_ have anyone here contributed to the perl 6 book? 21:25
PerlJam tk_: you can think of some of the things pioneered with Rakudo as the test run for what gets implemented in Parrot
masak tk_: most people who have contributed to the Perl 6 book hang out here, yes.
PerlJam masak: most? Who doesn't? ;) 21:26
tk_ is the virtual machine a new concept for perl? 21:26
is parrot the first perl virtual machine? 21:27
PerlJam tk_: that it's a separate, well-defined thing is new to Perl
masak PerlJam: I was pretty sure there are contributors of smaller things who aren't regulars here... :)
PerlJam tk_: but Perl 5 has a VM of sorts
tk_ PerlJam: oh?
masak everyone is so surprised Perl 5 has a VM! :)
tk_ i had to ask 21:28
masak oh, sure.
tk_ i hope perl 6 kills java
masak tk_: Perl 5 code gets compiled into some kind of op tree. the interpreter then runs the op tree.
PerlJam I hope Perl 6 doesn't kill anything
I hope that Perl 6 lives and thrives and makes its own way in the world 21:29
benabik Oracle's doing a pretty good job of killing Java. They don't need help. ;-)
masak I find it unlikely that everyone doing Java today will do Perl 6 tomorrow. :)
tk_ they are opposite
PerlJam benabik: we just have to cross our fingers that Oracles doesn't decide to help Perl 6 ;)
s/les/le/ 21:30
tk_ is a virtual machine defined in the specification of perl 6?
is it in java?
benabik Java's VM has a specification. I don't think the language spec is explicitly tied to it. (I could be wrong.) 21:31
PerlJam tk_: I don't think that Perl cares how things are implemented as long as they have the proper syntax and semantics.
[Coke] masak: I do java. I'd drop it for perl6 if I could make a convincing business case. (but sadly cannot) 21:32
masak tk_: you'd have a hard time making a fully compliant Perl 6 implementation that *didn't* run on a VM.
geekosaur tk_, there is no VM specified, and two of the current implementations are on different VMs (Parrot and CLR)
tk_ so the perl virtual machine is made so that perl can run cross - compatible? what are the advantages and disadvantages? 21:33
geekosaur er?
masak [Coke]: main reason I'm still doing helping with Perl 6 is to see it ease into the world a little each day. so that I (and others) can use it practically for more things.
tk_ geekosaur: i realize that one advantage of a virtual machine is that programs can be run independently of the system they are running on. but what are the _real_ advantages? 21:34
PerlJam masak: watching it ease into Perl 5 is interesting too :)
masak indeed.
but they can't steal some of our best stuff because backwards compat.
PerlJam depends on how well obra's plan for the future is executed. It may be that one day they break back-compat. because they can. 21:35
masak nod. 21:36
obra++
PerlJam and then we'll wake up one day and Perl 5 will have morphed into Perl 6 and there will only be Perl :) 21:36
jnthn tk_: All we really mean by "virtual machine" here is "something that provides services that a language implementation needs at runtime". Part of that is access to the underlying OS, part of it is garbage collection (freeing memory that is no longer in use), part of it is actually executing code, and so on. 21:37
tk_ nqp is like rakudo in that it compiles code for parrot? 21:39
PerlJam tk_: you can think of nqp as a mini-perl6 21:40
[Coke] ISTR jnthn++ had a nice slide for that. 21:41
dalek p/bs: 52c1b2b | jnthn++ | src/NQP/ (3 files):
Big refactor of block and symbol handling in NQP. This more fully shuffles it into NQP::World, matching the factoring that's worked out so nicely in Rakudo. It'll also make the upcoming multi refactoring much easier.
jnthn [Coke]: You may be thinking of the Rakudo architecture diagram from get_legacy_block_list() 21:42
er
from jnthn.net/papers/2011-tcpw-optimization.pdf
[Coke] jnthn: that seems sufficiently recent and pretty, yes. 21:45
tk_: try that URL, page through to "Rakudo Architecture".
tk_ ruby v python? 21:48
PerlJam ruby wins! 21:50
tk_ i havent used it, but i'm not sure if i can love anything as much as python 21:51
i also think that the industry needs to catch up with py3
masak good luck. 21:52
tk_ its unfortunate that i have to write code in py2
s/have/am forced
masak now you know how we feel, with Perl 5 and Perl 6 :) 21:55
tk_ "the perl 6 teleportation project" 21:56
tk_ in which caffeined perl 6 macho devs port every package to perl 6 21:57
[Coke]: 'we do a few fixups do world' ..??
is AST v World like .c v .h? 21:58
[Coke] tk_: context?
tk_ in the slide
'how rakudo runs programs' 21:59
jnthn tk_: The AST captures the execution aspects of the program (like statements). The World captures the declarations (classes, routines, etc.) 21:59
jnthn tk_: There's a phase when we've compiled the AST when we need to tweak some things. In Perl 6 you can run code at compile time, so when we parse a routine, we install a "compile it on demand" stub thingy. But before we run the program we have to replace that "compile on demand" stub with the real compiled output. 22:01
That phase is called "fixup time" in Rakudo.
nom: sub foo() { say "I get run!" }; BEGIN { foo() }; omfg !!!!! a parse error!! 22:02
p6eval nom bd5adb: OUTPUT«I get run!␤===SORRY!===␤Confused␤at /tmp/ss0bh2qdR4:1␤»
tk_ jnthn: is that the basis of JIT? in order to run at compile time? 22:03
jnthn tk_: No, JIT would be at a rather lower level than this.
tk_ jnthn: its a concept that needs a name 22:04
jnthn tk_: Though the "on demand" compilation is indeed what JIT compilers do.
tk_: But this is something else.
tk_: In Perl 6 we talk about BEGIN time. This is the time during compilation.
So the code I wrote there gets run at BEGIN time.
(and no, it's not something you do very often :-)) 22:05
(it's a power tool)
PerlJam parse-time thunking of compile-time information for proper run-time semantics
(but run-time could happen at any time ;) 22:06
jnthn Perl 6 kinda blurs the lines for the user, but if you're implementing the stuff you'd better know which you're doing at a given point in time. :) 22:08
tk_ pypy is so cool 22:10
it translates python into c
so does rakudo compile perl 6 into a subset of perl 6 for efficient compilation? 22:12
i.e. nqp
jnthn tk_: No, both Rakudo and NQP genrate an AST, and that is in turn transformed into code for the Parrot VM. 22:13
tk_: NQP being simpler allows us to more easily optimize it, and means it can be bootstrapped too.
Well, we could technically bootstrap all of Rakudo, but...I'd not fancy that. :) 22:14
tk_ what is bootstrapped
jnthn tk_: NQP can compile itself.
tk_: That is, it's written in NQP.
tk_ it said it was written in parrot
jnthn At least, mostly; the few parts left that aren't soon will be. 22:15
PerlJam tk_: bootstrapping is building just enough of a compiler to write a complicated compiler. Keep doing that until you have the compiler of appropriate complexity that you wan.t
to write a more complicated compiler.
tk_ rakudo uses nqp?
PerlJam tk_: ye
jnthn tk_: Rakudo is mostly written in NQP
tk_ so you could eliminate rakudo and write in nqp? 22:16
jnthn tk_: Yes, NQP is a language in its own right. 22:16
tk_ why wouldnt you do that, jnthn
jnthn tk_: It's a subset of Perl 6, with just the right things for writing compilers.
tk_ thats what i said
PerlJam tk_: NQP doesn't have assignment for instance. Only binding. 22:17
tk_ example? i dont
masak binding is tying a value to a variable. assignment is placing a value in a container (which in turn may be tied to a variable). 22:19
nom: my $a; my $b := $a; $a = 42; say $b
p6eval nom bd5adb: OUTPUT«42␤»
masak (the ':=' is binding, and makes $b and $a point to the same container)
so, um, binding is tying *something* to a variable -- either a value or a container. 22:20
nom: my $a := 42; $a = 5
p6eval nom bd5adb: OUTPUT«Cannot assign to a non-container␤ in block <anon> at /tmp/p6UiRlEkLx:1␤␤»
masak when tying $a directly to a value like that, it loses its ability to be assigned to. because the container that assignment needs is out of the loop -- there's just the variable and the value. 22:21
tk_ so declare and define is equivalent to bind and assign?
masak I'm not sure that analogy holds, no.
tk_ what else is := used for? 22:25
masak signature binding. 22:26
nom: sub foo($a) { say $a }; my $b = 42; foo($b)
p6eval nom bd5adb: OUTPUT«42␤»
masak the $a and the $b are distinct variables, but they need to point to the same value/container.
binding to the rescue. 22:27
tk_ thank you, lover.
masak has never been called "lover" on #perl6 before :) 22:28
tk_ normally i'd reserve that language for my usb port, but you're something special ;) 22:29
masak not gonna argue there. :)
pyrimidine it's getting hot in here :)
masak should've said "...I see my reputation precedes me" :) 22:30
tk_: we do want to let newbies feel they are loved when they arrive. we love to heap attention on them. 22:32
tk_ i'm speechless 22:33
tk_ has a sense of gratitude in his eyes
tk________ *ahem* 22:35
tk________ crosses legs
masak now that's perhaps somewhat below the level of discourse we try to keep around here. 22:38
keep in mind this is a family channel.
tk________ im trying to keep it g 22:39
masak nom: .say for (1, 1, 2, * + * ... *)[^20] 22:41
p6eval nom bd5adb: OUTPUT«1␤1␤2␤3␤5␤8␤13␤21␤34␤55␤89␤144␤233␤377␤610␤987␤1597␤2584␤4181␤6765␤»
tk______ masak: in the signature binding example you never used := 22:44
masak no, because the binding is done behind the scenes, through the function call. 22:48
but you asked about another example of binding, not of the actual ':=' syntax.
tk______ oh
OH 22:49
tk______ my comp sci teacher always used to say "java 5's new boxing feature hurr durr" which made primitive data types into objects 22:50
autoboxing
masak in Perl 6, Int is a full-featured object, and int is the native type. 22:52
generally, you'd use Int for full flexibility, and int for memory compactness/speed. 22:53
tk______ a:=42 is that an Int or int 22:55
masak std: a := 42 22:56
p6eval std 52f3895: OUTPUT«===SORRY!===␤Preceding context expects a term, but found infix := instead at /tmp/SaaPMU4bNU line 1:␤------> a ⏏:= 42␤Parse failed␤FAILED 00:01 109m␤»
masak it's not syntactically correct Perl 6.
you need to declare variables, using 'my' or equivalent. 22:57
benabik nom: my \a := 42; say a
p6eval nom bd5adb: OUTPUT«===SORRY!===␤Malformed my at line 1, near " \\a := 42;"␤»
benabik NYI, I guess.
tk______ masak: so, using = on an int would be illegal? 22:58
because it doesnt have a container
in reference to your earlier example
jnthn tk______: If you want to work with native types, you have to explicitly declare variables as being natives
my int $a = 42; # $a stores 42 as a native integer 22:59
my $a = 42; # $a stores 42 as a boxed Int
tk______ std: my int $a = 42; $b = 43
p6eval std 52f3895: OUTPUT«===SORRY!===␤Variable $b is not predeclared at /tmp/3JLRbG4vmK line 1:␤------> my int $a = 42; ⏏$b = 43␤Check failed␤FAILED 00:01 111m␤»
tk______ std: my int $a = 42; my $b = 43
p6eval std 52f3895: OUTPUT«ok 00:01 110m␤»
tk______ std: my int $a = 42; my $b = $a 23:00
p6eval std 52f3895: OUTPUT«ok 00:01 110m␤»
tk______ std: my int $a := 42; my $b = $a
p6eval std 52f3895: OUTPUT«ok 00:01 110m␤»
masak tk______: your question about natives, assignment, and containers makes a lot of sense. I don't know the answer, though.
tk______ std: my $a := 42; my $b = $a
p6eval std 52f3895: OUTPUT«ok 00:01 110m␤»
masak keep in mind that STD only checks syntax, it doesn't run your code.
skids jnthn: that kinda confuses me given the "native types dispatch to their container's methods" thing, which makes me expect "my $i = int(42)" to be the way to do things. 23:01
tk______ nom: my $a := 42; my $a = 5 23:02
p6eval nom bd5adb: OUTPUT«===SORRY!===␤Redeclaration of symbol $a␤at /tmp/EMNdolqru0:1␤»
tk______ nom: my $a := 42; $a = 5
p6eval nom bd5adb: OUTPUT«Cannot assign to a non-container␤ in block <anon> at /tmp/cDYF5FsQjy:1␤␤»
tk______ nom: my int $a := 42; $a = 5
p6eval nom bd5adb: OUTPUT«===SORRY!===␤Cannot bind to a natively typed variable; use assignment instead␤at /tmp/onl5kHz6Ht:1␤»
tk______ nom: my int $a = 42; $a = 5 23:02
p6eval nom bd5adb: ( no output )
masak seems it's binding that doesn't work for natives.
jnthn Correct. 23:03
Because there's no container.
tk______ so its Int by default
tk______ nom: my Int $a := 42; $a = 5 23:03
p6eval nom bd5adb: OUTPUT«Cannot assign to a non-container␤ in block <anon> at /tmp/HL7WKXUkdu:1␤␤»
jnthn skids: It's not quite like that - actually if you do a method call on a native type it has to box it first. 23:04
skids: When you declare "my int ..." you're saying "don't allocate a container, instead just make space in the lexpad for a native integer"
dalek p/bs: 0fad218 | jnthn++ | src/ops/nqp.ops:
Add an op for setting the signature as a single object.
 23:06
p/bs: e7be6cc | jnthn++ | src/guts/multi_dispatch.h:
Fix typo.
p/bs: 18f17e9 | jnthn++ | src/ops/nqp.ops:
Oops, can't remove this yet - gotta go round the bootstrap loop.
p/bs: 78fb2c3 | jnthn++ | src/guts/multi_dispatch.c:
Get multi dispatcher to recognize both new and old way of doing the signature for a moment.
p/bs: bfadf52 | jnthn++ | src/NQP/ (2 files):
Switch to using the new signature objects.
p/bs: 1ac3f88 | jnthn++ | src/guts/multi_dispatch.c:
Quick tweak and...the new signature objects work.
p/bs: 33fc4d2 | jnthn++ | src/stage0/ (6 files):
Update bootstrap.
p/bs: 7c8a11c | jnthn++ | / (4 files):
Rip out the NQPMultiSig PMC; we don't need it any more.
tk______ so if youre binding an integer you could just use native types for memory efficiency 23:06
jnthn Yeah *but* Int also provides big integer semantics. 23:07
skids so "my int $i" declares a virtual container that autoboxes into ... ? ... and then methods called on its contents (an immutable int) dispatch to the result of the autobox.
jnthn nom: say [*] 1..100
p6eval nom bd5adb: OUTPUT«93326215443944152681699238856266700490715968264381621468592963895217599993229915608941463976156518286253697920827223758251185210916864000000000000000000000000␤»
jnthn Much bigger than a native int can store.
skids: If you do $i.foo then it code-gens a boxing if $i into an Int. 23:08
And then what you said, yes.
s:2nd/if/of
tk______ so binding essentially erases the container of an Int? 23:09
nom: my $a = 5; $a := 42 23:10
p6eval nom bd5adb: ( no output ) 23:10
tk______ nom: my $a = 5; $a := 42; $a := 5; $a = 5
p6eval nom bd5adb: OUTPUT«Cannot assign to a non-container␤ in block <anon> at /tmp/Ti0vvEUFZh:1␤␤»
tk______ nom: my $a = 5; $a := 42; $a := 5
p6eval nom bd5adb: ( no output )
tk______ nom: my $a = 5; $a := 42; $a := [*] 1..100; say $a 23:11
p6eval nom bd5adb: OUTPUT«93326215443944152681699238856266700490715968264381621468592963895217599993229915608941463976156518286253697920827223758251185210916864000000000000000000000000␤»
tk______ nom: my int $a = [*] 1..100; say $a 23:12
p6eval nom bd5adb: OUTPUT«0␤» 23:12
tk______ nom: my $a = [*] 1..100; say $a
p6eval nom bd5adb: OUTPUT«93326215443944152681699238856266700490715968264381621468592963895217599993229915608941463976156518286253697920827223758251185210916864000000000000000000000000␤»
skids OK, so if "int $i" is contains the platform's MAXINT, and you call "$i++", then at some point after the boxing, the Int gets unboxed again, and that's when we'd get an error, or is Int supposed to have some wisdom about underlying REPRs? 23:13
tk______ is all this Int and boxing stuff defined in the spec? 23:14
jnthn skids: ++ is an operator. When we're done with natives, that'll just be a CPU-level increment. And it should check for overflow and throw also. 23:15
skids The binding stuff and such related material is kinda scattered around the specs, really needs a more congealed workup.
jnthn Note that the native types implementation is still quite new and experimental. 23:16
masak yes, the thing with native types and boxing is defined by the spec.
jnthn It works well in various places, and not so well in others just yet.
skids jnthn: that, I have definitely noticed :-)
jnthn skids: Yeah. Many improvements there are on my todo list. 23:17
skids WRT the ++ example, I was thinking more on a "theoretically if we didn't have an optimizer" specifications level. 23:19
masak 'night, #perl6 23:29
jnthn figures he'd best get some rest also 23:44
'night

Powered by Cro and the Raku® Programming Language.

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

Thank you!