»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.
timotimo in my opinion, there must be recognition of repetition and redundancy across sub/method call boundaries 00:00
anything that'll fall out of data flow analysis, really
anyway, since i really don't know nearly enough about this, i'll just go to bed and see what the next day will bring :)
00:10 FROGGS left 00:15 Psyche^ left 00:17 Psyche^ joined 00:44 dwarring left
japhb__ r: use Test; class A {}; multi infix:<eqv>(A $a, A $b) { True }; my A $one .= new; my A $two .= new; say $one eqv $two; is_deeply($one, $two); 01:07
camelia rakudo c6d5c9: OUTPUT«True␤not ok 1 - ␤# got: A.new()␤# expected: A.new()␤»
japhb__ GAH.
japhb__ shakes his fist at Rakudo
colomon why is deeply? 01:08
japhb__ colomon: That's the test to use when you want eqv comparison.
colomon but they're not eqv
it's just the got / expected that are wrong. 01:09
japhb__ colomon: look at the code again. ;-)
colomon oh
:\
japhb__ Yeah.
My best guess right now is that the infix:<eqv> cannot participate in the eqv multi used by the already-compiled Test class. 01:10
colomon oh!
that's totally it
that's the way p6 is supposed to work 01:11
if you want to shake your fist, you need to shake it at TimToady++
r: sub check($a, $b) { say $a eqv $b; }; class A {}; multi infix:<eqv>(A $a, A $b) { True }; my A $one .= new; my A $two .= new; say $one eqv $two; check($one, $two) 01:12
camelia rakudo c6d5c9: OUTPUT«True␤True␤»
colomon unless that's not it at alll...
japhb__ How in the world do you make use of multi candidates that aren't in the same compilation unit then?
colomon that's what lift is supposed to be for 01:13
and I totally agree it kind of sucks
r: sub check($a, $b) { say $a eqv $b; }; class A {}; my A $one .= new; my A $two .= new; say $one eqv $two; check($one, $two) 01:14
camelia rakudo c6d5c9: OUTPUT«False␤False␤»
japhb__ 'git grep lift' in a rakudo checkout gives nothing of value.
colomon japhb__: right, because lift is just hypothetical so far
TimToady and likely to go away anyway
colomon and may well not end up being named lift
TimToady! \o/ 01:15
japhb__ Ah, language designer in da house.
TimToady well, it's likelier to be macros that can implicitly carry down operators
but is_deeply should probably be parameterized for comparison
japhb__ So: how do I fix my code at 18:07 such that Perl 6 will Do What I Want?
colomon probably don't use is_deeply, I think. 01:16
TimToady japhb__: well, one of the two things I just said
diakopter how many timezones is that from me
japhb__ is_deeply is just an example of the broader problem.
TimToady either is_deeply gets a :by parameter
or we have macros that can do &infix:<cmp> = &CALLER:<cmp> 01:17
well, maybe can get away without full macros there 01:18
rather, &infix:<cmp> = &CALLER::infix:<cmp>
japhb__ eqv recurses through the proto at each level. Do you mean that you would do :by(&infix<eqv>) and thus get the caller's view of the proto?
colomon diakopter: I believe he means 9:07 Eastern
japhb__ diakopter: Are you not in your usual timezone? 01:19
TimToady he's at ppw
japhb__ Ah!
Are you also?
colomon ppw++
colomon is in Chicago for Legoland 01:20
japhb__ Very cool.
diakopter polish perl workshop is great 01:22
colomon pghpw.org/ppw2013/talk/5091 01:24
are you going to have slides online?
diakopter heh
considering I changed the topic today...
colomon oh? 01:25
diakopter yeah, to "WTFP6" 01:26
or "P6WTF"; I haven't decided 01:27
colomon moar bugs that bad today? ;) 01:28
diakopter no it's meant to identify with the audience
and the first Sunday talk 01:30
colomon diakopter++ # representing 01:31
TimToady we all show our passion in different ways :) 01:32
diakopter I'm not sure I'll actually be representing anyrhing 01:33
japhb__ r: class A {}; sub foo { say &CALLER::infix:<eqv>(A.new, A.new) }; multi infix:<eqv>(A $a, A $b) { True }; foo()
camelia rakudo c6d5c9: OUTPUT«No such method 'dynamic' for invocant of type 'Sub+{Precedence}'␤ in method at_key at src/gen/CORE.setting:9171␤ in sub postcircumfix:<{ }> at src/gen/CORE.setting:2414␤ in sub postcircumfix:<{ }> at src/gen/CORE.setting:2410␤ in sub foo at /tmp/wc8…»
japhb__ Now WTH does that mean?
TimToady it means "Now WTH does that mean?" 01:36
japhb__ TimToady: OK ... so what did I not understand about your suggestion? (Leaving aside the question of temporary assignment of &infix:<eqv>, I can't even *use* it right now) 01:37
diakopter colomon: but no, jnthn and FROGGS seemed to make huge progress today on the moarqp bootstrap 01:38
colomon \o/
TimToady right, it will take either a compiler to allow infix:<eqv> to be accessed as a dynamic lexical, or it will take some kind of semimacro to tell the caller to pass it implicitly somehow, or enable it to show up in the current lexical scope in a dynamically visible fasion
colomon very exiting.
diakopter yes, exciting too 01:39
japhb__ hmmm
colomon I'm sorry I've been so out of things lately. And I'm going to keep being too busy to p6, I should be $working right now.
diakopter colomon: thumbs up 7up 01:40
:D
japhb__ colomon: Seems to be the season for $day-job tuit constraint ....
diakopter ,i 01:41
colomon japhb__: for me it's as much life as $day-job, but yes, almost all my normally spare tuits are being funneled into contract $work right now. 01:42
japhb__ TimToady: in not treating infix:<eqv> as a dynamic lexical, is this a rakudobug or WAI?
TimToady r: class A {}; sub foo { say CALLER::['&infix:<eqv>'](A.new, A.new) }; multi infix:<eqv>(A $a, A $b) { True }; 01:43
camelia ( no output )
TimToady r: class A {}; sub foo { say CALLER::['&infix:<eqv>'](A.new, A.new) }; multi infix:<eqv>(A $a, A $b) { True }; foo()
camelia rakudo c6d5c9: OUTPUT«Cannot call 'Real'; none of these signatures match:␤:(Mu:U \v: Mu *%_)␤ in method Real at src/gen/CORE.setting:944␤ in method Real at src/gen/CORE.setting:3422␤ in method Real at src/gen/CORE.setting:944␤ in sub infix:<<> at src/gen/CORE.setting:3…»
japhb__ o.O 01:44
01:44 cognominal left
TimToady r: class A {}; sub foo { say CALLER::('&infix:<eqv>')(A.new, A.new) }; multi infix:<eqv>(A $a, A $b) { True }; foo() 01:44
camelia rakudo c6d5c9: OUTPUT«No such method 'dynamic' for invocant of type 'Sub+{Precedence}'␤ in method at_key at src/gen/CORE.setting:9171␤ in sub postcircumfix:<{ }> at src/gen/CORE.setting:2414␤ in sub postcircumfix:<{ }> at src/gen/CORE.setting:2410␤ in sub INDIRECT_NAME_L…»
TimToady anyway, we need to be able to do something like that, one way or t'other 01:45
japhb__ wonders if jnthn++ got partway through implementing it, just based on the odd error message. 01:46
TimToady r: class A {}; sub foo { say CALLER::myeqv(A.new, A.new) }; multi infix:<eqv>(A $a, A $b) { True }; my &myeqv = &infix:<eqv>; foo() 01:47
camelia rakudo c6d5c9: OUTPUT«No such method 'Any' for invocant of type 'Parcel'␤ in sub foo at /tmp/bykqP9x6HL:1␤ in block at /tmp/bykqP9x6HL:1␤␤»
japhb__ TimToady: It feels to me that multis need to by default stay "open", because something like infix:<eqv> can work its way through multis in quite a few compilation units before bottoming out doing comparison of complex data structures.
TimToady they do stay open, in a lexically scoped sense 01:48
we're just trying to use CALLER to get at the lexically scoped definition at the point of the caller 01:49
once it finds the right proto, it should happen right automatically 01:50
japhb__ I meant they'd need to stay open in a dynamic sense ... but maybe if every single eqv multi that wants to delegate always does CALLER::infix:eqv when it does so, the chain could be kept correct ...
Yeah, I'm seeing a little more clearly what you're going for.
TimToady whether we can get away with doing the completely dynamically, or need a bit of macroish help, I dunno 01:51
it would be nice to have the compiler aware enough of it that, should it inline the call, it can just throw away all the linkage
japhb__ Well, every eqv multi would have to do: local infix:<eqv> = &CALLER::infix:<eqv> before delegating
hmmm
TimToady that's why I'd like some syntactic relief in the signature, perhaps 01:52
but we're moving away from the lift idea, insofar as that was going to auto-hoist every operator, but we really want a little better control 01:53
japhb__ signature of ... the eqv multi candidates?
TimToady either in the proto sig, or in each multi
not all multis want to lift 01:54
eventually you find the multi that is the base case, and that wants the raw version of eqv
japhb__ 'proto infix:<eqv>(|) is lifted {*}' ?
TimToady not that proto
or not like that 01:55
01:55 colomon left
TimToady I think it's more likely that each multi has a way to talk about each operator as lifted or not 01:55
which is why we were discussing a hook directly into the parser such as ¤ might provide for ASTs 01:56
but it's not an AST, it's just an implicit caller's operator
japhb__ That seems to put a lot of burden on the casual extender (by which I mean, person who just wants to add a multi candidate for their new class)
TimToady so $a © eqv $b or some such 01:57
01:57 prevost joined
japhb__ And perhaps I'm thinking too narrowly about eqv, but I don't think that one bottoms out to "raw version of eqv", but rather bottoms out when the comparison begins to use other operators (eqv for Str, Str translating to eq) 01:57
01:58 colomon joined
japhb__ In other words, you start looking at a different proto. 01:58
TimToady eqv is perhaps a bad example
japhb__ yeah, perhaps 01:59
I can see the value of a (c) meta-operator, though
01:59 cognominal joined
TimToady $a ©+ $b is perhaps a more likely lifted operator 01:59
© for CALLER of course
japhb__ I understood. :-) 02:00
TimToady the point of © and ¤ is to provide direct hooks into the parser to let them slip things into the parse in place of a syntactic category
so there's be some kind of generic "caller infix" that this would pull in as a kind of metaoperator 02:01
japhb__ Under that scheme, would much of Any.pm, Cool.pm, Real.pm start needing their operations to use (c) metaops everywhere?
TimToady much as we might have macros that can replace other things than just nouns, but we need to be explicit about exactly how much of the AST we're replacing
02:02 logie joined
TimToady so the type of the ¤ needs to be either on the parameter or at the point of the "gnostic interpolation" 02:02
well, at some point we'll need to control genericity, or we're just re-inventing BAL 360 :) 02:03
but fershure you can't inline everything--that don't scale to good
*too 02:04
you'd like the modules that develop new ops to pay the price, not everyone else
so yeah, we might need people to register their interest in genericity to avoid such problems 02:05
02:08 mattp_ left
dalek p/moarboot: fa5dcf2 | coke++ | t/docs/opcodes.t:
test for MOAR opcodes, also
 02:09
02:09 mattp_ joined 02:13 FROGGS joined
benabik TEST MOAR THINGS! 02:15
"¤"? 02:17
[Coke] ?
.u ¤ 02:18
yoleaux U+00A4 CURRENCY SIGN [Sc] (¤)
02:18 FROGGS left
benabik [Coke]: TimToady was talking about ¤ and ASTs, and I've never seen this symbol before. Ack doesn't seem to find it in the specs, either 02:19
TimToady 'tis an IRC speculation of long standing 02:20
benabik Interesting that there is a generic currency symbol. I had always just used µ for "monetary unit"
Sadly the irclogs seem to use it as part of mojibake for the N/L symbol. 02:21
So it's fairly unsearchable.
TimToady you'd wanna scan for me discussing macro parameters and syntax categories with masak++ 02:22
or scan for "CURRENCY" :) 02:24
benabik As people go "what is that character"? Good idea.
*blinks* And now the irc logs are 403ing, which makes it even less searchable. I'll assume moritz++ is doing something to them. 02:25
TimToady not likely at this time of night...
japhb__ is thinking about 'class A { has Mu @.mus }; class B { has A @.as }; my @c := [1, "foo", B.new(as => [A.new(mus => ...)])];' where A, B, and @c are defined in separate compunits, and wondering how anything non-pessimizing could possible handle multi operations on @c correctly 02:26
TimToady the logs themselves eem fine to me
benabik Oh, it's not the newline character. Maybe color codes? It's in all the SORRY lines. 02:27
And they're back. Probably cosmic rays
TimToady someone needs to teach that program to understand color codes someday
02:28 prevost left
japhb__
.oO( Rule 3: "Well volunteered!" doesn't work on Larry. )
 02:28
TimToady Not when I was told at the beginning of this that they didn't like my programming style... :)
lue TimToady: what is the generic currency symbol proposed to mean? Or must we search the logs for an answer? 02:29
TimToady of course, all the VMs start looking just the same, because portable C forces you to do macros that way
benabik Searching for "currency sign" in the clogs seems to find it nicely. TimToady++ 02:30
TimToady it's intended to signal the parser that it should immediately replace an alternative in the grammar with something of known category
02:30 apejens joined
TimToady so where it looks for a bunch of <infix> it would know that ¤myoperator is an AST parameter declared somewhere as an infix 02:30
or ¤infix:myparameter or some such 02:31
lue imagines the texas version to be >()< based on his current font :)
02:31 wsri joined
TimToady the category could be marked on either the interpolation location or on the parameter decl 02:31
this is intended to replace and generalize the {{{$myparam}}} monstrosity
benabik wonders if the texas version is "if you're mucking around with macros, you can figure out Unicode" 02:32
lue Ah, so you keep the parser from trying to sift through alternatives by just giving it the answer right away, IIUC.
.oO( "if you're mucking around with macros, you can figure out Unicode"infix:myparameter )
benabik lue++
TimToady yes, but masak++ was dubious
when I floated that it should be difficult :) 02:33
benabik masak++ is good at dubious.
TimToady it needs to be something that highjacks the parser, so it needs to be rather unusual
since there's be an implicit alternative to almost any rule in the grammar 02:34
lue
.oO(for unusual, perhaps a quaternary operator, just to show you mean it. <something> ? <something> : <something> | myoperator :P)
 02:35
TimToady that's not what I call "obvious"
¤ is obvious :)
lue
.oO(And then maybe hide ¤ behind a use MONKEY_SMARTS; if safeguards are desired)
 02:36
TimToady in the sense of giving the parser an immedate WTF moment
lue ‽ would nicely sum up the parser's feelings :) ‽infix:right_alternative 02:37
TimToady well, we're still struggling feebly to stay in Latin-1, so that there's likely to be a compose key 02:38
though I guess that one has a compose
⸘Doesn't it‽ 02:39
lue Don't be silly. I only needed to type <Menu Key> ? ! for that! There was no key labelled "compose key" involved :)
TimToady :P
benabik .u ⸘
yoleaux U+2E18 INVERTED INTERROBANG [Po] (⸘)
benabik Okay, while I'm not too surprised interrobang is popular enough to make Unicode... Inverted Interrobang? 02:40
.u ‽
yoleaux U+203D INTERROBANG [Po] (‽)
lue TimToady: I think we're basically fine if we stay inside BMP for Core P6. ☺
TimToady 😺 02:41
lue (if only because there's a font out that can act as an ultimate fallback for that range: unifont)
TimToady imagines that it gives you half the glyph for surrogates :) 02:44
02:51 logie left
lue Huh. For some utterly strange and weird reason my "print the BMP" one-liner fails on character D800 :) 02:53
03:00 FROGGS joined 03:06 FROGGS left 03:08 FROGGS joined, lizmat left 03:13 lizmat joined 03:14 woolfy joined 03:15 prevost joined 03:18 FROGGS left 03:31 lizmat left 03:32 woolfy left
japhb__ My current conundrum -- Sum.pm: 'module Sum; sub zip-sum(Mu @a, Mu @b) is export { @a Z+ @b }' A.pm: 'class A {}; multi infix:<+>(A $a, A $b) is export { 42 }' main: 'use Sum; use A; say zip-sum([A.new], [B.new]);' How do you compile each of these files so they are default-efficient, but main still produces the correct answer? 03:32
(Aside from B.new should be A.new, of course) 03:33
03:34 lizmat joined 03:36 logie joined 03:37 woolfy joined
japhb__ .tell jnthn Thoughts on irclog.perlgeek.de/perl6/2013-10-06#i_7677518 through irclog.perlgeek.de/perl6/2013-10-06#i_7677853 very much appreciated. 03:44
yoleaux japhb__: I'll pass your message to jnthn.
japhb__ .tell jnthn In particular, I *think* much of the problem in this space can be summarized by irclog.perlgeek.de/perl6/2013-10-06#i_7677850 . 03:46
yoleaux japhb__: I'll pass your message to jnthn.
japhb__ TimToady: Do you consider the addition of new multi candidates to be a mutation of the language, or just the addition of new protos? In other words, there a linguistic reason (as opposed to performance) reason why my example from 20 minutes ago shouldn't Just Work? 03:53
*is there 03:54
03:58 preflex_ joined, ChanServ sets mode: +v preflex_, preflex left, preflex_ is now known as preflex 04:03 finanalyst joined 04:18 colomon left 04:25 logie left
japhb__ Furthermore, given -- A.pm: 'use B; class A { has B @.b }' B.pm: 'use C; class B { has C @.c }' C.pm: 'class A {...}; class C { has A @.a }' -- How in the heck would you sanely define working infix:<eqv> multis for each of these? 04:26
04:28 sisar joined 04:33 logie joined 04:48 fridim_ left 05:16 xenoterracide left, xenoterracide_ joined, prevost left
moritz \o 05:23
05:28 rindolf joined 05:29 xenoterracide_ left 05:37 rindolf left 05:45 logie left 05:46 darutoko joined 05:59 Celelibi joined 06:02 BenGoldberg left 06:10 logie joined 06:20 btyler left, FROGGS joined 06:58 logie left 07:23 FROGGS left 07:51 dmol joined 08:01 FROGGS joined
FROGGS o/ 08:07
arnsholt Does anyone have access to a 32-bit Linux machine? 08:16
I'm pretty sure NQP #28 is fixed, but can't confirm since it's supposedly a 32-bit thing
FROGGS arnsholt: I'm booting one atm 08:17
arnsholt Oh, awesome. If the NQP test suite runs cleanly on it, you can close #28 08:18
FROGGS k
08:22 kaare__ joined, kaare__ is now known as kaare_
FROGGS t/nqp/46-charspec.t (Wstat: 256 Tests: 0 Failed: 0) 08:25
Non-zero exit status: 1
Parse errors: No plan found in TAP output
ahh, these are due to missing icu 08:26
arnsholt FROGGS++ # Cheers! 08:29
FROGGS :o) 08:30
dalek rlito: 875d2eb | (Flavio S. Glock)++ | / (3 files):
Perlito5 - perl5 - pretty-printer - statement modifier
FROGGS arnsholt: btw, I'd like to keep #133 open until we actually have a parrot release that fixes it
arnsholt Yeah, that's probably a good idea 08:56
bbkr__ indeed :) lizmat++ colomon++, thanks for instant help 08:58
arnsholt And then we can both try to remember to close it when the next Parrot release is out
bbkr__ sorry
wrong widow :)
window*
09:03 xinming left 09:05 xinming joined 09:09 KroKite joined 09:14 spider-mario joined
arnsholt bbkr__: Could you give the stuff in NQP #32-35 a whirl? Are those warnings still there, or can we close the bugs? 09:14
09:17 yogan left, yogan joined
bbkr__ arnsholt: sure, testing it right now... 09:18
09:18 yogan left
arnsholt Cheers! 09:19
09:25 sisar left
bbkr__ so far 33, 32, 35 have popped in build messages :( 09:31
new one spotted: github.com/perl6/nqp/issues/136 09:33
arnsholt Oh, that one looks legitimate 09:35
bbkr__ this compiler is way too oversensitive 09:36
jnthn aye
It is, but for once the prec warning has pointed out a problem :)
arnsholt jnthn: What's your thoughts on NQP #46? Bug or WONTFIX? 09:37
bbkr__++ 09:40
dalek p: 8f719a5 | (Arne Skjærholt)++ | src/vm/parrot/ops/nqp_dyncall.ops:
Fix precedence oops in dyncall ops. Fixes #136.
09:41 KroKite left
bbkr__ I've dumped all warnings to a single file: gist.github.com/bbkr/6851782 09:42
If someone wants OS X shell account to fix them please PM me
jnthn arnsholt: Bug 09:44
arnsholt Okies 09:45
bbkr__ I've closed #34 09:46
arnsholt I tried to fix the contextual one, but the easy fix I tried didn't work
bbkr__: Many of those are artefacts of the C code being generated code, but the non-void ones sound legitimate as well 09:47
dalek p/moarboot: c847d0e | (Tobias Leich)++ | tools/build/Makefile-Moar.in:
whitespace cleanup
 09:48
09:55 yogan joined 09:56 grondilu joined
dalek rlito: 6f61350 | (Flavio S. Glock)++ | / (3 files):
Perlito5 - perl5 - pretty-printer - operators
 09:58
arnsholt bbkr__: Most of the non-void ones are due to clang not handling GCC's "doesn't return" annotations, but I've found a bug as well 10:01
10:05 [Sno] left
dalek p: b1acc0b | (Arne Skjærholt)++ | src/vm/parrot/pmc/sixmodelobject.pmc:
Add a missing return.

  bbkr__++ for finding this. Fixes #35.		 10:09
FROGGS ohh, nice one
10:18 hexcoder joined
hexcoder Hi, i have a stupid beginners question. When using latest rakudo in REPL mode (without args), how am i supposed to enter multi line snippets? 10:25
I tried pasting these 5 lines, but the block spanning 3 lines was not accepted. 10:27
my @array = 1, 2, 3, 'foo'; my $sum = @array[0] + @array[1]; if $sum > @array[2] { say "not executed"; }
Hm, here they are join into one line
Hm, here they are joined into one line 10:28
FROGGS that might be due to your irc client
hexcoder yes
FROGGS i'm not a REPL user so I dunno if you can have multi-line statements at all
why not paste it as one line in the repl?
r: my @array = 1, 2, 3, 'foo'; my $sum = @array[0] + @array[1]; if $sum > @array[2] { say "not executed"; } 10:29
camelia ( no output )
jnthn It doesn't do multi-line statements
hexcoder I thought for testing many small snippets , this was the way to go...
ok, thanks.
jnthn Each line has to be valid syntax on its own
FROGGS r: my @array = 1, 2, 3, 'foo'; my $sum = @array[0] + @array[1]; unless $sum > @array[2] { say "not executed"; }
camelia rakudo c6d5c9: OUTPUT«not executed␤»
jnthn However, variables or functions declared on previous lines are visible to subsequent lines.
FROGGS hexcoder: you can privmsg camelia and test there too if you want 10:30
hexcoder So I would have to write scripts. This is the easiest way?
ok, will try camelia...
FROGGS scripts have the benefit that you can change previous lines... that is sort of impossible with a REPL 10:32
arnsholt jnthn: I think we can close NQP #66, but I'm not quite sure. Care to have a look?
Or maybe not close, but I think I found the issue at hand at least
jnthn yeah. I bet that's gonna vary by platform too 10:34
hexcoder r: my @array = 1, 2, 3, 'foo'; my $sum = @array[0] + @array[1]; if $sum > @array[2] { say "not executed"; }
camelia ( no output )
arnsholt bbkr__: I'd love an OS X shell, BTW!
jnthn On JVM, WVal lookups resolve once. Moar does them pretty efficiently too
On Parrot I think it's actually a hash lookup...
arnsholt My old Mac laptop has been retired, which makes it harder to test OS X specific things 10:35
dalek rlito: 66f5d76 | (Flavio S. Glock)++ | / (3 files):
Perlito5 - perl5 - pretty-printer - operators
 10:43
bbkr__ arnsholt: did you get login info on private channel? 10:49
10:57 [Sno] joined 11:07 hummeleB1 joined 11:13 ssutch left
lizmat good *, #perl6! 11:16
11:19 SamuraiJack__ joined
moritz \o lizmat 11:25
11:28 finanalyst left 11:30 finanalyst joined 11:34 finanalyst left 11:38 colomon joined
hexcoder a question regarding read-only parameters for subs: sub try-to-reset($bar) { $bar = 2; }. Can this ever have a chance to succeed? If not, why let it compile? 11:38
jnthn No, it won't, we just aren't doing the static analysis needed to rule it out at compile time yet. 11:39
It'll probably be done in the optimization phase, when it's trying to prove other things about assignments. 11:40
hexcoder great, thanks.
jnthn Which is the same point it rejects sub calls with incorrect argument counts/types when it can.
lizmat r: sub a ($a) {}; a() 11:50
camelia rakudo c6d5c9: OUTPUT«===SORRY!===␤CHECK FAILED:␤Calling 'a' will never work with no arguments (lines 1, 1)␤ Expected: :($a)␤»
12:02 PacoAir joined 12:22 lizmat left, lizmat joined
dalek rlito: 413dc0d | (Flavio S. Glock)++ | / (3 files):
Perlito5 - perl5 - pretty-printer - number, apply
 12:24
12:24 lizmat left 12:28 lizmat joined
timotimo GlitchMr: i find it weird that you singled out heredocs allocating less cursors, when there were a whole bunch of commits removing all in all almost a million cursors (though some of those commits were in nqp) 12:35
or was that half a million? 12:37
dalek rlito: f861f2b | (Flavio S. Glock)++ | / (2 files):
Perlito5 - perl5 - pretty-printer - number, apply (fix)
 12:40
12:46 telex left 12:52 aindilis left 12:56 sisar joined
dalek kudo/nom: 889892a | (Elizabeth Mattijsen)++ | src/core/Baggy.pm:
Simplify Bag.grab|pick|roll, they're really just the same
 12:58
nwc10 good *, #perl6
timotimo (* *, *)("good", Whatever.new, "#perl6") 12:59
13:05 telex joined 13:20 baest_ left
masak good afty, #perl6 13:28
FROGGS o/
13:29 kathy joined, kathy is now known as kathyz
dalek kudo/nom: c6b15dc | (Elizabeth Mattijsen)++ | src/core/ (2 files):
Implement Bag|Mix.grabpairs, oddly enough exactly the same as Setty.grab
 13:34
grondilu maybe we need a PairContainer role or something? 13:36
moritz Associative? 13:37
masak r: say 40 + 2
camelia rakudo 889892: OUTPUT«42␤»
masak r: class Bird { method tweet { say "chirp!" } }; Bird.new.tweet 13:38
camelia rakudo 889892: OUTPUT«chirp!␤»
jnthn nqp-m: say('Moar frivolous evals!') 13:39
camelia nqp-moarvm: OUTPUT«Moar frivolous evals!␤»
masak nqp-moarvm! \o/
m: say('does this work, then?') 13:40
hm, I guess that should be reserved for Rakudo-on-Moar, anyway ;)
jnthn Maybe we should save that one for Rakudo on Moar :)
masak I think we should save that one for Rakudo on Moar.
13:41 kathyz left
diakopter let's save that one for Rakudo on Moar 13:41
TimToady: you think we should save that one for Rakudo on Moar?
tadzik we should save it for Rakudo/Moar 13:42
sisar I've been out of the perl6 loop for some time. Any progress on S09 implementation on any VM?
masak sisar: some, but not enough :)
sisar: when it does progress, it will on all VMs, I wager.
at least for Rakudo. 13:43
moritz sisar: rakudo-jvm and rakudo-parrot now both support packed native lists, and Buf/Blob are implemented as those
jnthn Well, Buf/Blob got improved which caused a bunch of the underlying work for native arrays to happen. And the postcircumfix changes happened which was also partly to help with native arrays :)
13:44 baest joined
diakopter guesses who kathyz is 13:44
grondilu r: my $b = Buf.new: (^10).roll(5); $b[4] = 0;
camelia rakudo 889892: OUTPUT«Cannot modify an immutable value␤ in block at /tmp/FwvL8Xd1s0:1␤␤»
grondilu n: my $b = Buf.new: (^10).roll(5); $b[4] = 0;
sisar masak, jnthn : i see.
camelia niecza v24-98-g473bd20: OUTPUT«Unhandled exception: Cannot call new; none of these signatures match:␤ Mu, *%_␤ at /tmp/pQmkQy7rp5 line 1 (mainline @ 4) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4583 (ANON @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4584 (m…»
jnthn diakopter: Given it ends in a z, I'd wager it's a Czech name :P
uh, Polish
argh ;)
grondilu Buf is still not mutable, though
jnthn grondilu: Yes, that's 'cus it needs the other piece we need for native arrays :) 13:45
diakopter guesses Chinese
sisar Is there a major raodblock to S09 implementation or is it just simply lack of tuits ?
(of course, any major hurdle can be broken with enough tuits :p ) 13:47
jnthn I guess partly lack of tuits; various bits of it need work at quite a few different levels of the stack. 13:48
sisar jnthn: ok
jnthn From my perspective, some of S09 falls into "nice to have" rather than "priority" also.
I suspect the native array stuff will happen in the not too distant future given most of the pieces are in place. 13:50
The PDL stuff I'd say is rather further out unless somebody jumps up to work on it.
sisar the R metaop, where is it _really_ useful ? Where does it really simplify an expression ? Can someone give a good example. 13:52
13:54 stevan_ joined
moritz sisar: usually for non-comunitative operators 13:54
r: say [R-] 1, 2, 10
camelia rakudo 889892: OUTPUT«9␤»
moritz uhm 13:55
r: say [-] 10, 2, 1
camelia rakudo 889892: OUTPUT«7␤»
moritz masak: wanna submit rakudobug? (at least I think that's one)
masak moritz: I'm just looking at it.
moritz: are you sure?
does R reverse the associativity? 13:56
jnthn not afaik
masak r: say (1 R- 2) R- 10
camelia rakudo 889892: OUTPUT«9␤»
sisar no, S03 says associativity is not reversed
masak r: say 1 R- (2 R- 10)
camelia rakudo 889892: OUTPUT«7␤»
masak moritz: so, non-bug.
jnthn STD doesn't reverse it either 13:57
sisar: Maybe it's useful in whatever-star cases
* / * is fine, but what if the things come in the wrong order? 13:58
* R/ * then does it
sisar jnthn: sorry, i still don't see it :( 14:00
where would such a thing come up ?
grondilu r: say 1 R- 2 R- 10
camelia rakudo 889892: OUTPUT«9␤»
grondilu r: say 10 - 2 - 1 14:01
camelia rakudo 889892: OUTPUT«7␤»
grondilu r: say 2 ** 3 ** 4 14:02
camelia rakudo 889892: OUTPUT«2417851639229258349412352␤»
sisar I'm looking at rosettacode.org/wiki/Runge-Kutta#Perl_6 , is using R at line 3 really helping/simplifying things ?
14:03 spider-mario left
moritz sisar: it means one doesn't need parens 14:03
sisar (I know it is a subjetive thing, but still )
grondilu sisar: I wrote this. I often write $n R/ [+] ... because this allows me to avoid using parens
lizmat cycling& 14:04
sisar moritz, grondilu : ok. I'll think about it. Thanks.
sisar wishes state scoped term definitions were implemented, then the Runge-Kutta code can have sigilless δy too 14:05
(at line 5 )
grondilu can't do that. Sigilless terms can not be rw variable 14:06
sisar grondilu: oh
grondilu you can write 'my \n = 42;' but then n is a constant. 14:07
though I guess there has to be ways to cheat 14:08
r: sub foo is rw { state $ }; foo = 4;
camelia rakudo 889892: OUTPUT«===SORRY!=== Error while compiling /tmp/SBZE274VQq␤Preceding context expects a term, but found infix = instead␤at /tmp/SBZE274VQq:1␤------> sub foo is rw { state $ }; foo =⏏ 4;␤»
grondilu r: sub foo is rw { state $ }; foo() = 4;
camelia ( no output )
grondilu r: sub foo is rw { state $ }; say foo() = 4; 14:09
camelia rakudo 889892: OUTPUT«4␤»
sisar grondilu: btw, output is slightly different from what is given. The printf "y(%2.0f) now gives output like y(0.) and y(1.) etc.
grondilu hum, let me check
ahh, I need to recompile rakudo first 14:10
r: printf("%d", pi) 14:12
camelia rakudo 889892: OUTPUT«3»
sisar r: printf "%2.0f", 2.000 14:13
camelia rakudo 889892: OUTPUT«2.»
grondilu maybe just replace %2.0f with %d
of %02d
I mean %2d
sisar grondilu: yup, i'll do that 14:14
sisar changes rosettacode
14:15 fridim_ joined
sisar done 14:15
grondilu sisar++ 14:17
14:18 stevan_ left
grondilu r: sub foo() is rw { state $ }; say foo() = 4; 14:20
camelia rakudo c6b15d: OUTPUT«4␤»
grondilu r: sub foo() is rw { state $ }; say foo = 4; 14:21
camelia rakudo c6b15d: OUTPUT«===SORRY!=== Error while compiling /tmp/JtiSdrO5M_␤Preceding context expects a term, but found infix = instead␤at /tmp/JtiSdrO5M_:1␤------> sub foo() is rw { state $ }; say foo =⏏ 4;␤»
grondilu r: sub term:foo() is rw { state $ }; say foo = 4; 14:22
camelia rakudo c6b15d: OUTPUT«===SORRY!===␤No such method 'ast' for invocant of type 'NQPMu'␤»
sisar grondilu: what does it mean when a sub is rw ?
grondilu it means that it returns a lvalue 14:23
diakopter "Rakudo on JVM is passing 99.28% of the spectests that Rakudo on Parrot does"
grondilu r: sub term:<foo>() is rw { state $ }; say foo = 4;
camelia rakudo c6b15d: OUTPUT«4␤»
diakopter what's the new number for this
sisar grondilu: ok, thanks.
tadzik we should change the remaning test descriptions to "I am the 1%"
diakopter "down with the 99%" 14:24
grondilu r: sub term:<foo>() is rw { state $ }; foo = 4; say ++foo 14:25
camelia rakudo c6b15d: OUTPUT«5␤»
jnthn diakopter: Don't know because something is wrong with the automated run
diakopter boo
grondilu sisar: so sigilless variables are possible, but it's a cheat
diakopter well is it below or above 100%
14:27 ajr joined 14:28 ajr is now known as Guest42553
jnthn diakopter: Probably a little below; once we start getting tests for the concurrency stuff I suspect it'll get ahead, though... 14:28
benabik r: my \foo; foo = 4; say ++foo 14:29
camelia rakudo c6b15d: OUTPUT«===SORRY!===␤No such method 'ast' for invocant of type 'NQPMu'␤»
benabik LTA error
14:30 Guest42553 is now known as ajr_, stevan_ joined
felher r: enum E < a b c>; my E $e = E.enums.keys[0] 14:32
camelia rakudo c6b15d: OUTPUT«Type check failed in assignment to '$e'; expected 'E' but got 'Str'␤ in block at /tmp/pbHmOWiv0g:1␤␤»
felher How do I get all E constants from the enum? E.enums.keys returns strings, not Es
jnthn r: enum E < a b c>; my E $e = E.enums.values[0] 14:33
camelia rakudo c6b15d: OUTPUT«Type check failed in assignment to '$e'; expected 'E' but got 'Int'␤ in block at /tmp/w5ZLw4llOT:1␤␤»
jnthn ah, yeah
r: enum E < a b c>; my E $e = E.^enum_value_list[0] 14:34
camelia ( no output )
jnthn But surely there's a better way... :)
Oh, E::.values works too I guess
r: enum E < a b c>; my E $e = E::.values[0]
camelia ( no output )
felher jnthn: okay, I'll use E::.values, thanks :) 14:35
jnthn yeah, that's a better way that relying on the enum MOP...
I wonder if there shouldn't be a better way though.
felher jnthn: how does E::.values work anyway? because E ist a package? 14:37
jnthn Yes
E:: gets the stash
felher jnthn: yeah, then I guess there ought to be a more explicit way. :) 14:38
14:44 BenGoldberg joined
dalek osystem: b2dc645 | (Pawel Pabian)++ | META.list:
Added Text::Markov
 14:47
osystem: e1158c6 | tadzik++ | META.list:
Merge pull request #27 from bbkr/master

Added Text::Markov
tadzik bbkr__++
moritz huh. 109s for rakudo-jvm to parse the setting 14:50
that's... SLOW
jnthn huh, yes
moritz considering that rakudo-parrot took only about 140s
jnthn It normally comes out well ahead here
moritz and that's a java-7-oracle JVM + javac 14:51
FROGGS I have 65s ommpared to 140s
jnthn 45s vs 100s here 14:52
FROGGS compared*
14:58 xenoterracide_ joined
dalek rlito: 46c565c | (Flavio S. Glock)++ | / (3 files):
Perlito5 - perl5 - pretty-printer - lookup, index
 15:00
[Coke] yawns. 15:10
hexcoder r: @@( <a b>, <c d> ) == @( <a b>, <c d> )
camelia ( no output )
hexcoder r: say @@( <a b>, <c d> ) == @( <a b>, <c d> )
camelia rakudo c6b15d: OUTPUT«True␤»
jnthn tries to work out what hexcoder is expecting that to mean and fails :) 15:11
hexcoder i am exploring slice context, but fail too.
:r my @a = @@( <a b>, <c d> ); say @a 15:12
FROGGS r: my @a = 'a' .. 'z'; say @a[0; 2] 15:13
camelia rakudo c6b15d: OUTPUT«WARNINGS:␤Useless use of constant integer 0 in sink context (line 1)␤c␤»
FROGGS r: my @a = ^10; say @a[0, 2]
camelia rakudo c6b15d: OUTPUT«0 2␤»
dalek rlito: 71da9bc | (Flavio S. Glock)++ | / (2 files):
Perlito5 - perl5 - pretty-printer - slices
FROGGS r: my @a = 'a'..'z'; say @a[0, 2] # that is what I meant 15:14
camelia rakudo c6b15d: OUTPUT«a c␤»
hexcoder no, i meant the slice contextualizer that moritz describes in his document Perl 5 to Perl 6 in chapter 'Context'
@() should flatten, and @@() should not. 15:15
FROGGS r: say @( <a b>, <c d> )
camelia rakudo c6b15d: OUTPUT«a b c d␤»
FROGGS hexcoder: I think @@() is not yet implemented
r: say @( <a b>, <c d> ).perl
camelia rakudo c6b15d: OUTPUT«(("a", "b"), ("c", "d")).list␤» 15:16
hexcoder ah, and suddenly everything makes sense :-))) thanks!!
FROGGS so, @() does the job of @@() atm
hexcoder: you're welcome :o)
hexcoder I agree, I could not even find it in the Synopsis. Or is it in already? 15:17
dalek rl6-roast-data: c35c60b | coke++ | / (5 files):
today (automated commit)
jnthn I think @@ went away, fwiw
FROGGS ahh, that explains it then :o) 15:19
masak greps the logs 15:20
yeah. it was removed in Feb 2010.
**@ is still alive and well, though.
hexcoder Thanks folks. I will tell Moritz... 15:21
jnthn yeah, but that ain't a conextualizer...
moritz hexcoder: did you read it on one of my websites?
hexcoder Yes
perlgeek.de/en/article/5-to-6
BenGoldberg r: say ^'z' 15:22
camelia rakudo c6b15d: OUTPUT«Cannot call 'Real'; none of these signatures match:␤:(Mu:U \v: Mu *%_)␤ in method Real at src/gen/CORE.setting:944␤ in method new at src/gen/CORE.setting:6814␤ in method new at src/gen/CORE.setting:6794␤ in sub prefix:<^> at src/gen/CORE.setting:6…»
hexcoder moritz: I will sent all of my findings combined to you. Stay tuned... 15:23
moritz stays tuned, ++hexcoder 15:24
[Coke] tries to think of a good way to debug this stupid rakudo.jvm failure on the test box. 15:25
[Coke] supposes he could attach visualvm and sit around for 2 hours.
jnthn [Coke]: Does it only happen under the eval server? 15:26
15:31 rindolf joined 15:37 kaare__ joined
[Coke] jnthn: that's an excellent thing to to test. 15:40
15:41 kaare_ left
[Coke] aha! 15:41
looks like this time my "log the eval serve" thing worked.
gist.github.com/coke/6855515 - OOME 15:42
of course, it has no useful details. :)
BenGoldberg r: my @a; say (@a xx *)[0]; 15:43
camelia rakudo c6b15d: OUTPUT«(timeout)»
BenGoldberg Should xx check if it's first argument is an empty list?
15:49 araujo left, araujo joined 15:50 sisar left
moritz [Coke]: you could check if t/spec/S02-types/array.t OOMs 15:57
[Coke]: that's the test file that causes problems on my boxes with parakudo 15:58
16:18 stevan_ left 16:22 xinming left 16:25 xinming joined, ssutch joined
masak BenGoldberg: seems to me `(() xx *)[0]` falls within "doctor, it hurts when I do this" territory. just like `loop {}` 16:34
TimToady r: say (() xx *).lol[^10].perl 16:36
camelia rakudo c6b15d: OUTPUT«(ListIter.new(), Any, Any, Any, Any, Any, Any, Any, Any, Any)␤»
TimToady lol should not flatten ()
moritz r: say (() xx *).lol[^10].elems
camelia rakudo c6b15d: OUTPUT«Inf␤»
moritz r: say (() xx *).lol[^10].tree.elems
camelia rakudo c6b15d: OUTPUT«Inf␤»
moritz r: say (() xx 5).lol.perl 16:37
camelia rakudo c6b15d: OUTPUT«LoL.new(ListIter.new())␤»
moritz r: say (() xx 5).lol.elems
camelia rakudo c6b15d: OUTPUT«1␤»
dalek rlito: 5a62cef | (Flavio S. Glock)++ | / (3 files):
Perlito5 - perl5 - pretty-printer - map
TimToady I'd think () xx 5 should return (),(),(),(),()
and lol should turn that into [],[],[],[],[] 16:38
grondilu thinks about 'loop {}' Something bugs him about it
TimToady or something equivalently itemy
loop is both a noun and a verb, so you can think of it either way 16:39
r: (loop {})[^10].perl.say 16:40
camelia rakudo c6b15d: OUTPUT«(timeout)»
grondilu currently, 'loop {}' hangs, right? I wonder if, for a language that is supposed to allow parallelism, it is not a bad idea. Maybe loop {} should do just nothing.
TimToady the imperative mindset is difficult to subvert :)
grondilu I guess it is
TimToady it should do the same thing as () xx * :)
by current spec, which people keep carping about :) 16:41
grondilu remembers that he once thought there should be an entire synopys dedicated to degenerate cases.
TimToady we do probably need to exempt loops at statementlist level from being lazy though 16:42
masak if you ask me, `loop {}` is behaving exactly as it should.
you ordered an infinite loop? well, you have it.
TimToady there's that imperative mindset again :)
grondilu :)
TimToady the question is not whether it's infinite, but whether it's lazy
jnthn loops kinda are imperative... :)
grondilu they are imperative 'inside' the loop. But from an external point of view, they are a member of a program that does not have to be imperative. 16:44
jnthn Maybe, but I think at least culturally loops are seen as an imperative construct, and the functional constructs are spelt more like "map" :) 16:45
masak the fact that I have an imperative mindset (and that I consider it "correct" here) was my entire point. 16:46
jnthn Heck, when I'm teaching Linq I even find myself saying "it's nice because we don't have a load of loops, just nice declarative code saying how to process the stream of things coming through the enumerable"... 16:47
16:48 diakopter left
BenGoldberg r: gather { loop {} }; 16:49
camelia rakudo c6b15d: OUTPUT«(timeout)»
BenGoldberg r: gather { loop {} }; 1;
grondilu wonders if loop {} is not the same kind of undefined notion as 0 * Inf
ingy want to be a Perl 6 programmer? so how many mindsets do you have? 3? hmmmm... stick with Perl 5
camelia rakudo c6b15d: OUTPUT«(timeout)»
BenGoldberg n: gather { loop {} }; 1;
camelia ( no output )
benabik Doesn't 0 * Inf = 0?
grondilu benabik: no 16:50
benabik r: print 0*Inf
camelia rakudo c6b15d: OUTPUT«NaN»
grondilu 0 * Inf is undefined.
BenGoldberg Why did gather { loop {} }; 1; hang in Rakudo?
rn: gather { die 'foo' }; 1; 16:51
camelia ( no output )
..rakudo c6b15d: OUTPUT«foo␤ in block at /tmp/lkkO7Msc9b:1␤␤»
BenGoldberg rn: gather { take 0; die 'foo' }; 1; 16:52
camelia ( no output )
..rakudo c6b15d: OUTPUT«foo␤ in block at /tmp/DlYXOBv0OP:1␤␤»
BenGoldberg rn: my @a := gather { take 0; die 'foo' }; @a;
camelia ( no output )
..rakudo c6b15d: OUTPUT«foo␤current instr.: 'throw' pc 402312 (src/gen/CORE.setting.pir:169118) (src/gen/CORE.setting:10504)␤called from Sub 'die' pc 42214 (src/gen/CORE.setting.pir:16280) (src/gen/CORE.setting:662)␤called from Sub 'die' pc 42113 (src/gen/CORE.setting.pir:16224)…»
BenGoldberg rn: my @a := gather { take 0; throw 'foo' }; @a;
camelia rakudo c6b15d: OUTPUT«===SORRY!=== Error while compiling /tmp/7HZWHGsU6d␤Undeclared routine:␤ throw used at line 1␤␤»
..niecza v24-98-g473bd20: OUTPUT«===SORRY!===␤␤Undeclared routine:␤ 'throw' used at line 1␤␤Unhandled exception: Check failed␤␤ at /home/p6eval/niecza/boot/lib/CORE.setting line 1502 (die @ 5) ␤ at /home/p6eval/niecza/src/STD.pm6 line 1147 (P6.comp_…»
BenGoldberg EINSUFFICIENTCAFFIENE 16:53
grondilu is fine with gather { take 0; die } dying.
BenGoldberg grondilu: in sink context? 16:54
grondilu yeah
though I should check what sink contect means
BenGoldberg Surely gather <anything> in sink context should be a no op.
It's like undef context in perl5 16:55
moritz there might still be side effects
BenGoldberg Err, !defined(wantarray) in perl5 :)
moritz yes, void context
lue hello world o/
BenGoldberg rn: my @a := map { die if $_ > 0 } ^Inf; say @a[0]; 16:56
camelia rakudo c6b15d: OUTPUT«Cannot call 'map'; none of these signatures match:␤:(&code, *@values)␤ in sub map at src/gen/CORE.setting:1719␤ in sub thread_junction at src/gen/CORE.setting:3123␤ in method AUTOTHREAD at src/gen/CORE.setting:3151␤ in sub map at src/gen/CORE.sett…»
..niecza v24-98-g473bd20: OUTPUT«(Any)␤»
BenGoldberg rn: my @a := map { die if $_ > 0 }, ^Inf; say @a[0];
16:56 zakharyas joined
camelia rakudo c6b15d: OUTPUT«␤ in block at /tmp/VYAgyhd1Sq:1␤␤» 16:56
..niecza v24-98-g473bd20: OUTPUT«␤»
grondilu rn: my @a := gather { take 1; say "bye" }; @a 16:57
BenGoldberg rn: map { die if $_ > 0 }, ^Inf;
camelia ( no output )
..rakudo c6b15d: OUTPUT«bye␤»
( no output )
..rakudo c6b15d: OUTPUT«No exception handler and no message␤␤current instr.: 'throw' pc 402312 (src/gen/CORE.setting.pir:169118) (src/gen/CORE.setting:10504)␤called from Sub 'die' pc 42298 (src/gen/CORE.setting.pir:16317) (src/gen/CORE.setting:665)␤called from Sub '' pc 156 ((…»
grondilu rn: my @a := 1..*; @a 16:58
camelia ( no output )
grondilu rn: my @a := gather for 1..* { .take; say "." }; @a 16:59
camelia ( no output )
..rakudo c6b15d: OUTPUT«(timeout).␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤.␤…»
grondilu BenGoldberg: you might be right. There should not be evaluations in sink context. 17:00
moritz grondilu: that's contrary to our current philosphy
which is: if you evaluate something in sink context, it's for its side effects, so it must be eagerly evaluated 17:01
grondilu ok
dalek rlito: 2bdccfe | (Flavio S. Glock)++ | / (4 files):
Perlito5 - perl5 - pretty-printer is the default perl5 backend
 17:04
hexcoder moritz: which of your email addresses should I use for feedback on 'Perl 5 to Perl6'? 17:21
17:24 diakopter joined
moritz hexcoder: [email@hidden.address] 17:29
dalek rlito: ac48c65 | (Flavio S. Glock)++ | / (2 files):
Perlito5 - perl5 - pretty-printer - more operators
 17:30
rlito: 8398d27 | (Flavio S. Glock)++ | / (2 files):
Perlito5 - perl5 - pretty-printer - tweak operators
hexcoder :moritz done 17:31
17:32 stevan__ joined 17:33 btyler joined, ajr_ left 17:34 ajr joined 17:35 ajr is now known as Guest44737, Guest44737 is now known as ajr_ 17:58 araujo left 17:59 araujo joined, Rotwang joined 18:02 ajr_ left 18:04 Rotwang left 18:20 SamuraiJack__ left
ingy sorear: ping 18:27
dalek ast: af7277f | (Elizabeth Mattijsen)++ | S02-types/ (3 files):
Add tests for .grabpairs where they are not supposed to work
 18:29
18:29 ajr_ joined 18:30 darutoko left 18:32 rindolf left
BenGoldberg rn: my @a := gather { take 1; say 'bye' }; @a.eager; 18:38
camelia rakudo c6b15d, niecza v24-98-g473bd20: OUTPUT«bye␤»
moritz r: 'abcd'.substr(my @a = 1, 4).say 18:42
camelia rakudo c6b15d: OUTPUT«cd␤»
FROGGS ? 18:43
jnthn that's like .substr(2) I guess
lizmat final value of the parcel ? feels like a p5ism
p5eval: my @a=(1,2,3,4); say $a[my @b=1,3] 18:45
p5eval lizmat: 41
lizmat p5eval: my @a=(1,2,3,4); $a[my @b=1,3]
p5eval lizmat: 4
FROGGS ahh, now I see it 18:46
xinming I saw the methdo ucfirst is renamed to tc for Str, What does tc mean here please? 18:55
jnthn xinming: title case 18:56
xinming got it, thanks
dalek rlito: 18ea7c3 | (Flavio S. Glock)++ | / (2 files):
Perlito5 - perl5 - pretty-printer - elsif
 19:05
masak lizmat: it's not final value of a parcel. it's its length. 19:09
lizmat p5eval: my @a=(1,2,3,4); say $a[my @b=1,2] 19:10
p5eval lizmat: 31
lizmat ok, so it's not a p5ism
masak I'm really glad we didn't inherit that one :) 19:13
r: say 'abcd'.substr(my @a = 1, 4); say 'abcd'.substr(2) 19:14
camelia rakudo c6b15d: OUTPUT«cd␤cd␤»
masak r: say [1, 4].elems
camelia rakudo c6b15d: OUTPUT«2␤»
masak r: say +[1, 4]
camelia rakudo c6b15d: OUTPUT«2␤»
masak r: say +my @a = 1, 4
camelia rakudo c6b15d: OUTPUT«2␤»
19:18 zakharyas left, sisar joined 19:32 Rotwang joined
lizmat r: my $b=<a b b>.BagHash; $b.grab(*).say; $b.say 19:33
camelia rakudo c6b15d: OUTPUT«b b a␤BagHash.new(a(0), b(0))␤»
lizmat oops, grab doesn't clean up 19:34
dalek kudo/nom: 08d0f14 | (Elizabeth Mattijsen)++ | src/core/Baggy.pm:
Make sure Baggy.grab cleans up after itself
 19:35
lizmat TimToady: on the subject of typed Bag/Mix: the problem is that rakudo currently cannot serialize any typed hashes/arrays 19:38
jnthn is aware of the pb, but for now is low on the list 19:39
jnthn lizmat: We determiend it wasn't just about the setting also, iirc? 19:45
lizmat jnthn is aware of the pb, but for now is low on the list 19:47
Stage optimize : 17.719
Code ref 'Str' does not exist in serialization context
so it happens in stage post
you want me to try this with e.g. lib/Test.pm and see what happens ? 19:48
19:49 japhb_ left
dalek kudo/nom: 2e2f5f0 | (Elizabeth Mattijsen)++ | src/core/Setty.pm:
Implement Setty.grabpairs for consistency
 19:50
kudo/nom: 7b98f86 | (Elizabeth Mattijsen)++ | src/core/ (3 files):
Implement .grabpairs where they're *not* supposed to work
moritz hexcoder++ # very good feedback on 5-to-6 19:52
japhb__ jnthn, did you have any thoughts on the multi problem I rambled about last night?
FROGGS gnight
lizmat jnthn: it *only* appears to be happening when compiling the setting 19:53
japhb__ gnight FROGGS
19:53 FROGGS left
lizmat adding a typed hash spec to lib/Test.pm and doing a "./perl6 --stagestats --target=pir --output=lib/Test.pir lib/Test.pm" just works fine 19:53
jnthn japhb__: Well, it's a language design issue rather than Rakudo not implementing current spec wrt multis, afaik. And yeah, it's a tricky issue...but multi-dispatch candidate lists being known lexically is *really* important for optimization, so we can't break that. 19:55
japhb__: Which somewhat constrains the solution space...
19:56 araujo left
ivanshmakov S17 mentions continuations, but doesn’t seem to either describe them itself, or point to another document describing them. Where do I find more information on that? perlcabal.org/syn/S17.html 19:58
20:01 japhb_ joined
lue ivanshmakov: unfortunately (if I remember correctly), S17 is severely undeveloped, so the stuff described inside is due for a lot more work. 20:02
They do seem to be mentioned once in S09, and a few times in S04 however.
lizmat indeed, S17 is mostly ~2005 ramblings, by yours truly among others :-( 20:03
20:03 araujo joined
lizmat is looking forward to get back to that once moarvm lands 20:03
jnthn Well, S17 probably wants to become a description of the stuff that's actually implemented so far :) 20:04
20:04 dmol left
jnthn The answer on continuations, though, is that Perl 6 doesn't expose them at user level. 20:04
The closest is gather/take, which is coroutine-powerful. 20:05
lizmat jnthn: that's what I mean :-)
jnthn lizmat: Yeah, but no need to wait for MoarVM to evolve that. ;) 20:06
perigrin ivanshmakov: the relatively ancient www.sidhe.org/~dan/blog/archives/000156.html helped me grasp the concept of continuations.
20:07 dmol joined, PacoAir left
arnsholt jnthn: NQP inherits all of its CFLAGS from Parrot, right? 20:07
jnthn yeah 20:08
Believe so
arnsholt 'Cause then I can punt #33 back to Parrot 20:09
lue That reminds me, can anyone tell me why my long-ago change to syn_index.html to have a link to S99 still doesn't show up on either feather's or perlcabal's synopses indexes? 20:10
arnsholt Not that compiler warnings are that big of a deal anyways
lue (it should appear directly underneath the row for S33 fwiw) 20:12
japhb__ jnthn: How much does it weaken the multi optimizations if we drop from "absolute knowledge of all applicable candidates" to "the list of applicable candidates changes rarely"? 20:15
20:18 araujo left, araujo joined, sisar left
lue
.oO(is it possible the index updater is not running? Both sites' indexes say they were last modified Wed 03 Apr 2013 02:01:04 AM PDT)
 20:19
jnthn japhb__: Means we can't inline at compile time.
japhb__: So a $native_int + $native_int would have to compile down to a multi-dispatch and we are at the mercy of JIT compilers doing inlines to make adding two native integers together work out, rather than code-generating a simple add instruction, which is what happens today with compile-time inlining. 20:21
japhb__ jnthn: What about a fast generation check that fails out of the straight line path, and succeeds with fallthrough to code generated on the assumption that it hasn't changed? Hmmm, come to think of it, for native ops, that's still a heckuva lot of overhead ... 20:22
Kinda does assume the existance of a JIT that can hoist that check higher, though. 20:24
dalek ast: 8379851 | (Elizabeth Mattijsen)++ | S02-types/ (3 files):
Add tests for where .grabpairs *should* work
 20:31
lizmat another ~70 tests or so
japhb__ TimToady: What are your thoughts on language decisions that penalize interpreters? "Heck no!" ? "It depends"? 20:32
jnthn japhb__: Yeah, and while we *may* be able to use invokedynamic to each the JVM's JIT enough to actually do it with Perl 6 multi-dispatch...and of course we can do as we please in Moar...anything else is a bit more tricky 20:35
20:36 PacoAir joined 20:38 PacoAir left 20:42 dmol left, dmol joined 20:45 stevan__ left
japhb__ Am I correct in assuming that no later-added multi candidate can contribute to an op that is *called* in the setting? 20:46
lizmat depends, afaik: if you add a multi-method to Any, it will only work on objects that *are* Any, not on any of its subclasses 20:48
jnthn The setting is the outermost lexical scope of the program, so what's visible there is what they see. Note that "later-added" doesn't actually add to the multi candidate lists of the operators as the setting sees them; rather, the dispatcher is derived...
lizmat: We're doing multi subs here...
lizmat ah, ok… :-) 20:49
lizmat goes back to lurk mode
japhb__ jnthn: Yeah, that's what I thought. I just wanted to make sure I wasn't being pessimistic on account of an incorrect mental picture. 20:51
20:54 ajr_ left
hexcoder privmsg camelia 20:59
privmsg camelia :macro circumfix:«< >» ($text) is parsed / <-[>]+> / {return $text.comb(rx/\S+/);} 21:02
grondilu is parsed is NYI, isn't it?
r: macro postfix:<????> is parsed { } 21:03
camelia rakudo 7b98f8: OUTPUT«===SORRY!=== Error while compiling /tmp/C6Cr6Q03pm␤Can't use unknown trait 'is parsed' in a macro+{precedence} declaration.␤at /tmp/C6Cr6Q03pm:1␤------> ␤ expecting any of:␤ rw parcel hidden_from_backtrace␤ pure de…»
hexcoder might be, but rakudo expects a block after that, so I thought it understands it.
21:04 skids left
grondilu it doesn't, but it recognizes it as a trait 21:04
hexcoder aha, I see.
21:10 dmol left 21:12 Rotwang left 21:16 skids joined 21:21 hummeleB1 left 21:22 kaare__ left
dalek kudo/nom: d45e4da | (Elizabeth Mattijsen)++ | src/core/Backtrace.pm:
Add :named so that we can filter on named subs/methods
 22:06
22:12 stevan_ joined 22:31 hexcoder left 22:36 stevan_ left
dalek p/moarboot: 431c3b1 | jnthn++ | src/vm/moar/ (3 files):
Properly handle --bootstrap on MoarVM backend.
 22:52
p/moarboot: 3f8a23c | jnthn++ | tools/build/Makefile-Moar.in:
Correct backend name in Moar makefile.
p/moarboot: ff77d36 | jnthn++ | src/vm/moar/stage0/ (9 files):
Update stage0 for MoarVM to get various fixes.
p/moarboot: 78f4d64 | jnthn++ | src/NQP/Compiler.nqp:
Tweak so that stage1 actually starts up.

Fails upon calling the first action method, however; some code-object related issue.
kudo/nom: 9947610 | (Elizabeth Mattijsen)++ | src/core/Backtrace.pm:
Add a .package method for convenience and abstraction
 23:02
lizmat sleep& 23:16
timotimo gnite lizmat
23:20 grondilu left 23:33 JJesus joined 23:41 cognominal left 23:47 dwarring joined, JJesus left

Powered by Cro and the Raku® Programming Language.

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

Thank you!