»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.
lizmat r: say (1/3).perl # this doesn't feel right 00:02
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«<1/3>␤»
lue r: say <1/3>.WHAT 00:03
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«(Str)␤»
lue rn: say <1/3>.WHAT
lizmat r: say (1/3).gist # this feels even worse
camelia niecza v24-98-g473bd20: OUTPUT«(Rat)␤»
..rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«(Str)␤»
rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«0.333333␤»
lue lizmat: <1/3> is supposed to be a RatStr. Although it should perhaps be just a Rat coming out of .perl 00:04
lizmat r: say (1/3).perl.eval # oddly enough, this *seems* to round trip 00:05
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«1/3␤»
lizmat r: say (1/3).perl.eval.WHAT # but alas, it is now a string
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«(Str)␤»
lue r: say <1/3>.perl; say +"1/3";
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«"1/3"␤0.333333␤»
lue r: say <1/3> + <2/3> 00:06
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«1␤»
00:06 synopsebot left
lizmat r: say 1/3 + 2/3 00:07
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«1␤»
00:07 hugme left
lizmat r: say 1 ÷ 3 00:07
camelia rakudo-parrot 874e35: OUTPUT«===SORRY!=== Error while compiling /tmp/RhIwd_xm2g␤Two terms in a row␤at /tmp/RhIwd_xm2g:1␤------> say 1 ⏏÷ 3␤ expecting any of:␤ postfix␤ infix stopper␤ infix or meta-infix␤…»
..rakudo-jvm 874e35: OUTPUT«===SORRY!=== Error while compiling /tmp/rtJR9UfTUC␤Two terms in a row␤at /tmp/rtJR9UfTUC:1␤------> say 1 ⏏÷ 3␤ expecting any of:␤ postfix␤ infix stopper␤ infix or meta-infix␤ …»
lue I forgot why it's RatStr instead of just a Str.
lizmat I don't thing we have any of the xxxStr thingies yet
lue lizmat: yeah, I'm wondering why we have them spec'd. I'm sure S02 will remind me. 00:08
jnthn val(...) is meant to produce them, iirc 00:09
lizmat preflex: .u ÷ 00:10
alas :-(
00:10 Juerd left
lue yep. It seems the only value to the *Str types is to have a "true" literal for Rat and Complex, even though it and constant folding differ only by the brackets used. 00:11
00:11 Juerd joined
jnthn lue: Also, information preservation, iirc 00:11
00:12 synopsebot joined
jercos Can pass an IntStr as an Int, wheras you can't a Str? 00:12
lue That could be. I think my point is that the *Str types are NYI possibly because coercion takes care of where it would otherwise be an issue :)
(responding to jnthn)
00:13 hugme joined, ChanServ sets mode: +v hugme
jercos programming: where coercion is considered better than inheritance. 00:13
00:15 beastd left
lue I'm not totally convinced that RatStr is better than Rat.Str when it comes to information preservation. But I can't say for sure the values of *Str (something to ask TimToady). 00:16
jercos beats making everything a FatRat :|
lue There's no FatRatStr though, AFAICT. If you really need a FatRat, it's gonna be a FatRat. 00:17
jercos Well I mean there doesn't need to be a FatRatStr though, because Strs aren't inherently size-bounded
lue The Str part of RatStr could handle a FatRat in string form, sure, but the Rat part couldn't. 00:18
jercos A RatStr should be able to behave like a Rat, but then when cast to a FatRat, retain the precision of the original string.
lue FatRats are already more precise than Rats, unless I'm misremembering. 00:19
jercos exactly. 00:20
RatStr and FatRat both have infinite potential precision, but a RatStr will degrade to a Rat, while a FatRat will upgrade Rats. 00:21
lue jercos: if the Str of RatStr is somehow more precise than the Rat part, that means the Str's contents don't fit in a Rat, thus a RatStr can't happen.
jercos: RatStrs are limited by the Rat part.
(because whatever value a potential RatStr is given must fit into both parts.) 00:22
jercos perlcabal.org/syn/S02.html#Allomorphic_Rats seems to imply otherwise
lue What the...? 00:24
r: constant pi is export = 3.14159_26535_89793_23846_26433_83279_50288; say pi.norm.nude; say pi.WHAT; 00:25
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«19634954084936207740391521145496893 6250000000000000000000000000000000␤(Rat)␤»
lue r: constant pi is export = 3.14159_26535_89793_23846_26433_83279_50288; say pi.Str; 00:26
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«3.14159265358979322257937383255921304␤»
lue r: constant pi is export = 3.14159_26535_89793_23846_26433_83279_50288; say pi.perl;
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«3.14159265358979322257937383255921304␤»
lizmat r: say 14/3.perl; say (14/3).perl # should the () really make that difference ? 00:28
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«4.666667␤<14/3>␤»
lizmat n: say 14/3.perl; say (14/3).perl # should the () really make that difference ?
camelia niecza v24-98-g473bd20: OUTPUT«4.666667␤<14/3>␤»
lue Rats (aka Rational[Int,Uint64]) can't overflow the numerator, right‽ 00:29
jnthn lizmat: Yes, 'cus method calls bind tighter
lue r: my $a = 0xFFFFFFFF_FFFFFFFF; say $a.WHAT; $a++; say $a.WHAT;
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«(Int)␤(Int)␤» 00:30
lizmat jnthn: so it binds to the 3 ?
lue jnthn: Ints don't overflow, right?
jnthn it's doing 14/(3.perl)
lue: Correct.
lizmat jnthn: gotcha
lue Which means this section of S02 never triggers: perlcabal.org/syn/S02.html#Allomorphic_Rats
lizmat wishes #perl6 a good * 00:31
lue (Even if that "numerator" was supposed to be "denominator", I can't quite parse what the hell that section of S02 is trying to say)
jnthn lue: Int doesn't overlfow, no. Though of course if you have a big enough number you can use all the RAM... ;)
lue good ♞ lizmat o/
jnthn No, the point of Rat is the numerator is ounboudned precision, but the denominator will never be more than a 64-bit int. 00:32
lue :) Tell me jnthn, does the section I linked make much sense, even with s/numer/denomin/
That is, too large Rats become *RatStr*s? Not FatRats or Nums?
jnthn lue: It makes sense to me. 00:33
lue: Correct, they become RatStr so you get the math performance you expect, but if you really meant to have all the precision and are willing to pay the performance penalty, you can recover the full precision of the constant.
lue gaaaah. My concept of the Rat/FatRat divide has been broken. I only see that if a Str that contains a too-precise Rat gets passed through, say, MAIN($args), but in that case a FatRatStr would make more sense. 00:34
jnthn FatRat can become *very* slow depending what you're doing.
No, the idea is to default to (cheaper) Rats but keep enough info around to get a FatRat built if needed. 00:35
lue would muchly prefer it become a FatRat or a Num.
If only to keep my sanity :)
lue fixes the "numerator" flub in any case. 00:36
jnthn Um, I think it may be right?
Though I find it weird
I'd just keep the original number text as the string... :)
colomon you can't overflow a numerator in a Rat 00:37
which I assume is what lue is fixing?
lue That section never triggers if I don't change it.
Mouq r: say (4/(2**64)).WHAT 00:38
r: say (4/(2**64+1)).WHAT
jnthn oh, I see what you mean...
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«(Rat)␤»
rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«(Num)␤»
jnthn Yeah, I misread how it was worded. Think I agree now :)
lue I like what rakudo's doing right now :) 00:39
Mouq r: say ((2**64+1)/5).WHAT
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«(Rat)␤»
colomon lue: what Mouq typed there isn't a literal. the rules only apply to literals.
00:39 bluescreen100 left
colomon (the "allomorphic rats" rules, I mean.) 00:39
lue forgets that it doesn't count as a Num unless e gets involved. 00:40
dalek ecs: b44eb86 | lue++ | S02-bits.pod:
[S02] Change numerator->denominator in Allomorphic Rats.

That way this section of the spec will ever do something, because you can't overflow a Rat's numerator.
colomon lue++ 00:41
lue colomon: then I like the way the constant-folding method does it :)
rn: say RatStr; 00:42
camelia rakudo-parrot 874e35: OUTPUT«===SORRY!=== Error while compiling /tmp/HB_OSPz_m1␤Undeclared name:␤ RatStr used at line 1␤␤»
..rakudo-jvm 874e35: OUTPUT«===SORRY!=== Error while compiling /tmp/JidAHc69St␤Undeclared name:␤ RatStr used at line 1␤␤»
..niecza v24-98-g473bd20: OUTPUT«(RatStr)␤»
colomon rakudo doesn't know RatStr.
lue n: say val("1/3").perl;
colomon n: say 3.14159_26535_89793_23846_26433_83279_50288.WHAT
camelia niecza v24-98-g473bd20: OUTPUT«val("1/3")␤»
niecza v24-98-g473bd20: OUTPUT«Num()␤»
colomon huh
lue n: say val("1/3").WHAT;
camelia niecza v24-98-g473bd20: OUTPUT«(RatStr)␤»
colomon n: say 3.14159_26535_89793_23846_26433_83279_50288
camelia niecza v24-98-g473bd20: OUTPUT«3.1415926535897931␤»
colomon n: say 3.14159_26535_89793_23846_26433_83279_50288.perl 00:43
camelia niecza v24-98-g473bd20: OUTPUT«3.1415926535897931e0␤»
lue n: say val("1/(2**64)").WHAT; say val("1/(2**64)")
camelia niecza v24-98-g473bd20: OUTPUT«(Str)␤1/(2**64)␤»
lue n: say val("1/{2**64}").WHAT; say val("1/{2**64}")
camelia niecza v24-98-g473bd20: OUTPUT«Num()␤1/18446744073709551616␤»
lue n: say val("1/{2**32}").WHAT; say val("1/{2**32}")
colomon gack, guess it's another bug for me to fix on that far-off day when I have time to niecza-hack.
camelia niecza v24-98-g473bd20: OUTPUT«(RatStr)␤1/4294967296␤» 00:44
jnthn sleep time for me... &
lue n: say val("1/{2**32+1}").WHAT; say val("1/{2**32+1}")
camelia niecza v24-98-g473bd20: OUTPUT«(RatStr)␤1/4294967297␤»
lue ♞ jnthn o/
colomon \o
lue I personally would prefer if too-big-for-Rat decimal literals were Num instead of FatRat. Because if you typed out a very big decimal number, chances are you're not thinking in fractions :) 00:45
(nor RatStr, for that matter.)
colomon but RatStr is the only choice that doesn't involve throwing away much of what you type 00:46
lue there's also NumStr, no?
colomon r: say 2**64
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«18446744073709551616␤»
colomon r: say (2**64).split 00:47
camelia rakudo-jvm 874e35: OUTPUT«Cannot call 'split'; none of these signatures match:␤:(Cool : Regex $pat, $limit = { ... }, :all(:$all), Mu *%_)␤:(Cool : Cool $pat, $limit = { ... }, :all(:$all), Mu *%_)␤ in any at gen/jvm/BOOTSTRAP.nqp:1201␤ in block at /tmp/jJCLdNnYmr:1␤ …»
..rakudo-parrot 874e35: OUTPUT«Cannot call 'split'; none of these signatures match:␤:(Cool : Regex $pat, $limit = { ... }, :all(:$all), Mu *%_)␤:(Cool : Cool $pat, $limit = { ... }, :all(:$all), Mu *%_)␤ in any at gen/parrot/BOOTSTRAP.nqp:1218␤ in any at gen/parrot/BOOTSTR…»
colomon r: say (2**64).comb
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«1 8 4 4 6 7 4 4 0 7 3 7 0 9 5 5 1 6 1 6␤»
lue (Doesn't bother me too much either way, though.)
colomon r: say +((2**64).comb
camelia rakudo-parrot 874e35: OUTPUT«===SORRY!=== Error while compiling /tmp/d6plr6Rnic␤Unable to parse expression in parenthesized expression; couldn't find final ')' ␤at /tmp/d6plr6Rnic:1␤------> say +((2**64).comb⏏<EOL>␤ expecting any of:…»
..rakudo-jvm 874e35: OUTPUT«===SORRY!=== Error while compiling /tmp/5_ERMeEiWn␤Unable to parse expression in parenthesized expression; couldn't find final ')' ␤at /tmp/5_ERMeEiWn:1␤------> say +((2**64).comb⏏<EOL>␤ expecting any of:␤…»
colomon r: say +((2**64).comb)
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«20␤»
colomon what I was trying (unsuccessfully) to get at there is, unless your number starts with a load of zeros, Rat is actually more accurate than Num. 00:48
so RatStr gives you something reasonable to work with, and allows you to get at any other form you'd want, with the possibility of even more accuracy. 00:49
(for FatRat conversion.)
lue Yeah, I just don't like how the Rat/FatRat divide I've always imagined gets utterly broken when it comes our ol' friend *Str. 00:50
*friends rather
I'd be stubborn and point out that FatRat supports the criteria of "precision" without allomorphic trickery, but I think its speed disadvantage has been stated well enough :) 00:53
lue & 00:54
00:54 mschreck left 00:55 mschreck joined
jercos lue: clearly we should just eschew efficiency all together, and store all numbers as rationals made of products of prime powers. :p 01:01
Mouq jercos: an to make things simple, we should store the primes using unary arithmetic, e.g., 5 gets 5 bits, 17 gets 17 01:03
*and
jercos ah, well, if you're storing prime powers, the primes themselves don't really get any representation in the stored form 01:05
just taxation
(3, 1, 1)/(0, 0, 0, 1) = 2 * 2 * 2 * 3 * 5 / 7 = 120/7 = ~17.142857 01:07
Mouq oh, duh. but of course we still need the numbers to tell which number prime we're talking about. but our primes can get arbitrarily big! We should store the primes using the same PrimeRat class we're storing the primes in!
colomon lue: just remember you *never* get a FatRat unless you explicitly ask for one. or you have one already. 01:08
jercos Mouq: I actually implemented that for laughs while playing around with dynamically generated arrays. I still don't have a good way to make a magic array of primes that continues extending indefinitely as it's accessed, without being at least as slow as testing is-prime for every number between the last known prime and the prime index requested. :| 01:12
For bonus points, the 0th prime number is taken to be -1, so negative numbers can be represented. >.>
Mouq jercos: sieve? 01:13
jercos a sieve would be nice, but that' 01:14
s rather hard to fit inside the closure for ...
01:17 dwoldrich joined, benabik joined, benabik left
jercos plus, I'd rather not take up n array elements to find a prime <= n. 01:17
memory is cheap, but not that cheap :| 01:18
01:28 dayangkun joined
timotimo rakudo is still pretty wasteful with regards to memory, it seems 01:34
but when moarvm comes around, that may change 01:35
also, sized arrays are in soon, after that we have native types as lvalues and then compact arrays
01:39 mschreck_ joined
timotimo that's the hope anyway :) 01:40
01:40 mschreck left 01:42 Grrrr joined 01:44 MikeFair_ joined 01:46 mschreck_ left 01:55 JimmyZ_ joined, JimmyZ_ left
lue colomon: that's a good thing to keep in mind. I still don't like how RatStr is special from the other *Str types though :) . 02:04
colomon lue: you wouldn't happen to remember how to pass a hash as a bunch of named arguments, would you? 02:19
02:19 ggoebel110 left
lue r: sub foo(:$bar, :$baz) { say "$bar $baz" }; my %h = <bar 1 baz 2>; foo(|%h); 02:20
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«1 2␤»
colomon brill!
lue++
lue r: sub foo(:$bar, :$baz) { say "$bar $baz" }; my %h = <bar 1 baz 2>; foo(%h);
camelia rakudo-jvm 874e35: OUTPUT«Too many positional parameters passed; got 1 but expected 0␤ in block at /tmp/X2wcpACsGg:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:1086␤ in any evalfiles at gen/jvm/stage2/NQPHLL.nqp:1292␤ in any command_eval at gen/jvm/stage2/NQPHLL.nqp:1196…»
..rakudo-parrot 874e35: OUTPUT«Too many positional parameters passed; got 1 but expected 0␤ in sub foo at /tmp/UFpjckVPNI:1␤ in block at /tmp/UFpjckVPNI:1␤ in any at /tmp/UFpjckVPNI:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPH…»
lue you're welcome :) 02:21
02:22 mschreck joined 02:23 xenoterracide left
colomon not actually helping my code. :\ 02:24
02:30 MikeFair_ is now known as MikeFair 02:39 ggoebel110 joined 02:44 mschreck left 02:52 mschreck joined 02:54 mschreck left, mschreck joined
colomon It's rapidly becoming apparent that when IO::Prompter last worked, Rakudo didn't actually check parameter types. 02:57
03:00 mschreck left 03:07 kaleem joined 03:08 SevenWolf left
colomon r: say $*IN.WHAT 03:08
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«(IO::Handle)␤»
03:16 Grrrr left 03:20 xenoterracide joined 03:22 Grrrr joined
colomon triumph! 03:31
03:35 Rix joined 03:36 mschreck joined 03:44 mschreck_ joined, mschreck left, mschreck_ is now known as mschreck 03:45 kaleem left 04:01 mschreck left 04:02 cxreg2 is now known as cxreg 04:03 mschreck joined 04:09 mschreck left 04:49 ssutch joined 04:52 zakharyas joined 04:53 preflex left 04:54 preflex joined, ChanServ sets mode: +v preflex 04:58 mschreck joined 05:07 jlaire joined 05:13 Mouq left 05:14 ggoebel110 left 05:26 mschreck left 05:27 [Sno] left, mschreck joined, SevenWolf joined 05:29 dayangkun left
TimToady <1/3> is always just a Rat; the RatStrs don't show up unless there is whitespace 05:43
numeric operations *never* produce RatStr
if you overflow a denominator on a computation, you'll either get a rounded Rat or a Num 05:44
IntStr, NumStr, and RatStr are there only for stuff that is ambiguous because it's coming through a text-only API, such as literal code, or command line, or terminal input 05:45
and they are only for hedging your bets until the first numeric operation 05:46
05:49 zakharyas left
lue TimToady: that Allomorphic Rats example implied that a long decimal causes RatStr though. 05:49
So is that section wrong no matter which half of the fraction is stated to overflow?
TimToady your program is *text* 05:51
lue (if overflow gets rounded Rat or Num, that is)
TimToady it's not numeric unless something makes it numeric
05:52 fridim__ left, mschreck left
TimToady nothing trying to make it overflow, that's just the defintion of the constant 05:52
a RatStr
05:52 mschreck joined
TimToady the RatStr is the only conversion that doesn't lose precision 05:53
s/RatStr/FatRat/
lue
.oO(The statement "your program is text" feels a lot like "you're surrounded by air", in that it's something you normally forget ☺)
TimToady well, and .Str, but that's not really a conversion
lue so that constant pi = <long decimal> is supposed to be a RatStr because the program itself is text, IIUC. 05:54
TimToady yes
it's one of those less obvious places where Perl 6 solves a problem with premature commitment by being a bit lazier than the average langauge 05:55
*guage
05:56 BenGoldberg left
TimToady though MI is not the only way to solve this; we could say that all numeric types derive from Str instead, but that has its own issues 05:56
violating Liskov, for one 05:57
lue Now that I realize there's air surrounding me, the decision to make too-large-for-Rat literals be RatStr makes a bit more sense. /me imagines "+<large decimal>" would force that rounded Rat or Num condition 05:58
06:00 mschreck left
TimToady on the use of <1/3> for a literal notation, it just kinda falls out of wanting to represent a list of numbers inside <> anyway; also the precedence issue if you forget the parens around (1/3) are worse for complex literals, which use the looser + 06:01
it just felt conceptually cleaner not to rely on constant folding for those kinds of literals 06:02
06:02 mschreck joined
lue Too bad there's nothing like 2.5 for Complex numbers though; 1i2 or something... 06:02
TimToady once you have val() as a "safe" eval, it seems reasonable to use it
n: say <1i2>.WHAT 06:03
camelia niecza v24-98-g473bd20: OUTPUT«(Str)␤»
TimToady n: say <1i+2>.WHAT
lue
.oO( eval == exciting val() )
camelia niecza v24-98-g473bd20: OUTPUT«(Str)␤»
TimToady n: say <1+2i>.WHAT
camelia niecza v24-98-g473bd20: OUTPUT«(Complex)␤»
TimToady 1i2 would be terribly ambigous; does the i go on the 1 or on the 2?
and as long as <> was getting val semantics anyway, might as well use the spaceless form as a literal 06:04
arguably, we're overloading brackets more than we ought, but there just aren't enough brackets in ASCII 06:05
lue yeah. I was just lamenting mathematics in general for using operators as the most basic representation of a type of number. 'i' could be any kind of separator really. I've thought of 1;2 as that before too.
TimToady we still haven't achieved the heights of Perl 5's overloading of {}
which as I recall meant at least six different things
06:07 mschreck left
diakopter eval {}{}{}{}{}{}{}{}{} 06:07
p5eval {}{}{}{}{}{}{}{}{}
TimToady block, hash, "${foo}bar", \x{dead}, /x{1,2}/, q{}, and I think something in formats 06:08
06:08 mschreck joined
TimToady and probably a few other things I've forgotten 06:08
diakopter lexical scope?
TimToady block
diakopter oh
TimToady well, {} could also be literal in a regex if it didn't have numbers 06:10
oh, duh, hash subscripts
diakopter s{ }{ } 06:11
TimToady that's a q{} variant in my head
though it certainly pays attention to the bracketness on left 06:12
diakopter what a head
TimToady (as does p6, albeit differently)
anyway, unoverloading {} is why we get things like "\x[dead] {$foo}" instead 06:14
so ${foo} is no longer a special case, it's just a block in a funny spot
lue we could always use a special literals bracket if you fear overloading. I'm partial to ⦃1/2⦄ (only because it's somehow my go-to "Unicode brackets"). Not ⟅1/2⟆, that's clearly for Bags :) .
.oO(If they weren't generic brackets, I'd be even more partial to ☃1/2☄)
 06:16
06:16 darutoko joined
TimToady there are lotsa lovely brackets in Unicode, but those seem like overkill for literals, especially when it already falls out of qw// semantics 06:16
well, qw:v or whatever 06:17
lue TimToady: I agree. But if you do someday fear <> overloading... :)
(on the overkill that is)
TimToady well, the main problem with our use of <> is that it requires space in front of binary infix < 06:18
but we're not optimizing for golf anymore, so p6 doesn't mind enforced spacing in certain spots 06:19
lue ‹› would be probably the least-shocking change-to I think. (Though it would interfere with if we want <quote word lists> and other <> bracketing uses to use ‹› as the non-Texas version someday.) 06:20
TimToady well, huffman coding kinda applies, if you consider typing
don't mind Unicode ops for sets, since those are much rarer than lists of numbers 06:21
or literals
using ‹› to indicate something like "always assume numeric types" might be a possibility, but while ‹› are a good choice for something like <>, they are also a bad choice because they're not quite visually distinctive enough 06:23
lue TimToady: Just as a small confirmation, NFG ord() and chr() are supposed to deal in not-exactly-codepoint numbers, right? ord("a􏿽xCC􏿽x88") would be potentially 0x11_0001 in that case, for instance. 06:24
That's why I prefer ‹› to be to <> what «» is to <<>> instead. But it was a thought that came to mind. 06:25
TimToady ä has a precomposed form, so is always NFC
(I think) 06:26
lue Oh? P6 does forced NFC by default?
(forced at time of string creation that is)
06:26 [Sno] joined
diakopter lue: yes 06:26
lue In any case, that was just the fastest multi-codes grapheme I could reach for.
(assume it represents a generic uncomposable grapheme :P) 06:27
TimToady NFG is defined as NFC plus extra canonicalization of things that "ought" to have a precomposed codepoint
lue diakopter: yes to which question? or both?
lue suspected NFG would prefer normal codepoints where possible. 06:28
06:28 SamuraiJack__ joined 06:29 SamuraiJack__ left
lue Now I wonder if there's a NFDG, NFKG, and NFKDG :) . Also if we want to invent other NFs, like NFP (codePoints), NFU (code Units), and NFB or NFO (Bytes or Octets). 06:30
At least I know now why you called in NFG instead of just, say, count-by-graphemes.
moritz \o 06:31
TimToady well, it's also a pun
lue o/ moritz
06:31 SamuraiJack joined
TimToady oh, the made-up graphemes will always be negative, not > 0x110000. 06:32
lue But... but... the ISO pledged to never go beyond 10_FFFF. Or are you suggesting they'll replace UTF-16 with a 16-bit version of UTF-8 someday? :)
TimToady the symmetry around 0 makes it possible to use signed types like int8, int16 for strands of the rope that aren't too high up in U-land 06:33
lue Honestly I just typed 0x11_0001 because I didn't feel like typing out enough bits for a negative number :P 06:34
TimToady -1 is shorter :)
TimToady admits that -1 arguably depends on constant folding :) 06:35
n: say <-1>.WHAT
camelia niecza v24-98-g473bd20: OUTPUT«(Int)␤»
TimToady n: say < -1 >.WHAT
camelia niecza v24-98-g473bd20: OUTPUT«(IntStr)␤»
lue unlike Rats, we don't need angle brackets for that literal. Just use ­APL's HIGH MINUS :)
TimToady that was not just APL; they trying to foist that notation on us when I was in grade school :) 06:36
thankfully it didn't stick 06:37
*tried
ah, the 60s...
jlaire J uses _1, that's like LOW MINUS 06:38
06:39 FROGGS[mobile] joined
TimToady anyway, the intent of NFG is not to try to encode all the codepoints into the integer, but to synthesize a codepoint that can be used to lookup the NFD form in a table somewhere 06:39
lue TimToady: would NFG and "str".decompose ever exist in the same space? I imagine issue would arise with things like ord("ä".decompose) (would ord NFG-ify its copy of the string?) 06:40
TimToady you just get an NFD string out of it, is all
if Str is really NFG, then what we have now is what we might end up calling Uni, where it's just known to be codepoints 06:41
if we know the canonicalization, it might be a more specific type such as NFD or NFC
but we make no guarantees about the operations available on those types, as of now. 06:42
those types are more like utf8 in that respect
lue has been assuming NFG to be a fancy way of saying "count by graphemes" until now, so how NFG fits with the rest of NF* and all that is a bit fuzzy. 06:43
If only there were a synopsis on Unicode that would clarify this... :)
TimToady it really means "store by graphemes", so we avoid any hint of variable-width characters
or surrogate monstrosities 06:44
06:44 mschreck left
TimToady nor does the concept of combining characters exist in NFG 06:44
except insofar as we allow people to do matching on them in regex, but that will probably require a special notation 06:45
lue So beyond storage, does NFG have any other effect on the basis of counting characters? That is, does NFG turn on count-by-graphemes, or is it that the count-by setting happens to correlate with the storage method by default?
06:45 rurban2 left
TimToady there's no "count by" needed when you basically have an array of integers, one per character 06:45
06:46 mschreck joined
TimToady not in the sense of "count by" meaning "cheat on a variable-width encoding" 06:46
for regex we'll need a way of saying "match this base character where it also contains this set of one or more combining chars" 06:47
we haven't specced that yet that I know of
lue So pragmas controlling how to view strings make no sense (e.g. use graphs; or use codes;). It'd be more useful to change the storage of the string's contents to change how you see things. 06:48
(and things like $str.codes would act as if the string stored its contents as codepoints if not already.)
TimToady that's the current way of looking at it; in any case, it's more type-based than having much to do with lexical scope
moritz or to have separate types 06:49
moritz too slow
TimToady that being said, different languages happen in different lexical scopes, and a different language might not support NFG
lue pragmas would at least control the default storage. So things like use NFC; or use NFKD; or use NFG; (and perhaps other invented NFs like use NFP; to store by codepoints) 06:50
TimToady our current Perl 6 language, for instance, is really in a 'use codepoints' kind of mental space
TimToady hopes we can switch over to NFG semantics without too much heartburn
yes, defaults can certainly be the province of pragmas
lue wonders if at somepoint "str".ords is better written as "str".encode :) 06:51
TimToady if you say "use NFD;' then presumably all your literal strings are stored as NFD, for instance
but we need to work out the type system to say much beyond that 06:52
lue I assume I can think of NFG as NFC++ then, right? 06:53
06:53 logie_ joined
TimToady does . always match a codepoint in a Uni string? 06:53
that's how I think of it
moritz thought it always matched a grapheme
TimToady supposed to, in NFG-land
but all is fair if you predeclare 06:54
lue for ^0x11_0000 { say chr($_) ~~ /./ ?? True !! False } should test it :)
06:54 xenoterracide left
lue wouldn't be surprised if at some point Str becomes just a default form of String[things], like Rat is Rat64 is Rational[Int,UInt64] 06:55
TimToady to take an extreme example, if you say 'use v5;' it is no longer the case that . matches a grapheme
that might well happen for other languages, including non-NFG P6 variants 06:56
but we really need to play with NFG first to see how non-NFG will work out
lue Once NFG happens in an implementation, I predict a lot major and unsettling changes will occur to how we do strings.
TimToady this seems likely 06:57
06:57 mschreck left
TimToady for instance, the need to treat individual characters as potentially complex entities 06:58
(when matching)
06:58 xenoterracide joined
TimToady currently we have :m and :samemark, which are crude ways of saying 'ignore the combiners' 07:01
lue Would it be right to think of codepoints and below as low-level enough that an invented NF* doesn't make the most amount of sense? Or have we generalized NF* as ways to store strings (from the Unicode meaning of "(de)composing characters", with potentially some transformations to the exact contents?
TimToady they are basically pragmas writ small
TimToady doesn't quite understand the question 07:03
lue TimToady: another way to put it is: in addition to things like "str".graphs, "str".codes, and so on, would we have "str".composeds and "str".decomposeds(:compatability) to go with NFC and NFKD ? 07:04
TimToady at the Uni level, the meaning of the string is sufficiently characterized by Unicode properties that we could conceivably do some amount of autocoercion
lue "would there be NFP (codePoints) and such in addition to NFG and NFKC?" is yet another form of the question. 07:05
TimToady you can't make up information that isn't there though
I don't understand what your NFP would do that an undifferentiated Uni string wouldn't 07:07
NFC and NFD, as constrained variants of Uni, are measured in .codes 07:09
lue NFP would store the strings much as they are now: by unicode codepoints. I don't know, my mind feels eager to blur the line between the meaning of NF* and what your idea of the base unit of a string some more it seems :) .
TimToady just as utf8 is measured in bytes 07:10
my mind is blurry enough without outside encouragement :)
lue So NF[K]?[C|D] is effectively "codepoints" then. Then encoding forms would be how you count by "code units", and buf8 would be how you count by "bytes" or "octets" :) 07:11
(and NFG of course is "graphemes")
TimToady I wouldn't use "encoding" to talk about NF*
07:12 kaleem joined
TimToady in sense they are, but using that term for them will confuse people 07:12
if I understand the unification you're going for, it's that NFG.codes is the same as NFG.chars 07:13
a "codepoint" in NFG just happens to be grapheme
you can't tell how many real codepoints it would produce if you converted to NFC or NFD, because those are fuzzy in the same way that, er, the number of seconds in a month is fuzzy... 07:14
you must at least say "if it *were* converted to NFD, what would it be?"
07:15 arcterus left
TimToady much as with dates you have to say, "one month later, supposing today is Jan 31" 07:15
though the metaphor is strained
07:15 arcterus joined
lue no, "Then encoding forms would be how you count by "code units"" refers to the UTF- family. A UTF-32 code unit is a 32-bit number, a UTF-16 code unit 16-bit, etc. 07:16
TimToady you can't know the length in .codes without know the canonicalization, just as you can't know the length in .bytes without knowing whether it's going to be UTF-8 or UTF-16 or UTF-32 07:17
lue So "default utf-8".units would be different that "default utf-8".convert-to(utf-32).units, assuming you had multi-unit characters in the UTF-8 string.
bytes is where .encode() comes in :)
TimToady: it's late here, so tomorrow I'll write what I'm thinking in pseudo-provisional-S15 kind of way, so I can explain myself more clearly than one-line-at-a-time IRC :) 07:18
TimToady at this point, specifying either a default encoding or default normalization seems smelly to me
07:18 ingy left
lue But don't we do utf-8 and NFC++ by default? :) 07:18
♘ all o/
TimToady o/ 07:19
07:19 berekuk joined 07:32 darutoko- joined, berekuk left 07:33 darutoko left, xenoterracide left, xenoterracide joined 07:41 FROGGS left 07:47 berekuk joined 07:55 markov joined 08:10 darutoko- left, darutoko joined 08:11 dmol joined 08:15 nnunley left
dalek pan style="color: #395be5">perl6-examples: a01a165 | electricface++ | 99-problems/P10-topo.pl:
Update P10-topo.pl
 08:18
pan style="color: #395be5">perl6-examples: 878d9f1 | moritz++ | 99-problems/P10-topo.pl:
Merge pull request #13 from electricface/patch-1

Update P10-topo.pl
08:35 FROGGS[mobile] left 08:49 arcterus left, Mouq joined 08:50 arcterus joined 08:53 Mouq left 08:56 denis_boyun joined 09:04 denis_boyun left 09:10 [Sno] left 09:12 nnunley joined 09:13 denis_boyun joined 09:17 Alina-malina left 09:18 Alina-malina joined 09:30 denis_boyun left 09:32 denis_boyun joined 09:38 woolfy joined 09:40 berekuk left 09:41 mschreck joined 09:42 denis_boyun left 09:47 hummeleB1 joined 09:51 dakkar joined
lizmat good *, #perl6! 09:52
09:52 denisboyun joined
lizmat gives up on trying to use the hotel wifi to reach apparently blocked sites, such as github, or perl6's advent site, or any other number of sites 09:53
I suspect some kind of IPv6 tunnel failure at either the hotel, or the hotel's provider 09:54
instead, I'll just start our decommute a bit earlier
09:55 fhelmberger joined 09:56 tgt joined
hoelzro lizmat: do you need a proxy? 09:58
I can lend you one
10:00 denisboyun left 10:06 lizmat left 10:07 woolfy left 10:09 denisboyun joined
jnthn morning-ish... o/ 10:11
10:12 FROGGS[mobile] joined 10:15 denisboyun left 10:16 FROGGS[mobile] left 10:17 nnunley left
nwc10 good UGT, jnthn 10:17
10:33 Mouq joined
masak good antenoon, #perl6 10:34
preflex masak: you have 1 new message. '/msg preflex messages' to read it.
masak ooh, a message!
preflex: messages
preflex lue said 11 hours, 54 minutes and 22 seconds ago: gist.github.com/lue/7741574 has the same Temporal stuff, but in a less sensitive location :)
moritz a masak! 10:36
masak I suppose everything went well with the advent post?
haven't looked yet :) 10:37
moritz yes, it was published
10:37 Mouq left
masak \o/ 10:37
masak is teaching
jnthn masak: Hope it's not too testing a class... 10:38
10:43 tipdbmp left
masak jnthn: they are assertive but also responsive :) 10:44
jnthn So long as they ain't mockists, or you'll have to send 'em to AAA for help... :P 10:45
10:47 denisboyun joined
masak none of them have shown any mockist bent yet. but it's early in the course still. 10:50
masak .oO( To Mock a Mockist Bird )
10:51 dmol left 10:53 denisboyun left 10:59 fhelmberger left, fhelmberger_ joined, SamuraiJack_ joined 11:00 SamuraiJack left 11:04 tipdbmp joined 11:08 mschreck left
masak wow, Java is so inept. 11:08
11:08 mschreck joined
jnthn It has class...but not much else. 11:09
masak Q: How do I sort an array of doubles in reverse order? 11:11
A1: With difficulty.
A2: Don't do that.
moritz .sort.reverse ? :-) 11:12
jnthn theArray = theArray.OrderByDescending().ToArray(); // C# answer ;)
masak moritz: hahaha *sob* 11:14
11:15 mschreck left, SevenWolf left 11:26 denis_boyun joined
colomon Hmmm, DateTime::TimeZone and IO::Prompter both working with latest smoke test, but now Time::Duration has started failing. 11:44
11:44 dmol joined 11:45 dmol left 11:46 dmol joined 11:53 xinming left 11:56 xinming joined 11:57 dakkar_ joined, silug_ joined, atta_ joined, dakkar left, dylanwh left
hoelzro I take it that command line switches are all loaded into %*COMPILING<%?OPTIONS>? 11:57
automatically?
11:57 dylanwh joined, silug left
hoelzro I have a Perl6::Grammar object in NQP, and I'd like to tell it to use a NULL setting 11:57
moritz hoelzro: just look at how rakudo does it 11:59
hoelzro that's the thing; I'm trying
moritz if %*COMPILING<%?OPTIONS><setting> eq 'NULL' { 12:00
is what Perl6::Grammar and Perl6::Actions use 12:01
which a git grep NULL found pretty much instantly in rakudo 12:02
jnthn hoelzro: Gotta head out right now, but I'm curious what your "null setting" use case is...given that you in that case probably are compiling a setting. 12:03
12:04 mschreck joined
jnthn bbs & 12:05
12:13 fridim__ joined 12:16 dwoldrich left 12:19 dmol left 12:21 dmol joined, Mouq joined 12:23 dmol left 12:25 Mouq left 12:26 dmol joined
masak lue: well, the problem with "humanistic counterpart" is this: Instant is well-defined, DateTime is well-defined (if you squint), Duration is well-defined (since Instant is), but Interval is not well-defined at all. 12:27
lue: and just saying that we'll have such a type doesn't fix the problems of non-well-definedness. it pulls them up to the surface as puts them in the spec. 12:28
lue: I suggest you start from three (or so) use cases, and we can discuss those, and how they are potentially problematic.
lue: use cases, as in follow an Interval from beginning to end. 12:29
lue: naming-wise, it bothers me that the pairing is Instant/Duration and DateTime/Interval. I/D and D/I makes less mnemonic sense than I/I and D/D. but that is a smaller point compared to the conceptual difficulties.
12:36 FROGGS joined 12:37 xenoterracide left 12:39 araujo left
masak lue: as to the week syntax in constructors, I simply disagree that we need it, just because ISO 8601 has it. 12:42
lue: I invoke Postel's Law. recognizing many different input formats makes the class weaker, not stronger.
12:43 dmol left
FROGGS a clear design is worth much, yes 12:43
it will grow weak over time anyway 12:44
12:44 dmol joined 12:47 mberends joined 12:54 araujo joined, dmol1 joined, dmol left 12:56 rurban1 joined 13:03 ggoebel110 joined 13:05 SamuraiJack_ left, SamuraiJack_ joined 13:07 PZt left
hoelzro moritz: right, but I want to set setting to 'NULL' 13:09
not ask if it is
jnthn: I doubt I'll be sticking with a NULL setting; I'm just playing around
moritz hoelzro: my %*COMPILING; %*COMPILING<%?OPTIONS><setting> := 'NULL'; 13:11
13:11 SamuraiJack_ left
hoelzro hmm 13:11
I didn't think it'd be that simple =/
moritz well, that's the first thing I'd try
hoelzro I was also curious to see if it was some magic populating %*COMPILING<%?OPTIONS>
or if something crawls over ARGV and populates it that I'm not seeing 13:12
moritz yes, that happens
in HLL::Compiler.compile 13:13
hoelzro ah ha
that's why I wasn't finding it
moritz and method command_line calls process_args which in turn delegates most of the wrok to HLL::CommandLine::Parser 13:14
13:23 mschreck left 13:24 denis_boyun left
hoelzro hmm 13:28
my VM *really* doesn't want to compile CORE.setting
how much memory does it take to compile CORE.setting with nqp-p these days, anyway? 13:30
I have 2GB of RAM on this VM 13:31
but top reports that ./perl6-p is only taking 25% of that
of course I don't have the swap column toggled on =/
13:33 denis_boyun joined
masak FROGGS: well, especially in this case (with Interval) where there are *known* pitfalls that haven't been accounted for. 13:35
moritz hoelzro: around 1.5GB, give or take
hoelzro hmm 13:36
then I don't see why my VM is complaining
FROGGS masak: yes, that I why I'd talk to drolsky first, and would ask him nicely for some kind of spec/design
masak FROGGS: he has weighed in on exactly these issues on p6l at least once. 13:37
FROGGS: his comments (and moritz++' work) were what led to a separate Date class.
FROGGS: and it also made me aware of the subtleties involved in having a type representing a (human-datetime) interval. 13:38
FROGGS well, we could have both, durations like humans think about it and durations like machines tend to think about it 13:39
but it is way more than that
don't know how his tuit supply is though, and if he is willing to invest more time 13:40
example: I use P5's spans and spansets a lot at work to calculate fees, so I know what I'd need for rewriting that in P6, but I guess I have not covered all absurdities that happen when it comes to times 13:42
masak FROGGS: one well-known source of confusion in CPAN's DateTime::Duration is that it adds the units (IIRC) smallest-first. 13:44
FROGGS that might be the reason why I only supply a single value 13:45
masak so Feb 28 + (1 month, 1 day) is not Mar 29, but Apr 1.
FROGGS: right, and "only supply a single value" is what .delta does!
problem solved.
FROGGS :o)
13:45 kivutar left 13:47 dmol1 left 13:48 kaleem left
FROGGS can we do spanset math already? I guess not 13:49
because that would be so much nicer as my well written P5 code 13:50
jnthn is back
FROGGS welcome back, jnthn! \o/ 13:51
jnthn masak: Wouldn't Postel's Law mean being liberal in what you can pass in? ;)
13:55 atroxaper joined, rurban1 left
hoelzro shrugs 13:58
compiles fine when I turn off swap.
(don't try this at home, kids!)
14:01 dakkar_ is now known as dakkar 14:05 davecv joined, dave1c joined, nebuchadnezzar joined
Woodi hallo :) 14:07
14:07 kaare_ joined
Woodi I think Duration and Interval are the same thing... Duration is just term more in Time department... 14:08
14:09 Mouq joined
Woodi some others here (clicking phase is nice): translate.google.pl/?hl=pl&tab...l/interval 14:09
but if we talk about human representations then chapter Builder from GangOfFour applies... 14:10
(I openned it randomly yesterday @ caffeteria :) 14:11
14:14 Mouq left
Woodi btw. do Perl6 clousures are allowed to close any varible they can access during clousure creation or just values from function creating clousure ? 14:15
timotimo i expect they close over everything they can access 14:16
jnthn Anything visible
14:16 pdurbin left
timotimo right, because you can theoretically access more than is visible :P 14:17
Woodi en.wikipedia.org/wiki/Closure_%28c...science%29 shows just newly created me think... so we have better clousures ? :)
14:18 rurban1 joined 14:20 PacoAir joined 14:29 kaleem joined 14:31 lowpro30 joined 14:32 PerlJam joined 14:36 jnap joined 14:46 berekuk joined 14:48 lowpro30_ joined, mtk left 14:51 lowpro30__ joined, lowpro30 left 14:52 mtk joined 14:54 lowpro30_ left
masak jnthn: dangit! somehow I always get Postel's Law turned around! :( 14:59
jnthn masak: Well, that's easily fixed. Just be more liberal about finding time to read about it, and conservative in how often you talk about it. :D 15:00
The thing that always bothers me a tiny bit about Postel's Law is that it's kinda the opposite of what we learn from a security perspective: be picky about what you accept. :) 15:01
masak yeah, I guess the assumption is something like "if the user is known-friendly"
my current view of it (which may be what is confusing me) is "be picky about what you expect, but be hesitant in flagging user input as wrong" 15:02
15:03 kivutar joined
jnthn dalek: you working? :) 15:04
colomon I believe in it very strongly for my $work. Read any old crappy files written out by other CAD systems; always try to write out 100% correct files from ours. 15:05
15:05 kaleem left 15:06 kaare_ left
masak it strikes me that the situation is very different depending on whether it's the one-and-only interaction between client and server, or whether we know that we are in an interactive, iterative environment like an event loop. 15:08
I think Postel's Law applies best to the former. 15:09
15:10 lowpro30__ left 15:11 kaare_ joined
colomon Which is my situation, I guess. It's not like people trying to read a CAD file can go to the people who wrote the program that wrote the file and tell them to fix their bugs in a timely fashion. 15:11
"Dear AutoCAD. You are incorrectly specifying the angles of cones in your STEP files. Please get a fix by this afternoon to my friends that wrote this file for me." 15:12
15:13 cognominal left 15:14 cognominal joined
Woodi tell that to photo camera manufacturers - everyone have it's own, secret RAW light catching format :) 15:16
colomon (Note: I don't think that was the actual AutoCAD bug I ran into a couple of years back. And luckily I do have contacts there that I could report the bug to.) 15:17
masak colomon: interestingly, it also shows why XML's über-strictness in browsers is the wrong behavior.
colomon: (because it is not the page visitor's fault that the page is faulty)
colomon right
15:21 arcterus left 15:30 btyler joined
moritz founds strict XML parsing very helpful for the IRC logs, made my fix any possible cross-site scripting holes 15:32
masak moritz: yes, but that's you as the author, not you as the visitor... 15:33
colomon sure, and having a STEP file reader which complains at the slightest glitch is really handy for tuning my STEP file exporter to properly follow the (insanely complicated) spec. 15:34
moritz masak: well, in some cases my visitors alerted my "page $url doesn't display here", and I fixed that too :-)
masak moritz: I see. you have very nice visitors, then. :) 15:37
hoelzro hmm 15:39
apparently you can't subclass Perl6::Grammar from within Perl 6 =/
I'm guessing that's because it's from NQP land
masak realizes he would like a super-easy way to mark a (user) class as being a value type in Perl 6 15:41
moritz hoelzro: yes, jaffa4 recently had the same problem 15:43
masak: +1
masak mainly to make them easily comparable with, hm, eqv, I guess.
hoelzro moritz: do you know if they solved it? 15:44
moritz with === I guess
hoelzro: no
hoelzro hmm
masak moritz: yeah, === is better.
moritz and eqv already looks into the attributes by default, I believe 15:45
masak r: class A {}; #`{magic invocation here}; say A.new === A.new # should print "True"
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«False␤»
moritz r: class A { has $.x }; given A.new(x => 42) { say $_ eqv A.new(x => 42); say $_ eqv A.new(x => 5) }
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«False␤False␤»
moritz hm, doesn't 15:46
masak no, only array elements and hash entries, IIRC.
jnthn As a thing to beat, .Net makes you inherit from ValueType... 15:49
hoelzro hmm 15:50
well, my code *almost* works in NQP
but I get a null PMC access =/
and I've determined that it's trying to call unshift on @END_PHASERS
I take it I have to somehow set that up before calling Perl6::Grammar.parse? 15:51
masak jnthn: that feels like a simple solution. 15:52
hoelzro ah, I see it in main.nqp 15:53
15:53 rurban1 left 15:54 rurban1 joined 15:57 Mouq joined 15:58 denis_boyun left, Psyche^ joined
masak decommutes 16:01
16:01 Mouq left 16:02 Psyche^_ left 16:07 berekuk left 16:09 berekuk joined
hoelzro hmm 16:24
where does hll-config in main.nqp come from?
jnthn Think it's generated 16:25
hoelzro grrr
jnthn But I think you might get away with an empty one too... 16:26
hoelzro I'm just trying to run main.nqp via nqp-p at the moment
(as a way of sanity checking things)
16:26 PZt joined
hoelzro it's proving...difficult. 16:26
hmm 16:27
my lack of --vmlibs is probably not helping =)
nope, that doesn't help either. 16:28
FROGGS hoelzro: what do you wanna sanity check? 16:35
hoelzro well, I'm trying to extend the Perl6::Grammar from an NQP program for an experiment I'm trying
FROGGS r: say nqp::gethllsym('perl6', '$COMPILER_CONFIG')
camelia rakudo-jvm 874e35: OUTPUT«("codename" => "", "name" => "rakudo", "release-number" => "", "build-date" => "2013-11-30T23:10:59Z", "version" => "2013.11-20-g874e358").hash␤»
..rakudo-parrot 874e35: OUTPUT«("name" => "rakudo", "version" => "2013.11-20-g874e358", "release-number" => "", "codename" => "", "build-date" => "2013-12-01T00:20:06Z").hash␤»
hoelzro I can't convince NQP to do what I want
FROGGS do you want to access that?
hoelzro so I'm just trying to get main.nqp to run atm
FROGGS ahh, hmm
hoelzro which it stubbornly refuses to do. 16:36
hmm 16:37
I may have figured it out
hmm
16:37 fhelmberger joined
hoelzro is there an nqp op for asking about an object's type? 16:37
like nqp::type(@args)?
moritz I'd try nqp::what
nqp: my $x := nqp::what([]); say($x.HOW.name($x)) 16:38
camelia nqp-parrot: OUTPUT«Can only use get_what on a SixModelObject␤current instr.: '' pc 50 ((file unknown):43) (/tmp/eTdISkhp_s:1)␤»
..nqp-moarvm, nqp-jvm: OUTPUT«NQPArray␤»
hoelzro that angered NQP
moritz jvm++
moar++
hoelzro cries
I can't wait for the Moar port to be done. 16:39
moritz then don't wait; help!
16:39 SamuraiJack_ joined
hoelzro heh 16:41
jnthn nqp-j: sub foo() { }; foo(1)
hoelzro I'd like to
16:41 fhelmberger_ left
camelia nqp-jvm: OUTPUT«Wrong number of arguments passed; expected 0..0, but got 1␤ in foo (/tmp/CPPvYxcvRD)␤ in (/tmp/CPPvYxcvRD:1)␤ in (gen/jvm/stage2/NQPHLL.nqp:1100)␤ in eval (gen/jvm/stage2/NQPHLL.nqp:1086)␤ in evalfiles (gen/jvm/stage2/NQPHLL.nqp:1292)␤ in command_…» 16:41
hoelzro learning more about moar is on my TODO list
16:41 ajr joined
hoelzro decommute & 16:41
16:41 ajr is now known as Guest68285 16:42 Guest68285 is now known as ajr_, fhelmberger left 16:44 thou joined 16:45 atroxaper left, kaleem joined 16:50 berekuk left 17:01 berekuk joined 17:09 salv0 left 17:10 salv0 joined
hoelzro fg 17:11
perigrin hoelzro: no.
I refuse to be foregrounded by you.
:)
hoelzro =)
17:12 dmol joined
timotimo ohai 17:12
what do i do today? 17:13
FROGGS fix pow_I ?
timotimo i feel like more sleep ...
FROGGS :o)
timotimo who, me?
FROGGS ohh, well, that would work too
if you want to... should not take too long
timotimo is that a moar problem? 17:14
FROGGS yes
timotimo i could certainly have a quick look
FROGGS see the commit msg: github.com/MoarVM/MoarVM/commit/091a2369e
timotimo but my head is kind of achy
FROGGS it should return a num if it returns NaN or Inf, and int in all other cases I think
no problem, I can do that too later 17:15
hoelzro so I'm looking at this conditional declaration of MAIN in main.nqp
sub MAIN(@args) for Parrot
sub MAIN(*@args) for !Parrot
what's the deal with that? 17:16
because I actually can't get it to work on Parrot unless I change it to slurpy args
timotimo need to get home first
colomon hoelzro: might indicate we need another test….
timotimo what are you working on? workshop exercises? 17:18
jnthn FROGGS, timotimo: There's a ticket also
(about the pow_I issue)
hoelzro timotimo: me?
jnthn hoelzro: Some annoying oddity about pre-compiled vs run from command line also, iirc... 17:19
timotimo I'd rather fix something that brings us closer to hello world :P
hoelzro ah, that would do it
timotimo yes you
hoelzro I'm working on an idea I had over the weekend
I was trying to do it in Perl6 land, but I can't subclass Perl6::Grammar from there =/
so it's NQP land for me 17:20
17:20 Mouq joined
timotimo that is something i hope we can get to work, though 17:20
either that or making rakudo self host on p6 :P 17:21
hoelzro =)
17:22 arcterus joined
arnsholt hoelzro: If it's any consolation, eliminating that problem with arrays not being sixmodelobjects on Parrot is something we want to kill (there's even a bug for it), but it's a bit of a drudge to actually complete 17:24
17:30 kaleem left 17:32 Rotwang joined
hoelzro I see 17:33
hmm 17:36
masak interesting comment to my post: perl6advent.wordpress.com/2013/12/0...mment-3319
(I don't agree, by the way)
hoelzro now I'm getting a null PMC access in get_pmc_keyed
I discovered that it's trying to access %*COMPILING
but I already have a my %*COMPILING in my MAIN
(on a separate note, %*COMPILING just makes me think of "winning!") 17:37
17:37 SamuraiJack_ left
hoelzro masak: you could just use a pragma-like module to add that behavior if you wanted it, though 17:37
right? 17:38
(by you I mean the commenter)
oh, I see now 17:39
%*COMPILING<%?OPTIONS> must also be a hahs
seems reasonable.
jnthn I think S12 or S06 already spec a pragma for that. 17:40
masak hoelzro: yes, I guess so. 17:43
hoelzro hmm 17:45
ok
so my basic NQP interpreter thingy works.
now I'm trying to extend Perl6::Grammar
and NQP yells at me
"P6opaque only supports type changes where the MRO of the original type is a suffix of the MRO of the new type"
any tips?
17:45 zakharyas joined
hoelzro I don't even know what types I'm changing 17:46
jnthn hoelzro: The debugger manages to subclass Perl6::Grammar. I remember having that error somewhere along the way. 17:47
hoelzro ok, I shall look
arnsholt masak: Interesting comment, but I agree with you
Any method that requires a defined self will fail whenever it tries to do whatever it is that requires it to be defined. Perhaps a bit LTA, but one less thing to remember if you want a method callable on the type object 17:48
TimToady the way to make a value type is to define .WHICH such that it returns the same identity for the same value 17:49
jnthn Hm, that means if you want a generic solution, just write a module and then "does ValueTypeStuff" :) 17:50
Where the role provides a WHICH
TimToady sure, that's one way to do it :) 17:51
17:51 arcterus left
jnthn dalek: why u no report my commits? :P 17:51
dalek: restart
TimToady otoh, you also ought to worry about immutability if you have a value, so you can't just treat value-ness as a special sauce
jnthn ?
Hmm 17:52
dalek: help
Ah well...
TimToady unless, of course, we have a generic way to disable all publicly-visible mutations (caches being a notable exception) 17:53
hoelzro ah, got it
you have to override comp_unit
jnthn Ah, seems nothing is responding on feather3.perl6.nl:5000/
hoelzro and populate %*LANG
jnthn hoelzro: Ah, yeah, 'cus that's the braid source
hoelzro this is proving to be quite a useful experiment =) 17:54
17:55 dakkar left
masak moritz: did you see the good hpmor news today? :) 17:57
17:57 lumimies joined
TimToady sometimes wonders whether %*LANG is a hack to avoid doing the work of mixing in all your sublanguage switchers 17:57
hoelzro mwahahaha
it works! 17:58
TimToady or, to look at it another way, will %*LANG eventually prevent some optimizations that would be possible if you looked up sublanguages by method rather than by hash? 18:00
or is the hash preventing a combinatorial explosion of lexers? dunno offhand... 18:01
well, not inclined to change anything until we run into a real problem, but just had to mention that I do wonder about that from time to time 18:03
18:03 PerlJam left
hoelzro how do you iterate over an NQP hash's keys, again? 18:05
masak TimToady: flood filling demands we try both! :P
TimToady that only works with a flood of volunteers :) 18:06
doubtless someday there will be a huge east/west schism over this issue 18:07
18:07 PerlJam joined
TimToady since it's one of those "who's to master" issues :) 18:08
hoelzro hmm 18:12
perl6.org is quite tempermental today.
[Coke] I see dalek, but no commit messages? 18:16
jnthn [Coke]: The thing that github should call in feather is down.. 18:17
[Coke]: Maybe not started again after a feather restart?
[Coke] ah, probably didn't survive the reboot. danke. 18:18
:)
TimToady actually, not being able to derive the Perl 6 grammar in P6-Land is a huge blocker to language mutability and sane macros; it negates our answer to Lisp's homoiconicity until we can actually parse Perl 6 in real Perl 6 18:19
[Coke] regarding the java vs. parrot status in the advent article - java's been in the lead for weeks.
github.com/coke/perl6-roast-data/b..._rates.csv
jnthn [Coke]: Thanks, I'll need to work on that article soon, I guess. :) 18:20
[Coke] jnthn: someone quoted the stats in the first article, I think. 18:21
I can't be here AND check that at the same time, stupid network.
TimToady yes, I wondered about that 99.9% bit 18:22
[Coke] if I want to play with moar on rakudo, it's the moar-support branch in rakudo? (should perl Configure.pl --gen-nqp --backends=moar work?) 18:23
FROGGS hoelzro: grep for nqp::iter* in nqp 18:26
hoelzro ah, thank you FROGGS
jnthn [Coke]: It doesn't work yet so there's not much to play with... :) Builds the setting, mind...
shop; bbiab 18:27
[Coke] jnthn: I can't even get it to build moar. 18:28
(or run Configure.pl without warnings). I'll see about adding some cleanups for that.
... after $dayjob. 18:29
hoelzro: a doc patch adding iter to the docs/ops* would be most welcome. 18:30
hoelzro I think that can be arranged =) 18:31
18:33 sizz joined, ashleydev joined 18:34 sizz_ left
FROGGS hoelzro++ 18:35
hoelzro hmm
I wonder if we should submit Perl 6 repos for the 24 PRs thing?
18:36 colomon left
FROGGS 24 PRs thing? 18:36
18:37 pecastro left, ajr_ left 18:38 ajr joined, colomon joined, ajr is now known as Guest67784, darutoko left 18:39 Guest67784 is now known as ajr_, denis_boyun joined 18:40 pecastro joined 18:48 Duchess joined 18:49 ajr_ left 18:51 Esylt left
hoelzro 24pullrequests.com 18:56
18:59 berekuk left, colomon left
FROGGS that would be awesome! 19:01
19:01 colomon joined, berekuk joined
hoelzro =( 19:07
FROGGS ?
hoelzro I said "Perl 6" as the programming language
invalid entry: "must be a programming language"
fools!
FROGGS :(
hoelzro I'll just say Perl
FROGGS wait, maybe we can patch it :o) 19:08
hoelzro hehe 19:09
good idea!
19:11 ajr_ joined 19:14 berekuk left
FROGGS PR sent 19:16
Rotwang is NativeCall broken in 2013 11 or is it just me?
FROGGS what is wrong?
its tests pass on my box, and Inline::C which uses it works too
hoelzro FROGGS: wow, nice! 19:18
PR sent for nqp =) 19:20
Rotwang FROGGS: I'll try to make some minimal example 19:21
hoelzro EXPR/OPER are special rules, right?
FROGGS Rotwang: that would help 19:22
hoelzro they're provided by some convenience routines or something?
FROGGS hoelzro: they live in HLL::Compiler or so
hoelzro ah ha
FROGGS they are methods
19:22 sqirrel joined
hoelzro hmm 19:23
it looks like if I define OPER in my actions class, it doesn't get called =(
FROGGS maybe it must go into your grammar? 19:24
that is what I would expect
jnthn OPER?
FROGGS because action methods are called when their grammar token/rulen has matched 19:25
hoelzro OPER or EXPR
I only care about the former atm
but I could see myself caring about the latter
hmmm
since this is just a hack
I could override OPER in the grammar
good thinking FROGGS =)
FROGGS hoelzro: PR merged! 19:26
hoelzro: both, the 24prs and the nqp one... 19:27
what happened to dalek?
19:29 nnunley joined
Rotwang FROGGS: > hai.c && gcc -shared -fPIC -o libhai.so hai.c; perl6 -e 'use NativeCall;sub hai() is native("./libhai.so") { * };hai' 19:30
this is what I get: paste.lisp.org/display/140342 19:31
19:31 nnunley left
FROGGS hmmm 19:33
Rotwang: can you paste hai.c too? 19:34
Rotwang FROGGS: > hai.c
it is empty
FROGGS > hai.c && gcc -shared -fPIC -o libhai.so hai.c; perl6 -e 'use NativeCall;sub hai() is native("./libhai.so") { * };hai'
Cannot locate symbol 'hai' in native library './libhai.so'
that is what I get
Rotwang weir
d
FROGGS yea
h
hoelzro FROGGS: thanks! 19:35
FROGGS hoelzro: pleasure!
Rotwang: and mine fails in line 124, while your fails in 120 19:36
rjbs has just parted company with diakopter, who left saying, "I should really move to Philadelphia."
19:36 berekuk joined
Rotwang FROGGS: yeah in your case it fails as it should it seems 19:36
FROGGS Rotwang: github.com/jnthn/zavolaj/pull/30 19:37
Rotwang: do a panda update, and panda install NativeCall
Rotwang FROGGS: ok good to know [;, thanks 19:39
FROGGS I hope it helps
Rotwang I should probably use it without extension anyway 19:41
FROGGS yes, if you want it to be portable for example 19:42
TimToady once again in the situation of having to blow away install and nqp and rebootstrap all nqp compilers, sigh... 19:51
hoelzro gruuu
NQP doesn't support captures, does it? 19:52
ex. method EXPR(|args)
moritz hoelzro: no, you have to use EXPR(|@pos, |%named) 19:55
hoelzro ah ha
19:58 dalek left, dalek joined, ChanServ sets mode: +v dalek
TimToady grumbles about not-quiteness 19:59
FROGGS TimToady: what happened?
TimToady same hangup trying to compile m-BOOTSTRAP, which no amount of submodule sync or remaking or git pulling would fix 20:01
eventually I was in a state of having a gen/moar/.gitignore that something installed as a managed file, and git was very unhappy with me, so I gave up 20:02
lue hello world o/
TimToady but it's aggravating to have to regen parrot and jvm stuff in order to regen the moar stuff
and from prior experience, just regenning moar means it thinks you want to throw away the parrot and jvm stuff, and it gets very confused 20:03
FROGGS yeah 20:04
TimToady this gen stuff is very all or nothing
FROGGS :o(
TimToady and very much against the spirit of 'make'
moritz how about just running 'make m-clean' in nqp?
TimToady didn't help 20:05
FROGGS rp: say <.. .>.first( *.IO.w ) # <--- I love that btw
camelia rakudo-parrot 874e35: OUTPUT«.␤»
FROGGS I hope it gets better when we merge moar-support into nom
TimToady thing is, a make m-clean can't fix a mis-gen
20:06 zakharyas left
FROGGS btw, you can be lucky that you don't add ops to moarvm... because there is a perl6 script that updates the op lists 20:07
TimToady or two gens that get out of sync via git
20:07 eiro left
FROGGS which can be funny when you have no perl6 binary because you tried to build perl6-m 20:07
hoelzro hmm 20:10
HLL::Actions and Perl6::Actions both hook into EXPR
but my EXPR method isn't getting called
FROGGS and NQP::Actions to mention a third :o)
20:10 dmol left
hoelzro noted =) 20:11
hmm 20:12
FROGGS hoelzro: do you have a EXPR token in your grammar=
?
hoelzro it's just inheriting from Perl6::Grammar
my EXPR method seems to be called, because adding make 5; to my EXPR method blows everything up. 20:13
20:13 dmol joined
FROGGS hmmm 20:13
20:13 rurban2 joined
hoelzro hmm 20:13
or not
20:13 rindolf joined
hoelzro (throwing an exception didn't work) 20:13
FROGGS but your EXPR method is in your grammar now, right? like the one in Perl6::Grammar 20:14
hoelzro oh, no 20:15
I removed it, considering that the other actions methods have EXPR methods
20:16 rurban1 left
hoelzro hmm 20:17
if I add EXPR to my grammar
I'm not sure how to populate $/
FROGGS self.orig should do 20:18
err, no
I think you don't have that at that point
hoelzro crap.
moritz tries to restart all of dalek
20:19 dalek left
hoelzro I think I'll take a break on this. 20:19
20:19 dalek joined, ChanServ sets mode: +v dalek 20:21 eiro joined
moritz wow, more than 800 page views on both days of the advent calendar 20:22
FROGGS wow 20:23
dalek p: 4128a67 | moritz++ | .gitignore:
.gitignore MANIFEST, which is ignored by the release process
 20:25
moritz dalek works again \o/ 20:26
FROGGS: two (iirc) years ago we were slashdotted and tweeted by tim o'reilly, and had 11k views on one day :-) 20:28
FROGGS O.o
that is nice :o) 20:29
moritz ah, it was 2010 20:34
20:38 kst` is now known as kst
Rotwang I've noticed that: say "foo" »~» "bar"».grep: { False }; 20:42
runs with no end on my machine
jnthn p: say "foo" »~» "bar"».grep: { False }; 20:43
camelia rakudo-parrot 874e35: OUTPUT«(timeout)»
jnthn p: say "foo" »~» ();
camelia rakudo-parrot 874e35: OUTPUT«(timeout)» 20:44
jnthn Well, that golfed quickly...
Rotwang so I guess it shouldn't happen? 20:45
lue ooc, is "str".encode("UTF-32") supposed to return a utf32 object, or would I need to do type conversion, e.g. "str".encode("UTF-32").utf32 or "str".utf32 ?
20:46 awwaiid joined
masak we still have a big huge gap in the advent calendar on the 4th, 5th, 6th, 7th, and 8th. 20:46
I refuse to believe that Perl 6 is in a less active state than last year. that doesn't accord with anecdotal evidence. 20:47
I think there are volunteers out there that haven't stepped forward yet. :)
jnthn: you're up for tomorrow. are you writing a post?
Rotwang jnthn: should I make a bug report or something? 20:48
raydiak perhaps perl 6 has become so useful that people are distracted by using it more than last year? :)
masak raydiak: nice try :P 20:50
moritz masak: for me it's mostly depletion of good ideas 20:51
many simple-to-explain ideas have already been covered on the advent 20:52
dalek : b170e30 | lue++ | misc/perl6advent-2013/schedule:
[advent-2013] Signup for Day 4, heredocs.
moritz ++lue
masak lue++ 20:54
lue I should probably remove the heredocs idea from the brainstorm list, right?
masak yeah.
moar volunteers!
moritz: I started skimming the synopses in order, looking for ideas for my posts. when I was done, I had gotten to mid-S03. ;) 20:55
dalek : 0b818bb | lue++ | misc/perl6advent-2013/topic-brainstorming:
[advent-2013] Remove heredocs from brainstorm list.

No sense making someone believe some topic's not be covered! :)
20:56 zakharyas joined
jnthn Rotwang: Yes, I'd file a bug report 20:57
20:57 kaare_ left
Rotwang ok 20:57
jnthn masak: Yes, I'm working on my most rajt nau
uh, post 20:58
.oO( a bridge to fail... )
masak how most appropriate. :)
moritz
.oO( bridge over troubled bytecode )
dalek : c9e5e17 | moritz++ | misc/perl6advent-2013/ (2 files):
[advent] claim day 7, with Date and maybe DateTime
 21:00
lue Oh yeah! masak: does at least Instant.new($seconds) and Duration.new($seconds) sit right with you? :) [maybe some epoch issues, but aside from that...] 21:03
21:04 kivutar left
moritz 13 unclaimed slots 21:09
timotimo oh lue, heredocs are cool :) 21:11
masak lue: not sure there'll even be epoch issues -- need to think about that. but the immediate question becomes: why do you need *two* separate classes for that?
lue: in fact, that's a good question anyway. 21:12
21:12 stevan_ left, ggoebel111 joined
lue masak: yeah, I was thinking of "why isn't Duration just Instant.delta" as a counterargument to the "delta is good enough" for Interval the other day :) 21:12
21:13 _thou joined, ajr joined, ajr is now known as Guest76769
nwc10 jnthn: is origin/indy-args supposed to work with current rakudo HEAD? 21:14
jnthn nwc10: I've no idea how things work out with it and Rakudo. 21:15
nwc10 ah OK
jnthn nwc10: It's in a branch for a reason ;)
It's a fair upheavel.
nwc10 Stage start : 0.000 21:16
java.lang.NoSuchMethodError: 1B74F280B2FD5C2A2EDF6F91C8F3FCF2B7BE1643.qb_2(Lorg/
perl6/nqp/runtime/CompilationUnit;Lorg/perl6/nqp/runtime/ThreadContext;Lorg/perl
6/nqp/runtime/CodeRef;Lorg/perl6/nqp/runtime/CallSiteDescriptor;Lorg/perl6/nqp/r
untime/ResumeStatus$Frame;)V in trait_mod:<is> (gen/jvm/CORE.setting)
[etc]
gah, my terminal. Line breaks when I don't want them
and not when they should be there.
jnthn yeah, that
sucks
nwc10 anyway, goes boom like that currently.
21:16 lowpro30 joined
jnthn Not quite sure where that error comes from, but yes, it's the kind of error I can quite easily relate to what I'm doing. :) 21:17
Which is making it so some common invocation patterns don't have to go playing with an Object[] arg array.
21:18 dmol left 21:19 dmol joined, xenoterracide joined
FROGGS r: my @a = regex { \w }, regex { \d }; say @a.first( 1 ~~ * ) # isn't that supposed to work? 21:19
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«␤»
21:20 ajr_ left, thou left, ggoebel110 left 21:21 berekuk left
FROGGS r: my @a = regex { \w }; say so @a.first( "a" ~~ * ) 21:21
camelia rakudo-jvm 874e35: OUTPUT«Cannot call 'match'; none of these signatures match:␤:(Cool:D : $target, *%adverbs)␤ in any at gen/jvm/BOOTSTRAP.nqp:1201␤ in method Bool at gen/jvm/CORE.setting:12368␤ in block at /tmp/s8bdpuZlUc:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:…» 21:22
..rakudo-parrot 874e35: OUTPUT«Cannot call 'match'; none of these signatures match:␤:(Cool:D : $target, *%adverbs)␤ in any at gen/parrot/BOOTSTRAP.nqp:1218␤ in any at gen/parrot/BOOTSTRAP.nqp:1209␤ in method match at gen/parrot/CORE.setting:3435␤ in method Bool at gen/…»
Mouq r: (1 ~~ *).say
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«True␤»
Mouq It doesn't doesn't code-ify
21:23 lowpro30_ joined
lue r: (1 ~~ *).WHAT.say 21:23
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«(Bool)␤»
lue r: {1 ~~ *}.WHAT.say
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«(Block)␤»
FROGGS r: my @a = regex { \w }; say so @a.first( *.ACCEPTS("a") )
Mouq I think that's appropriate, that way you can say given $something { when * { ... }}
camelia rakudo-jvm 874e35: OUTPUT«Cannot call 'match'; none of these signatures match:␤:(Cool:D : $target, *%adverbs)␤ in any at gen/jvm/BOOTSTRAP.nqp:1201␤ in method Bool at gen/jvm/CORE.setting:12368␤ in block at /tmp/AwuHnS454E:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:…»
..rakudo-parrot 874e35: OUTPUT«Cannot call 'match'; none of these signatures match:␤:(Cool:D : $target, *%adverbs)␤ in any at gen/parrot/BOOTSTRAP.nqp:1218␤ in any at gen/parrot/BOOTSTRAP.nqp:1209␤ in method match at gen/parrot/CORE.setting:3435␤ in method Bool at gen/…»
jnthn I think it's intentional that ~~ doesn't auto-curry 21:24
Or auto-closure, or whatever we call it... :)
FROGGS what about my last example?
r: my @a = regex { \w }; say [||] @a>>.ACCEPTS("a") 21:25
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«｢a｣␤␤»
FROGGS I thought it would do that
jnthn FROGGS: Yes, was something odd about that .first one...not sure what's up there 21:26
21:26 lowpro30 left
timotimo FROGGS: my headache decided not to subside, so I haven't actually looked at pow_I. did you get a chance? 21:30
FROGGS timotimo: no, still working on module installation... 21:31
timotimo: it is not pressing, so we don't have to do it today
21:32 SevenWolf joined
timotimo all right 21:32
21:33 stevan_ joined
raydiak is there a clean way to make a coercive type declaration like (MyClass(Hash) $obj) work, so you can optionally pass a hash instead of having to create an instance of MyClass manually? or would I be better off writing it as a multi which instantiates MyClass and then passes it to the other multi? 21:37
jnthn We don't have coercion types implemented yet anyway, I'm afraid... 21:38
The multi approach sounds good.
raydiak ah, thanks! you may have just saved me a lot of frustration :)
jnthn Note that a multi foo(%bar) { } will accept anything Hash-y rather than just a Hash (to be precise, anything Associative) 21:39
21:40 Rix left
raydiak if MyHash inherits from Hash, then will multi foo(MyHash %bar) be called in preference to multi foo(Hash %bar) if I pass a MyHash? I'm still a bit fuzzy on exactly how it chooses which multi to call 21:42
timotimo it should
but isn't MyHash %foo something different from what you want?
i remember there being some gotcha for @ and % when giving it a type annotation 21:43
jnthn MyHash %foo means "%foo of MyHash" which is usually not what you mean.
raydiak oh right...%foo is MyHash?
jnthn Um...dunno. I'd just write a straight MyHash $bar in that case... 21:44
raydiak on one side, it means the container...on the other, it means the contents...but I forget precisely how it goes
jnthn After all, anything can go into a scalar and you always are talking about the exact type there... 21:45
And yes, it respects subtyping.
raydiak you're right...I played around with this for a while before I started Pray...ran in to weirdness with sigils and abandoned it in favor of a more simplistic model until I had a better hold on what I was doing 21:46
lue raydiak: it is %foo is MyHash unless S02 is wrong, that is :) 21:47
raydiak when coersion types work, will there be a simple way to do what I was talking about, without having to much about with the meta model to add a coercion routine to the Hash class?
jnthn lue: Yes, but I tend to guide people wrt what I know works ;) 21:48
lue: And I know for sure the "is" thingy doesn't :) 21:49
raydiak: Yes, the way to do it is to override invocation on the type object...
lue I suppose that is AReasonableWayToHelpPeople.
raydiak s/much about/muck about/
jnthn: I'm still trying to figure out what that means :) 21:51
21:51 Ben_Goldberg joined
jnthn r: class A { method &.(A:U: (%h)) { say 'here'; A.new } }; say A({}) # I think... 21:51
camelia rakudo-parrot 874e35, rakudo-jvm 874e35: OUTPUT«here␤A.new()␤»
raydiak ahhh, I'm starting to get it 21:52
jnthn++ timotimo++ lue++ # slowly bringing clarity to my fuzzy understandings :) 21:53
how does an advent post about building an array-based class sound? 21:54
walk through creating a simple Vector class that can be used interchangably with and as Arrays 21:55
masak sounds nice. 21:56
timotimo yeah, i'd like to read that
raydiak done deal, then, I'll go mark it down 21:57
masak \o/
timotimo already has 3 paragraphs on his post, but it seems unnecessarily long and focused too much on me :P 21:58
jnthn raydiak++
My post is coming along.
masak timotimo: add more kittens? 21:59
:P
dalek : b983ca2 | raydiak++ | misc/perl6advent-2013/schedule:
claim day 8: Array-based objects
raydiak left myself enough time for research and peer review since I'm still about chest-deep in ignorance 22:02
22:03 airdisa joined
masak 'night, #perl6 22:05
raydiak \o
22:05 airdisa_ joined 22:07 lizmat joined
lizmat has decommuted for today 22:07
jnthn lizmat: Hope it was a fine journey. 22:08
22:08 airdisa left
lizmat apart from needing 1.5 hours for the first 10 km., it was fine 22:08
45 minutes to spare was not enough, in the end we arrived 80+ minutes later than originally planned 22:09
jnthn That's a slow first 10km... 22:11
lizmat yeah, some accident on the outgoing road to Folkestone, percolated all the way down to the centre of London 22:13
ah well, at least we had enough time to make nice pictures of the Tower Bridge complex
:-)
lizmat continues to unpack and re-organize stuff 22:14
jnthn :)
22:19 Ben_Goldberg left 22:20 BenGoldberg joined 22:22 raiph joined
jnthn perl6advent.wordpress.com/?p=1897&a...eview=true # help me review :) 22:24
FROGGS jnthn: Error 404 - Not Found
well, at least spelled correctly :o)
22:25 arcterus joined
nwc10 404 here too. I like the way that the real URL is 404-or-not depending on logged-in-ness 22:26
22:26 ggoebel111 left, sqirrel left, Ben_Goldberg joined, BenGoldberg left 22:27 [Sno] joined
FROGGS hmmm, I thought I had write access.... apparently not 22:27
PerlJam jnthn: your 3rd point under "Why the JVM?" almost says this but what about "every other major dynamic language has an implementation that targets the JVM except Perl"?
22:28 BenGoldberg joined
[Coke] can someone invite me to the p6avent wordpress site? I have a login, but no privs. 22:29
FROGGS same for me...
jnthn PerlJam: Added "Furthermore, every other major dynamic language is on the JVM. So, why not Perl too?" 22:30
raiph (same 4 me on advent site) 22:31
22:31 Ben_Goldberg left 22:32 lowpro30_ left
PerlJam jnthn: I read the whole thing and didn't really see anything to comment on. Looks good! jnthn++ 22:32
[Coke]: what's your login? 22:33
FROGGS: you too? (or email0
FROGGS PerlJam: tleich
[Coke] PerlJam: will at coleda dot com
PerlJam invitations send 22:34
er, sent
jnthn re-read it a couple of times and is happy, unless anybody else finds anything :)
PerlJam: thanks for reading :)
22:35 kivutar joined
PerlJam jnthn: thanks for the advent article! :-) 22:35
jnthn It's scheduled.
Not sure what timezone this thing runs in...
Oh, looks like UTC...
22:37 stevan_ left
[Coke] jnthn++ looks good! 22:38
well, and so does the article.
PerlJam++ #ivite 22:39
invite, even
PerlJam [Coke]: btw, I saw a guy that kinda looked like you in a leather jacket at DFW yesterday. He was turned away from me and was thinking, "that's Coke!", then he turned completely facing me and he wasn't you. :) 22:40
and *I* was thinking ...
[Coke] if I ever head to DFW, I'll let y'all know here first. :) 22:42
~~ 22:43
lue jnthn++ don't see any problems. 22:47
22:48 rurban2 left 22:49 arcterus left 22:51 arcterus joined
lue Buf isn't supposed to be a class, is it? perlcabal.org/syn/S32/Containers.html#Buf 22:53
22:59 PacoAir left
jnthn Buf's a role, in Rakudo and by spec 23:03
at least, in S002 it is ;)
So yeah, I'd say S32 has it wrong there. 23:04
Buf.new([127], :size(16)) # should be Buf[int16]([127]) or so 23:05
23:09 jlaire left
lue jnthn: I'm fine with using Buf.meth as a way to say "things that do the Buf role.meth" though 23:10
jnthn Well, and if you Buf.new it just puns, too
TimToady sigh, after complete reconf, my compile still blows up in m-BOOTSTRAP with: Heap corruption detected: pointer 0x2aaaac255650 to past fromspace 23:14
jnthn TimToady: Then that probably is memory corruption, or some similar GC-related nasty (most probably a missing mark or wb somewhere) 23:15
TimToady: Which is sigh indeed, especially for me given I'm probably the one who'll be finding it... 23:16
...and made more fun given it doesn't happen here where I've got debugging tools to investigate :/ 23:17
23:18 rindolf left 23:21 jlaire joined
TimToady lemme see if I can reproduce it on my home machine, which I could give you a login on 23:21
23:26 stevan_ joined 23:29 xinming left 23:30 xinming joined
TimToady 'course, if the GC characteristics depend on available memory, that machine has more, so might not fail 23:30
lue Does the <:UniProp> syntax function like <rule> or <.rule> ? /me is guessing you'd have to do <.:UniProp> for non-capture 23:31
TimToady : doesn't capture 23:32
only initial alpha captures
the whole point of the . form is to remove the alpha from the front :)
other than that, they're both method calls
lue So you'd have to do (<:Prop>) or $<rule>=<:Prop> then. 23:33
TimToady (well, without . it will check for a lexically scoped method first, but it's still a call to a method)
<rule=:Prop> is fine
lue $<rule>= keeps me from having to remember how b in <a=b> gets interpreted :P 23:34
jnthn TimToady: No, it's not that smart yet, partly to try and make these things reproducable in the early days...
TimToady exactly as if the a= weren't there, it's recursive
jnthn: well, will still take a while to get to that point 23:35
23:35 stevan_ left
jnthn TimToady: No hurry, I'm not doing too well at staying concious here :) 23:35
23:35 stevan_ joined
TimToady figgered you'd pass out first :) 23:35
could be a bad patch of memory on my laptop, for all that 23:37
lue TimToady: good to keep in mind. I'll get more comfortable with <a=b> the less I see it as a simple <subrule> caller I'm sure.
23:37 Rotwang left
TimToady it's a recursive prefix, like <!...> or <?...> 23:38
diakopter "every other major dynamic language"
you can remove "major" and "dynamic" there
TimToady well, Swahili is a major language 23:39
jnthn Not sure I'm feeling dynamic enough for such a major change :P
TimToady but he's trying to name our major "competetors" without naming them :) 23:40
*competitors even
FROGGS TimToady: try to rebuild moar only with --optimize=1, and then just make m-install in nom
thas does it for me 23:41
jnthn FROGGS: Oh, you can reproduce the bug too?
FROGGS: And it's optimization level related?
FROGGS jnthn: yes, that is the bug from sunday or so
diakopter .. but major dynamic languages aren't Perl 6's competitors...
TimToady they think they are ;) 23:42
23:42 btyler left
TimToady no soap, still crashes 23:43
FROGGS :/ 23:44
jnthn FROGGS: Did you think a certain change was to blame, or it "just started happening?"
Sorry if I'm forgetting stuff. Brain is still an exhausted mess from November's $dayjob excesses... 23:45
FROGGS jnthn: I dunno which commit did it
jnthn k
FROGGS but my gut feeling tells me that this is the memory corruption that I have for days when I do -O3, just that it fails slightly different now 23:46
23:46 arcterus left, xenoterracide left
jnthn FROGGS: But there was a point we didn't reliably have the issue? 23:47
FROGGS the day before I "stashed" the stdhandles
(or the commit before)
lue Hey, I wrote a very rough first draft of S15! → gist.github.com/lue/7761244
23:48 woolfy joined
jnthn git --grep=std 23:49
lue (gah, I forgot the pragmas!) 23:50
TimToady kaboom, reproduce it 23:51
*ced
23:51 kivutar left
FROGGS jnthn: this breaks it: 05eaefe8431547a1b7eabfed44b959c3051ba4d2 23:52
TimToady jnthn: what do you want for an account name on my machine?
FROGGS .oO( beer? ) 23:53
maybe rootbeer :o)
jnthn TimToady: jnthn? :) 23:54
rootbeer is funny though :P
TimToady I don't know if it allows questions marks :P
lue
.oO(do usernames allow a ? at the end)
Bah, TimToady!
jnthn Uh...is it me or does MVM_file_get_stdstream do something terrible with an interior pointer... 23:55
oh hell, yes.
MVMOSHandleBody * const body = &result->body; 23:56
That gets a pointer into the middle of a GC-managed thing
TimToady cool!
jnthn body->u.handle->data = body;
That stores it.
FROGGS ha!
jnthn And later on, when the object moves...guess what... :)
FROGGS and I did not even wrote that bit! \o/
jnthn So yeah, that's sure fire memory corruption...
FROGGS jnthn**10
jnthn Me either. Wish I'd caught it in an earlier read-through of the code though. 23:57
I guess I assumed it was held at an extra LoI, a bit like SCRef does...
Of course, then it'd not need the & 23:58
It happens for both the TTY and named pipe case, it seems
But I wonder if other IO related things do it also...
TimToady guesses jnthn++ won't be wantin' that account... :)
23:59 zakharyas left
TimToady but it's there already, so anytime you want a Linux account on a 6-core machine... 23:59

Powered by Cro and the Raku® Programming Language.

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

Thank you!