»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg camelia perl6: ... | irclog: irc.perl6.org | UTF-8 is our friend!
Set by sorear on 25 June 2013.
00:00 atta joined, sunnavy joined, rurban joined 00:03 FROGGS joined 00:04 Vlavv` joined
[Coke] rt.perl.org/Ticket/Display.html?id=120413 00:06
TimToady * doesn't mean "infinite", it means "arbitrarily large"
which in this case, as in the case of subscripts, is naturally limited 00:07
00:08 jnap1 left
TimToady you probably aren't going to read an infinite number of elements from xx * either; it will generally be limited some other way :) 00:08
00:12 Psyche^ joined
[Coke] S02-types/mix.t has some dubious tests. 00:13
github.com/perl6/roast/blob/master...mix.t#L262 00:14
unless roll is guaranteeing the spread of returned values.
00:15 colomon joined
timotimo i'm pretty sure there's a way to properly test such things 00:15
00:16 Psyche^_ left
[Coke] github.com/perl6/roast/blob/master...mix.t#L277 - those 2 tests fail on rj. 00:17
00:18 rurban left
timotimo huh, so negative values? 00:19
r: my $m = {a => 100000000000, b => 1, c => -100000000000}.Mix; my @a = $m.roll: 100; say +@a.grep(* eq 'a'); say +@a.grep(* eq 'b'); 00:20
camelia rakudo-jvm 882e33: OUTPUT«0␤100␤»
..rakudo-parrot c27c7d: OUTPUT«100␤0␤»
timotimo wait what
r: my $m = {a => 100000000000, b => 1, c => -100000000000}.Mix; my @a = $m.roll: 25; say @a.grep(* eq 'a').join("")
camelia rakudo-parrot c27c7d: OUTPUT«aaaaaaaaaaaaaaaaaaaaaaaaa␤»
..rakudo-jvm 882e33: OUTPUT«␤»
timotimo r: my $m = {a => 100000000000, b => 1, c => -100000000000}.Mix; say $m.total 00:21
camelia rakudo-parrot c27c7d, rakudo-jvm 882e33: OUTPUT«1␤»
timotimo r: say 1.rand 00:22
camelia rakudo-parrot c27c7d: OUTPUT«0.730592795272486␤»
..rakudo-jvm 882e33: OUTPUT«0.5370525054723231␤»
[Coke] timotimo: once you've golfed it, please open an RT. ;)
timotimo the thick plottens
[Coke] looks like mixhash might have the same issue 00:23
timotimo they derive the same method from Mixy
r: my @pairs = (hello => 100, goodbye => -100); my $rand = 5.rand; my $seen = 0; my $roll; for @pairs -> $pair { next if ( $seen += $pair.value ) <= $rand; $roll = $pair.key; last; }; say $rand; say $roll; 00:24
camelia rakudo-parrot c27c7d: OUTPUT«2.63166102625959␤hello␤»
..rakudo-jvm 882e33: OUTPUT«2.481465557745795␤hello␤»
timotimo r: my @pairs = (hello => -100, goodbye => 100); my $rand = 5.rand; my $seen = 0; my $roll; for @pairs -> $pair { next if ( $seen += $pair.value ) <= $rand; $roll = $pair.key; last; }; say $rand; say $roll;
camelia rakudo-parrot c27c7d: OUTPUT«1.59497233502165␤(Any)␤»
..rakudo-jvm 882e33: OUTPUT«4.354170603807108␤(Any)␤» 00:25
timotimo r: my $m = {a => 100000000000, b => 1, c => -100000000000}.Mix; my @a = $m.roll: 25; say @a.join("")
camelia rakudo-jvm 882e33: OUTPUT«bbbbbbbbbbbbbbbbbbbbbbbbb␤»
..rakudo-parrot c27c7d: OUTPUT«aaaaaaaaaaaaaaaaaaaaaaaaa␤»
timotimo ... huh
[Coke] r: my $ascii-chars = [~] chr(0)..chr(0xFF); say $ascii-chars.comb(/<:Punctuation>/).join
camelia rakudo-parrot c27c7d, rakudo-jvm 882e33: OUTPUT«!"#%&'()*,-./:;?@[\]_{}¡«·»¿␤» 00:26
timotimo r: my $m = {a => 100000000000, b => 1, c => -100000000000}.Mix; say $m.pairs.perl;
camelia rakudo-parrot c27c7d: OUTPUT«("a" => 100000000000, "b" => 1, "c" => -100000000000).list␤»
..rakudo-jvm 882e33: OUTPUT«("b" => 1, "a" => 100000000000, "c" => -100000000000).list␤»
timotimo oh, there's a b in the middle, didn't even see that
[Coke] ^^ that fails for me locally and on roast.
timotimo [Coke]: mumble mumble something something lue changed the test to fix the expectation
[Coke] er, s/fails/doesn't pass the spectest/ 00:27
timotimo i'm not aware how negative values are supposed to work, to be honest 00:28
lue I will valiantly hold that ¶ and § *are* punctuation characters. (They surround the « )
[Coke] dammit, I just amex commercial email: "① Join ② Write a review ③ Get $5 credit"
timotimo the specs don't say anything about mixes and negative quantities, at least nothing that i found directly 00:29
lue (They surround the « in the test's generated list, I should say)
timotimo if negative values mean "never gets rolled" ... 00:30
[Coke] timotimo: spec or not, wierd that rj & rp disagree
timotimo then the negative values should be culled from the roll. and they should not contribute to the total that's used to figure out what number to roll
well, here's what happens:
the total is the sum of the values, which is 1
it rolls a number between 0 and the total, then looks through the list of pairs 00:31
the first one, summing up all seen values, that's bigger than the random number gets taken
00:31 rurban joined
[Coke] ¯\_(ツ)_/¯ 00:31
00:31 benabik joined
timotimo but since the number is, wrongly, between 0 and 1 instead of -1000000000000000 and 1000000000000000000001, it will just take the first low number 00:31
... er 00:32
well, i know that's not quite accurate, but you could add and subtract ones and multiply by -1 in different places in my description and eventually you'll find the truth
[Coke] lue: should emacs git skips go in teh project, or in your local git file? (I vote the latter, but not vehemently)
timotimo point is: the algorithm behaves erratically if negative values are in the mix. 00:33
[Coke] timotimo: but why? is there any parrot specific code anywhere?
it's not failing on parrot, and is repeatably failing on the jvm.
timotimo i guess it's the hash that's used to store the data
[Coke] timotimo: ah, fair enough.
timotimo r: my $m = {a => 100000000000, b => 1, c => -100000000000}.Mix; say $m.pairs.perl;
camelia rakudo-parrot c27c7d: OUTPUT«("a" => 100000000000, "b" => 1, "c" => -100000000000).list␤»
..rakudo-jvm 882e33: OUTPUT«("b" => 1, "a" => 100000000000, "c" => -100000000000).list␤»
dalek kudo/nom: e455e82 | jonathan++ | src/vm/jvm/core/Threading.pm:
Flesh out Thread a little more; add $*THREAD.

This provides support for giving threads names, getting a thread ID, provides a sensible stringification and enables $*THREAD as a way to get a Thread instance for the currently executing thread (including making it work on the initial thread, so that isn't a special case).
timotimo this is the trigger, the error stems from not having specced mixes well enough :P 00:34
dalek ast: 362acca | jonathan++ | S17-concurrency/thread.t:
Some initial tests for the Thread class.
lue [Coke]: I always just put them in the project I'm working on. I *could* define it globally, but I haven't :) .
timotimo huh. we could totally have a sub "isn't" rather than "isnt" 00:36
[Coke] so save yourself the trouble for next time and put them in your file. :)
jnthn timotimo: I actually type isn't when first writing the tests :P
[Coke] r: sub isn't() { say "is so"}; isn't;
jnthn *typed
camelia rakudo-parrot c27c7d, rakudo-jvm 882e33: OUTPUT«is so␤»
[Coke] wanders off 00:37
dalek kudo/nom: 78e4c21 | jonathan++ | src/vm/jvm/core/Threading.pm:
Explain code's funkyness.
 00:39
00:40 mattp_ joined
jnthn 'night, #perl6 00:47
timotimo r: my @foo = <foo bar baz>; my $rex = @foo.keys.map({ qq/ ["\@foo[$^idx]" || \{ die "\@foo[$^idx] is missing" }] / }).join(".*?"); say "foo meep bar meep baz" ~~ /<$rex>/ 00:48
camelia rakudo-jvm 882e33: OUTPUT«Cannot call 'postcircumfix:<[ ]>'; none of these signatures match:␤:(\SELF, int $pos)␤:(\SELF, int $pos, Mu \$BIND)␤:(\SELF, int $pos, :delete(:$delete)!, *%other)␤:(\SELF, int $pos, :exists(:$exists)!, *%other)␤:(\SELF, int $pos, :kv(:$kv)!, *%ot…»
..rakudo-parrot c27c7d: OUTPUT«Cannot call 'postcircumfix:<[ ]>'; none of these signatures match:␤:(\SELF, int $pos)␤:(\SELF, int $pos, Mu \$BIND)␤:(\SELF, int $pos, :delete(:$delete)!, *%other)␤:(\SELF, int $pos, :exists(:$exists)!, *%other)␤:(\SELF, int $pos, :kv(:$kv)!, *…»
timotimo r: my $foo = "hi"; my $pat = '$foo'; "hi" ~~ /<$pat>/ 00:50
camelia rakudo-parrot c27c7d: OUTPUT«Nominal type check failed for parameter '$var'; expected Any but got NQPMu instead␤ in method INTERPOLATE at gen/parrot/CORE.setting:12135␤ in regex at eval_0:1␤ in method ACCEPTS at gen/parrot/CORE.setting:12307␤ in method ACCEPTS at gen/pa…»
..rakudo-jvm 882e33: OUTPUT«Nominal type check failed for parameter '$var'␤ in method ACCEPTS at gen/jvm/CORE.setting:12277␤ in block at gen/jvm/CORE.setting:12175␤ in block at gen/jvm/CORE.setting:7216␤ in block at gen/jvm/CORE.setting:7080␤ in method gimme at gen/j…»
timotimo well, that blew up in a different way the last time i tried it ...
diakopter timotimo: isn't it, like, early morning there? :) 01:23
timotimo errr, i should go 01:24
01:24 raiph joined
diakopter timotimo: eh? 01:25
01:26 risou_awy is now known as risou 01:28 ggoebel13 joined 01:29 nofks joined 01:33 ssutch joined 01:41 grondilu joined
grondilu rn: constant %h = <foo bar>; say %h<foo>; 01:42
camelia niecza v24-98-g473bd20: OUTPUT«bar␤»
..rakudo-jvm 882e33: OUTPUT«postcircumfix:<{ }> not defined for type Parcel␤ in block at /tmp/77bOoURBF_: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␤ in any …»
..rakudo-parrot 78e4c2: OUTPUT«postcircumfix:<{ }> not defined for type Parcel␤ in method gist at gen/parrot/CORE.setting:11956␤ in method gist at gen/parrot/CORE.setting:1014␤ in sub say at gen/parrot/CORE.setting:12856␤ in block at /tmp/EkvCe1k94M:1␤ in any at /tmp/…»
01:42 risou is now known as risou_awy
grondilu rn: constant %h = hash <foo bar>; say %h<foo>; 01:42
camelia rakudo-parrot 78e4c2, rakudo-jvm 882e33, niecza v24-98-g473bd20: OUTPUT«bar␤»
grondilu I thought the conversion was implicit, same as with:
rn: my %h = <foo bar>; say %h<foo>; 01:43
camelia rakudo-parrot 78e4c2, rakudo-jvm 882e33, niecza v24-98-g473bd20: OUTPUT«bar␤»
grondilu see? no need to use 'hash' here
BenGoldberg rn: constant h = hash <a b>; say h<a>; 01:44
camelia rakudo-parrot 78e4c2, rakudo-jvm 882e33, niecza v24-98-g473bd20: OUTPUT«b␤»
BenGoldberg rn: constant !h = hash <a b>; say !h<a>; 01:45
camelia rakudo-parrot 78e4c2: OUTPUT«===SORRY!=== Error while compiling /tmp/8OPpioJVBl␤Missing initializer on constant declaration␤at /tmp/8OPpioJVBl:1␤------> constant ⏏!h = hash <a b>; say !h<a>;␤ expecting any of:␤ statement lis…»
..rakudo-jvm 882e33: OUTPUT«===SORRY!=== Error while compiling /tmp/IcCIcSQjrZ␤Missing initializer on constant declaration␤at /tmp/IcCIcSQjrZ:1␤------> constant ⏏!h = hash <a b>; say !h<a>;␤ expecting any of:␤ statement list…»
..niecza v24-98-g473bd20: OUTPUT«===SORRY!===␤␤Missing initializer on constant declaration at /tmp/dwmeiy39wu line 1:␤------> constant ⏏!h = hash <a b>; say !h<a>;␤␤Negation metaoperator not followed by valid infix at /tmp/dwmeiy39wu li…»
BenGoldberg rn: constant @h = hash <a b>; say @h<a>;
camelia rakudo-parrot 78e4c2, rakudo-jvm 882e33: OUTPUT«b␤» 01:46
..niecza v24-98-g473bd20: OUTPUT«Unhandled exception: Cannot use hash access on an object of type List␤ at <unknown> line 0 (ExitRunloop @ 0) ␤ at /home/p6eval/niecza/lib/CORE.setting line 359 (Any.at_key @ 9) ␤ at <unknown> line 0 (ExitRunloop @ 0) ␤ at /tmp/AF2Gh9gT4N li…»
BenGoldberg rn: my \h = hash <a b>; say h<a>;
camelia rakudo-parrot 78e4c2, rakudo-jvm 882e33, niecza v24-98-g473bd20: OUTPUT«b␤»
01:52 bluescreen100 left 02:05 rurban left 02:09 colomon left
Mouq rn: my term:<!h> = hash <a b>; say !h<a>; 02:11
rn: my \term:<!h> = hash <a b>; say !h<a>;
camelia niecza v24-98-g473bd20: OUTPUT«===SORRY!===␤␤Malformed my at /tmp/pZZxUL91cQ line 1:␤------> my⏏ term:<!h> = hash <a b>; say !h<a>;␤␤Parse failed␤␤»
..rakudo-parrot 78e4c2: OUTPUT«===SORRY!===␤Type 'term:<!h>' is not declared␤at /tmp/Ac5r0hplvp:1␤------> my term:<!h>⏏ = hash <a b>; say !h<a>;␤Malformed my␤at /tmp/Ac5r0hplvp:1␤------> my term:<!⏏h> = hash <a b>; sa…»
..rakudo-jvm 882e33: OUTPUT«===SORRY!===␤Type 'term:<!h>' is not declared␤at /tmp/p2jtM_YEqr:1␤------> my term:<!h>⏏ = hash <a b>; say !h<a>;␤Malformed my␤at /tmp/p2jtM_YEqr:1␤------> my term:<!⏏h> = hash <a b>; say !…»
niecza v24-98-g473bd20: OUTPUT«b␤» 02:12
..rakudo-parrot 78e4c2, rakudo-jvm 882e33: OUTPUT«===SORRY!===␤No such method 'ast' for invocant of type 'NQPMu'␤»
Mouq :( That used to work on rakudo
r: my \term:<º> = {say 'alive'}; º()
camelia rakudo-parrot 78e4c2, rakudo-jvm 882e33: OUTPUT«===SORRY!===␤No such method 'ast' for invocant of type 'NQPMu'␤» 02:13
02:14 grondilu left 02:31 GlitchMr left, araujo left 02:32 araujo joined 02:33 johnny5_ left 02:34 johnny5_ joined 02:37 GlitchMr joined
retupmoca github.com/perl6/nqp/pull/147 <-- makes it easier for me to write my git-tracking gentoo ebuilds 02:39
02:41 PacoAir left 02:42 PacoAir joined 02:47 dayangkun left 02:56 raiph left 02:57 colomon joined 03:02 rurban joined 03:03 benabik left 03:06 vael left 03:07 rurban left 03:09 colomon left 03:13 araujo left 03:14 araujo joined 03:18 colomon joined 03:26 risou_awy is now known as risou 03:29 colomon left 03:33 nofks left 03:35 johnny5_ left 03:36 johnny5_ joined 03:42 johnny5_ left 03:43 risou is now known as risou_awy, colomon joined, johnny5_ joined 04:03 rurban joined 04:08 rurban left 04:19 Mouq left 04:20 jeffreykegler joined 04:27 jeffreykegler left 04:46 GlitchMr left 04:47 GlitchMr joined 04:48 logie joined 04:52 logie left
TimToady n: r: my \term:<º> = {say 'alive'}; º() 04:53
camelia niecza v24-98-g473bd20: OUTPUT«===SORRY!===␤␤Undeclared routine:␤ 'º' 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_uni…»
TimToady nr: my \term:<º> = {say 'alive'}; (º)() 04:54
camelia niecza v24-98-g473bd20: OUTPUT«alive␤»
..rakudo-parrot 78e4c2: OUTPUT«===SORRY!===␤No such method 'ast' for invocant of type 'NQPMu'␤»
..rakudo-jvm 882e33: OUTPUT«java.lang.OutOfMemoryError: Java heap space␤ in (gen/jvm/CORE.setting)␤ in (gen/jvm/ModuleLoader.nqp:349)␤ in load_setting (gen/jvm/ModuleLoader.nqp:340)␤ in (src/Perl6/World.nqp:304)␤ in load_setting (src/Perl6/World.nqp:300)␤ in (src…»
TimToady nr: say 'º' ~~ /\w/; # why º() doesn't work
camelia rakudo-parrot 78e4c2, niecza v24-98-g473bd20: OUTPUT«｢º｣␤␤» 04:55
..rakudo-jvm 882e33: OUTPUT«java.lang.OutOfMemoryError: Java heap space␤ in (gen/jvm/CORE.setting)␤ in (gen/jvm/ModuleLoader.nqp:349)␤ in load_setting (gen/jvm/ModuleLoader.nqp:340)␤ in (src/Perl6/World.nqp:304)␤ in load_setting (src/Perl6/World.nqp:300)␤ in (src…»
TimToady (other than the bug)
04:55 preflex_ joined, ChanServ sets mode: +v preflex_
TimToady nr: my \term:<°> = {say 'alive'}; °(); # this should work because ° is not \w 04:57
diakopter hee
camelia niecza v24-98-g473bd20: OUTPUT«alive␤»
..rakudo-parrot 78e4c2: OUTPUT«===SORRY!===␤No such method 'ast' for invocant of type 'NQPMu'␤»
..rakudo-jvm 882e33: OUTPUT«java.lang.OutOfMemoryError: Java heap space␤ in (gen/jvm/CORE.setting)␤ in (gen/jvm/ModuleLoader.nqp:349)␤ in load_setting (gen/jvm/ModuleLoader.nqp:340)␤ in (src/Perl6/World.nqp:304)␤ in load_setting (src/Perl6/World.nqp:300)␤ in (src…»
04:57 preflex left
TimToady (other than the bug :) 04:57
04:57 preflex_ is now known as preflex
TimToady (the reason º() can't work is that 'º(' is a longer token than 'º' is 05:04
)
05:04 rurban joined 05:06 rurban1 joined 05:09 rurban left 05:10 rurban1 left 05:11 rurban joined 05:13 johnny5_ left 05:14 johnny5_ joined, BenGoldberg left 05:18 SamuraiJack_ joined 05:23 johnny5_ left 05:24 johnny5_ joined 05:27 risou_awy is now known as risou 05:40 Vendethi_ joined 05:43 risou is now known as risou_awy 05:47 johnny5_ left 05:48 johnny5_ joined
dalek kudo/nom: 8e9470c | moritz++ | src/Perl6/Grammar.nqp:
revert accidentally commited work towards no-strict, Mouq++
 06:02
06:08 rindolf joined 06:09 rurban left 06:17 rindolf left, darutoko joined 06:33 daniel-s__ joined 06:39 rurban joined 06:41 FROGGS left 06:42 xenoterracide left 06:44 rurban left 06:49 pecastro left 06:53 FROGGS joined 06:57 xenoterracide joined, dansamo joined 07:09 kaleem joined 07:15 ssutch left 07:19 ssutch joined 07:24 ssutch left 07:28 risou_awy is now known as risou 07:33 arnsholt_ is now known as arnsholt 07:40 rurban joined 07:45 risou is now known as risou_awy, rurban left
dalek p: f81e259 | (Andrew Egeler)++ | tools/build/ (5 files):
Add DESTDIR support for JVM/Moar

install-jvm-runner and install-moar-runner were previously not aware of DESTDIR. DESTDIR was also loaded into PREFIX instead of being used directly in *-install.		 07:55
p: 1168608 | moritz++ | tools/build/ (5 files):
Merge pull request #147 from retupmoca/master

Add DESTDIR support for JVM/Moar
07:59 FROGGS[mobile] left 08:04 baest_ is now known as baest, zakharyas joined 08:13 lizmat joined 08:26 Vendethi_ left
masak good moarning, #perl6 08:28
FROGGS morning 08:31
08:35 iSlug joined 08:36 dansamo left 08:41 rurban joined
lizmat good *, #perl6 08:43
trying to wake up after AmsterdamX meeting + barring afterwards 08:44
08:44 Vendethi_ joined
masak wow, .pm meeting *and* gymnastics! 08:44
:P
lizmat mostly of the wrist :-) 08:45
08:46 rurban left
masak very important muscles in that area 08:46
can also be used for calligraphy
lizmat or darts: www.youtube.com/watch?v=sHnBppccI0o
masak or employee slapping 08:49
FROGGS hehe 08:51
masak www.satirewire.com/news/032700/sati...appy.shtml 08:52
FROGGS I need some lunch ideas for tomorrow, any suggestions? :o) 08:55
masak FROGGS: fry some vegetables in yummy sauces. 08:56
FROGGS hmm, vegetables doesn't work that for my kids... except potatoes of course 08:57
I'd like it though
08:58 markov joined
FROGGS ahh, but I remember that I wanted to make a carrot+apple salad 08:58
08:59 wbill left
FROGGS I think I do mashed potatoes and sausages to that 09:00
lizmat sounds like "stamppot" to me :-)
or "stoemp" as they would say in Belgium
FROGGS no, not quite
lizmat: more like that: www.seelenfarben.de/lb201305/kocht130530.htm 09:01
masak .oO( is "stamppot" at all related to "adapot"? )
FROGGS but just that the mashed potatoes actually look tasty 09:02
masak: I have to think about the old Tripod movie when I hear adapot 09:03
(... which was pretty scary when I was a kid)
uhh www.youtube.com/watch?v=kMirslevFmw 09:04
lizmat FROGGS: indeed, not at all like stamppot :-) 09:07
FROGGS well, without other vegetables, you see *g*
09:08 jaffa4 joined
jaffa4 hi all 09:08
masak hi jaffa4
jaffa4 I tried to compile rakudo star. it crashes badly
masak :/
jaffa4 when I execute perl Configure.pl --gen-parrot 09:09
masak huh.
that early?
FROGGS jaffa4: no-paste/gist the output?
moritz jaffa4: how did it crash?
jaffa4 .Can't spawn ".\test_2328.exe": Inappropriate I/O control operation at lib/Parrot/Configure/Utils.p
exception code 0000005
exception code c0000005 09:10
fault Module Name :StackHash_0a9e
FROGGS okay
jaffa4 :\standalone\rakudo-star-2013.10\rakudo-star-2013.10\parrot\test_2328.exe<<<<< 09:11
FROGGS jaffa4: do you have ActivePerl or strawberryperl installed?
(Perl 5 that is)
jaffa4 I have everything, perl 16_2 activeperl is active now in the path 09:12
FROGGS jaffa4: if you have another one, can you switch to it and test again? 09:13
I don't know what else to suggest
I installed a parrot 5.9.0 yesterday without problems (on a win7 x64) 09:14
JimmyZ_ I guess you should shutdown the antivir
antivirus 09:15
test_2328.exe may be consider as virus if you didn't shutdown it 09:16
FROGGS true
jaffa4: but also, what linenumber follows: .Can't spawn ".\test_2328.exe": Inappropriate I/O control operation at lib/Parrot/Configure/Utils.p
JimmyZ_ It may be blocked by the anti-virus 09:18
jaffa4 it works withj Straberry
at lib/Parrot/Configure/Utils.pm line 88.
what is the state of Perl6 new implementation> 09:20
FROGGS jaffa4: please rephrase the question 09:21
jaffa4 There is new virtual machine to be made
for rakudo.
FROGGS jaffa4: ahh, yes, MoarVM
moritz nqp is bootstrapped on moarvm
rakudo doesn't yet build on moarvm
but it's being worked on
FROGGS it is not a new Perl6 implementation, because rakudo is the implementation, which is left almost untouched 09:22
jaffa4 new backend
JimmyZ_ new World
FROGGS rakudo compiles a few things of itself already though on MoarVM
moritz new world order
jaffa4 What is going to be better?
FROGGS jaffa4: for what task? 09:23
jaffa4 in the virtula mechine?
FROGGS jaffa4: what do you want to achieve?
if you need to execute Perl6 of a coffee maker you want JVM backend
JimmyZ_ Everything. 09:24
FROGGS but in my dreams MoarVM will beat all other backend, but, well, dreams
jaffa4 on a coffee maker?
FROGGS JVM runs everywhere
moritz doubts that most coffee makers will have enough RAM to run rakudo
FROGGS and if you want to run perl6 in your browser, you want to wait for the JavaScript backend
JimmyZ_ buy new RAM :P 09:25
FROGGS jaffa4: it really depends on the task
jaffa4 running an IDE?
moritz I guess the JVM will beat any other backend for long running programs in the short and mid term
jaffa4 or command line tool?
FROGGS atm parrot might be better for short scripts or one-liners 09:26
jaffa4 for distribution?
..\..\..\..\3rdparty\dyncall\dyncallback\libdyncallback_s.a(dyncall_callback.o)(.text+0x333):dyncall_callback.c: undefined reference to `dcCallbackThunkEntry'
FROGGS for long running daemons the JVM is better
JimmyZ_ FROGGS: btw: #define add_entry missing MVMROOT
FROGGS JimmyZ_: in config.c.in? ohh... 09:27
jaffa4 I cannot compile again... see my last paste
JimmyZ_ FROGGS: yeah
root temp push will be better
FROGGS jaffa4: what windows version is that? and what C compiler?
JimmyZ_: patch it :o)
jaffa4 gcc (GCC) 3.4.5 (mingw-vista special r3), wiin 7 64bit 09:28
FROGGS somebody else uses gcc on windows?
09:29 risou_awy is now known as risou
FROGGS damn it, must go now... 09:29
bbi1h
09:30 FROGGS[mobile] joined, [particle]1 left 09:34 frdmn joined 09:38 Vendethi_ left 09:39 Vendethi_ joined 09:42 rurban joined 09:44 dakkar joined, Vendethi_ left 09:45 risou is now known as risou_awy 09:47 rurban left 10:12 k1lldash9 left 10:20 Vendethi_ joined 10:34 daxim joined 10:37 colomon left 10:38 colomon joined
xenoterracide outside of source code reading is there a way to get a list of what I can do with the metaobject? e.g. something like Obj.^.^methods 10:39
masak r: class A { method foo {} }; say A.^methods 10:42
camelia rakudo-jvm 882e33: OUTPUT«(timeout)»
..rakudo-parrot 8e9470: OUTPUT«foo␤»
10:43 rurban joined
masak p: class A { method foo {} }; say A.HOW.^methods 10:43
camelia rakudo-parrot 8e9470: OUTPUT«No such method 'gist' for invocant of type 'NQPRoutine'␤ in method gist at gen/parrot/CORE.setting:6730␤ in method gist at gen/parrot/CORE.setting:1014␤ in sub say at gen/parrot/CORE.setting:12856␤ in block at /tmp/hvdqRGiKrd:1␤ in any a…»
masak p: class A { method foo {} }; say .Str for A.HOW.^methods
camelia rakudo-parrot 8e9470: OUTPUT«No such method 'Str' for invocant of type 'NQPRoutine'␤ in block at /tmp/zwDsfq7m1O:1␤ in any at /tmp/zwDsfq7m1O:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPHLL.nqp:1133␤ in any evalfiles at gen/p…»
masak p: class A { method foo {} }; say .name for A.HOW.^methods
camelia rakudo-parrot 8e9470: OUTPUT«archetypes␤new␤new_type␤parameterize␤add_fallback␤compose␤roles␤role_typecheck_list␤concretization␤is_composed␤can␤setup_junction_fallback␤find_method_fallback␤has_fallbacks␤set_name␤name␤set_docs␤docs␤ver␤auth␤set…»
masak xenoterracide: like that?
10:44 rurban1 joined 10:46 spider-mario joined 10:47 rurban left 10:48 rindolf joined 10:49 rurban1 left 10:52 pdurbin left
lizmat p: class A { method foo {} }; say A.HOW.^methods>>.name 11:04
camelia rakudo-parrot 8e9470: OUTPUT«archetypes new new_type parameterize add_fallback compose roles role_typecheck_list concretization is_composed can setup_junction_fallback find_method_fallback has_fallbacks set_name name set_docs docs ver auth set_ver set_auth set_stash_type add_stash a…» 11:05
11:05 Ulti left
xenoterracide masak: yeah 11:06
lizmat mostly offline while on route to Salzburg& 11:07
11:07 lizmat left
masak lizmat: is Salzburg the result of two neighbor cities Säureburg and Baseburg that got too close together? :P 11:14
11:17 woolfy left 11:21 Vendethi_ left 11:30 risou_awy is now known as risou
masak moritz: I'm doing some very intense iterative TDD development right now at $dayjob. I find 'git commit -a' to be very suitable for that. 11:36
(because I keep editing the same test file and implementation file, and it's kind of a closed system where no other files change) 11:37
11:37 panchiniak left 11:38 fhelmberger joined 11:42 fhelmberger left 11:45 fhelmberger joined, rurban joined 11:48 risou is now known as risou_awy
dalek kudo/nom: 1255967 | jonathan++ | src/vm/jvm/core/Threading.pm:
Fix outstanding bug in scheduler.
 11:48
kudo/nom: d9e79fb | jonathan++ | src/vm/jvm/core/Threading.pm:
Add Scheduler role; provide uncaught error model.

Previously, unhandled exceptions in scheduled items could just vanish. Now they will not, and there is an API for setting a per-item handler as well as the default global fallback one. Since much of this can be done independent of an individual scheduler, this is mostly pulled out into a role.
ast: c4a73f6 | jonathan++ | S17-concurrency/scheduler.t:
Add some initial scheduler tests.
 11:49
11:49 rurban left 11:52 dansamo joined 11:58 Vendethi_ joined 12:01 telex left 12:02 telex joined 12:07 xinming left 12:09 pernatiy left 12:11 xinming joined 12:15 tipdbmp left 12:19 tipdbmp joined 12:30 skromnica joined, skromnica left 12:34 kaleem left 12:38 FROGGS left 12:40 FROGGS joined
FROGGS jnthn: you are giving a talk about concurrency? 12:41
12:46 rurban joined
[Coke] masak,moritz: I almost -always- use git commit -a when I'm working. very very occasionally I'll use -p if I want to skip whitespace or debug changes. 12:49
FROGGS++ (observing jnthn's tuits are sometimes talk shaped) 12:50
12:50 rurban left
nwc10 FROGGS: act.useperl.at/apw2013/talk/5183 12:51
[Coke] heads off to a day in the ColdFusion/SQL/XHTML mines. 12:54
13:01 FROGGS left, FROGGS joined
masak [Coke]: I don't really like -a, and most of my workflows stay clear of it. 13:01
13:06 kaleem joined 13:07 ajr joined, ajr is now known as Guest87010 13:08 Guest87010 is now known as ajr_
arnsholt My use of commit -a varies 13:08
Sometimes I've been focussed on one thing, in which case I'll often use -a 13:09
But if I've been hacking on various things and need to partition stuff into several commits, I don't (obviously)
13:10 panchiniak joined 13:12 SamuraiJack_ left
masak right. 13:16
arnsholt In the latter case add -p tends to crop up a fair bit as well =) 13:17
masak my threshold for "need to make sure I know what I'm committing" is probably much lower than other people's.
arnsholt Could be
I usually do that with diff 13:18
13:20 pernatiy joined
masak right, me too. 13:23
I usually do status + diff + 'git commit dir1 dir2 ...'
hoelzro git-add -p all the way! 13:24
masak :)
13:28 jnap joined 13:31 risou_awy is now known as risou 13:32 pernatiy_ joined
arnsholt Yeah, that's pretty similar to my workflow as well 13:34
13:35 pernatiy left 13:39 btyler joined 13:41 benabik joined 13:47 risou is now known as risou_awy, [particle] joined 13:50 Vendethi_ left 13:54 bluescreen10 joined 14:03 rurban joined 14:05 Vendethi_ joined
jnthn FROGGS: How'd you guess... :) 14:14
14:22 pernatiy joined 14:23 pernatiy_ left 14:32 Vendethi_ left
FROGGS[mobile] jnthn: I am Trelawny's best boy :o) 14:33
14:38 Tom_____ joined 14:39 Tom_____ is now known as tgt
masak FROGGS[mobile]: that's a seriously scary notion... :) 14:40
14:40 zakharyas left
FROGGS[mobile] masak: that is unintended I'd say 14:40
14:42 JTAA6C59 joined 14:46 zakharyas joined 14:54 xinming left 14:56 xinming joined 15:01 xinming left, xinming joined 15:11 pernatiy_ joined 15:12 denis_boyun joined 15:14 denis_boyun left, pernatiy left 15:16 Vendethi_ joined 15:18 denis_boyun joined 15:22 rindolf left 15:36 denis_boyun left 15:43 denis_boyun joined 15:44 wbill joined, Vendethi_ left 15:45 minimus joined 15:59 kaare_ joined 16:02 pernatiy joined 16:03 kaleem left 16:05 pernatiy_ left
masak mjd++ # perl.plover.com/yak/Moonpig/ 16:06
diakopter rjbs++ # MoonPig 16:10
rjbs mjd++ moonpig++ 16:12
We're actually just discussing Moonpig in IM, mjd and I. 16:13
16:14 logie joined
masak I'm glad to note that many of the hard-won lessons of those slides are also themes of our courses at $dayjob. :) 16:15
the "heartbeat" design is interesting. 16:16
16:18 tobyink joined
rjbs I forget whether, in those slides, heartbeat explosion is covered. 16:18
16:19 logie left
rjbs A ledger gets heartbeaten. It sends it to all its components in turn. Some of them (consumers) produce synthetic heartbeats across date ranges to catch up. 16:19
(Hm. Did we end up with heartbeats for those, or distinct sub-events? I don't recall.)
tobyink Are attributive parameters (S06) actually useful for much, other than initializers? 16:22
jnthn tobyink: That's where I typically use them. They're pretty nice there :) 16:23
tobyink I've ported most of the Perl 6 method signatures stuff to Perl 5 in metacpan.org/release/Kavorka but attributive parameters seem more effort than they're worth. I think in Moose/Mouse/Moo people tend to just rely on the default initialization. 16:25
16:26 fhelmberger left, fhelmberger joined
jnthn Yeah, they kinda make the object model and parameter handling intersect, and while in Perl 6 we have both under control at the language level, I guess in Perl 5 that's a more open-ended problem... 16:28
tobyink If you restrict it to working only in Moose classes, it's not technically a big problem. You just tie a scalar variable to the Moose::Meta::Attribute's get_value/set_value methods. A twigil can theoretically be handled (rafl has a module on github), but I'd probably stick with the "$" sigil. 16:31
16:34 nnunley joined
TimToady well, we've gotten burned before on things that are more ambiguous in p5 than in p6, like smartmatch vs types 16:35
if someone uses $attr in a signature, what do they really mean? with the twigil the intent is clearer 16:36
tobyink I was thinking along the lines of ($self: $attr is attribute) 16:37
TimToady ah...well, that is almost longer than what it is replacing
16:38 nnunley left
tobyink And/or allowing ($self: $.attr) in the signature, but establishing a plain $attr variable in the scope of the sub. 16:38
16:38 nnunley joined, [Sno] left
TimToady but then do you auto-assign or not? 16:38
16:39 fhelmberger_ joined 16:41 [Sno] joined
tobyink I don't know. I think the use cases for it in p5 are minimal, and the issues make it more effort than it's worth. My time is probably better spent on implementing return types or somesuch. 16:41
16:41 SamuraiJack_ joined 16:43 fhelmberger left 16:44 fhelmberger_ left
TimToady probably agrees :) 16:45
dalek p: f72fb2b | (Tobias Leich)++ | src/vm/moar/QAST/QASTCompilerMAST.nqp:
handle custom_args case by allowing any number of args
 16:46
tobyink Which brings me to another question. How do return types work in list context?
moritz tobyink: t
sorry 16:47
tobyink: Lists are also objects in Perl 6
tobyink: so if a subroutine returns a list of things, its return type should better be List
16:47 pernatiy_ joined
moritz ♪ what a wonderful world ♫ 16:48
tobyink So the return type would be something like List<Int>?
timotimo it's spelled List[Int] here, or List of Int, but probably you want Positional[Int] instead, so it will accept parcels, arrays and lists? 16:49
16:50 pernatiy left
diakopter r: my Positional[Positional] @b; @b := []; 16:52
camelia rakudo-jvm 882e33: OUTPUT«Type check failed in binding; expected 'Positional[Positional[Positional]]' but got 'Array'␤ in block at /tmp/J7YWQfpFku: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 a…»
..rakudo-parrot d9e79f: OUTPUT«Type check failed in binding; expected 'Positional[Positional[Positional]]' but got 'Array'␤ in block at /tmp/B8yDFzFRcr:1␤ in any at /tmp/B8yDFzFRcr:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPHLL.n…»
timotimo hmm. even if we get filename sanitation for the comparator, we'll still get different line numbers for things due to the conditional compilation infrastructure we have 16:53
moritz and what diakopter declares is an Array of Positional of Positional :-)
tobyink Hmmm... I'll think about how that would translate to p5.
timotimo i wonder if it makes sense to make that thing emit blank lines to match up line numbers between outputs?
r: my Positional[Positional] $b; $b := []; 16:54
moritz tobyink: probably to either declare a List return type, or to not apply a return type at all
camelia rakudo-jvm 882e33: OUTPUT«Type check failed in binding; expected 'Positional[Positional]' but got 'Array'␤ in block at /tmp/bNEx_IvsKY: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/st…»
..rakudo-parrot d9e79f: OUTPUT«Type check failed in binding; expected 'Positional[Positional]' but got 'Array'␤ in block at /tmp/oQ3yi9Ch1S:1␤ in any at /tmp/oQ3yi9Ch1S:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPHLL.nqp:1133␤ …»
timotimo hm. well. our constants are never typed, right?
moritz they aren't
arnsholt Not yet, IIRC 16:55
timotimo i'm wondering if and when i should start thinking about making that change ... there's stuff like turning <foo bar baz> into a Positional[Str] at compile time, but also there's the thing where <1/5> is supposed to be something like a RatStr
moritz rjbs: as someone who has to maintain the accounting & billing code at $work (mostly a single subroutine with 2.5k lines of tightly coupled Perl code), I really enjoy seing a sane design in that area
arnsholt Due to package variables not having containers or something along those lines 16:56
timotimo oops. i didn't mean "our constants" in monospaced font
i mean "the constants we create at compile time as a result of parsing"
my bad. perl is too much like english to be embedded into english without some kind of quoting :P
moritz giving those narrower types looks very dangerous 16:57
my $x = [1, 2, 3]; $x.push: 'foo'; # BOOM!
timotimo ah, hm.
16:57 kaleem joined
rjbs moritz: I feel your pain. 16:57
timotimo so it should only happen for actual constants
rjbs moritz: We still have one legacy billing system in place, an old version of FreeSide. Pain. 16:58
timotimo but that's hard to figure out. when is a <foo bar baz> just passed into a function and then introspected and when is it re-used to house some integers as well?
i suppose i should drop it for the time being.
ah, "allomorphic" is the word i was looking for for the other part of the questio 16:59
moritz timotimo: <foo bar baz> is a parcel (of non-containers), and thus immutable
16:59 pernatiy_ left, panchiniak left
moritz rjbs: but you are replacing it, I hope? 17:00
TimToady tobyink: please note that the return model is very different in p6, since list context is lazily applied to the return value, not known (mostly) at compile time
there's no wantarray in P6
rjbs moritz: Hopefully. It is much less painful than the old system we replaced already, so there is not as much pressure. 17:01
TimToady what's really returned is a parcel, which hasn't committed to list/scalar context yet
rjbs TimToady++ # no wantarray
17:02 cbk1090 joined
TimToady that being said, if a return type is declared, it can supply context sooner than that 17:02
such as moritz++'s suggested List return 17:03
cbk1090 TimToady, do you have a bit of time to help me with you XML module....? or maybe can point me to some example code. 17:04
s/you/your
TimToady goodness, that's ancient history
cbk1090 TimToady, sorry I like it 17:05
is there something better?
17:05 daxim left
TimToady well, depends on the application 17:05
cbk1090 I'm using it for my database program I'm making
TimToady if you need to manage the whole tree, then my old module is okay; there are others like XML::Twig that just look for various subtrees to process 17:06
17:06 ajr_ left
timotimo S02:4590 says the user may want to write a list of numbers with angle brackets (and i agree that it's a nice way to do it), but then S02:4595 speaks about using + or ~ as the prefix to force interpretation as a number. does that mean the user should +<< < 1 2 3 4 > if they want a list of numbers rather than strings in this case? 17:07
synopsebot Link: perlcabal.org/syn/S02.html#line_4590
TimToady would have to reread the docs to remember all the APIs though...
timotimo (the links were off by one paragraph >_>)
17:07 ajr joined, ajr is now known as Guest51379
cbk1090 I want to make a line like this: <record id="1" date="10/30/2013"><topic>some topic</topic><remarks>some text here</remarks></record> 17:07
TimToady timotimo: yes, a hyper would be appropriate for that 17:08
17:08 Guest51379 is now known as ajr_, rindolf joined
timotimo installs chrome.google.com/webstore/detail/...e-ntp-icon to better use the synopses 17:08
17:09 kaleem left
timotimo huh. that doesn't seem to work at all 17:09
TimToady cbk1090: unfortunately, I switched to using YAML more than a decade ago for most such things, and don't remember much (or actively forgot it :)
timotimo hum. i don't think hypers can be compile-time evaluated yet 17:11
TimToady well, they can be, but aren't
cbk1090 TimToady, ok no prob. I'll keep plugging away at it. Thanks for providing it as a Module. :-)
TimToady (Tim Bray)++ put me up to it 17:12
[Coke] afternoon, sixers.
TimToady interestingly, doing XML and seeing the impedance mismatch between C structs and Perl hashes is part of what clued me into the need for representational polymorphism in P6 17:13
cbk1090 oh, well i just need it for a text based database. 17:14
TimToady it's probably okay for that, if you don't push it on performance
moritz recommends base64-dumps of sqlite for text based databases
cbk1090 nothing fancy just a place where i can store my passwords and keep track of fuel cost, ext... 17:15
17:15 sftp_ joined
cbk1090 also planning on using same database as a way to publish a simple blog. 17:15
17:15 ajr_ left
cbk1090 <topic>blog</blog><remarks>this would be like a blog entry record.</remarks> 17:16
17:16 sftp left
cbk1090 <topic>blog</topic><remarks>this would be like a blog entry record.</remarks> 17:16
thats right
TimToady plugging in the expat parser to a suitable repr might be a good show of how polyrep is a win in p6 17:18
if someone is looking for a small project :) 17:19
arnsholt And it'll give us an XML parser, without having to wrap LibXML, which is kind of big 17:20
masak +1
17:22 tobyink left 17:23 zakharyas left
jaffa4 when are you going to have Rakudo Star 10 for WIndows? 17:26
17:26 cbk1090 left 17:27 cbk1090 joined
masak whenever someone rolls an .msi, I guess. 17:27
moritz: did you know the trick described in this post? blog.plover.com/prog/intervals.html -- I didn't.
PerlJam "trick"? 17:28
weird. I just thought that's how you do it. (and that this was the first time MJD had run across the problem)
17:28 sftp joined 17:29 cognominal left, sftp_ left
masak PerlJam: if I hadn't read that post, I would have implemented it like mjd did the first time around (but with all three cases, including the smaller-in-larger case) 17:31
FROGGS jaffa4: that is on my list for tonight/tomorrow 17:32
jaffa4 FROGGS: I see, how are you going with perl5 on rakudo? 17:33
FROGGS jaffa4: stalled a bit atm, because spectesting takes about 50 minutes... my hopes go into moarvm backend and therefore shorter test cycles 17:34
TimToady well, you could fix things one test at a time, and then check for regressions when you go to bed :) 17:36
FROGGS jaffa4: and when jnthn++ has some time for me, there are some minor issues that need to be fixed in order to make it a proper module 17:37
jaffa4 FROGGS: what is its state anyway? 17:38
FROGGS TimToady: the problem is that when I fix/implement one thing, the regressions pop up in very strange places... so I have to run a spectest almost every time
jaffa4: it passes 4407 of around 40000 tests, so, it is already somewhat usable :o) 17:39
jaffa4 FROGGS: cannot you run them in parallel?
FROGGS one major problem is that it is missing the Perl5 ore modules
core*
jaffa4: I could, but this breaks some tests (just two or three) 17:40
TimToady it also kinda sounds like you've started to hit the point where the spec tests are forcing you to implement some of the insanity the same way p5 does it :)
FROGGS using four cores it takes about 20 minutes
ingy TimToady: ♥
FROGGS TimToady: exactly
jaffa4 How about distributed testing? 17:41
FROGGS did I mention that I *hate* indirect object syntax?
jaffa4: I thought about it, but that's all
diakopter FROGGS: just lookahead a few.. :)
ingy TimToady: YAML*.pm is under heavy dev right now, so if you have any 10 years frustrations lemme know 17:42
TimToady mostly having to guess about Unicode, and usually guessing wrong 17:43
17:51 jeffreykegler joined
dalek p: d7151b3 | (Tobias Leich)++ | / (2 files):
mapped nqp::getpid
 17:51
17:52 tobyink joined 17:55 SamuraiJack_ left 17:56 cbk1090 left 17:58 darutoko left
timotimo TimToady: do you have something to say about how negative values in a Mix should influence the selection of elements from the Mix? 17:58
TimToady erroneous, maybe 17:59
only Mixy[Unum] supports .roll ? 18:01
timotimo well, the current way it works is b0rked indeed. there's a spectest that seems to suggest negative means "will not be picked, ever" 18:02
Mixy is currently non-parametric, though 18:04
18:04 dakkar left 18:11 tobyink left
TimToady not that Unum exists either... 18:13
timotimo right. should .roll die if there's a negative number in the pairs? should there be a test case to check that it doesn't live? 18:14
dalek ast: aa0f747 | jonathan++ | S17-concurrency/promise.t:
Some initial tests for Promise.
 18:15
timotimo questhub.io/realm/perl/quest/52714c...c02600004d - you can cheer me on if you want :P (implement eqat and get it into master) 18:16
(hum, maybe dalek could report quests being created with perl6 as the tag in the perl realm? would that be interesting?) 18:17
masak timotimo: we could try it. maybe as a one-liner. 18:19
also, "go go go timotimo! \o/" :) 18:20
timotimo: something like <dalek> questhub: nqp: implement and use eqat op on all backends -- questhub.io/realm/perl/quest/52714c...c02600004d
timotimo the dalek information is in Mu, right? 18:21
how to set it up and stuff?
moritz speaking of which, the code generation for scanning regexes on the JVM still doesn't use index 18:22
18:22 tgt left
moritz timotimo: I don't think so 18:23
18:24 Tom_ joined
moritz timotimo: I know it involves git://github.com/Infinoid/dalek-plugins.git somehow 18:25
18:26 dansamo left
timotimo i was refering to the documentation of the webhooks thing we've got for the commits 18:27
something similar should be doable for questhub, no?
retupmoca timotimo: are you talking about this: github.com/perl6/mu/blob/master/mi...k-push.txt 18:28
ooh, nqp commit-bit 18:32
it must only be rakudo that requires the paper signed? 18:33
jnthn retupmoca: Correct 18:35
rjbs wonders when we'll be able to port Marc L.'s modules to perl 6.
lue hello world o/ 18:40
18:41 FROGGS left, FROGGS joined 18:42 wbill left 18:46 wbill joined
TimToady doesn't classify modules by who wrote them, which is probably an autistic failure mode :) 18:47
TimToady doesn't much track actors names either: "the guy who played Boromir in the LotR" 18:49
rjbs Normally, I have no reason to classify them as such. It's only when the person who wrote them has a personality that overshadows the code... 18:50
TimToady doesn't know anyone like that :) 18:51
rjbs Best line from the Washignton Post in a while: "Two engineers with close ties to Google exploded in profanity" 18:54
TimToady I'd probably have heard any explosions from here, given how close I live to Google... 18:55
lue how are the Profanity city PD dealing with it? 18:56
retupmoca PCPD? 18:58
rjbs I hate that fucking place.
TimToady utters a string of confanity
rjbs Unicode 7 needs more grawlix.
I think it's crazy that we have no "PLANET SATURN."
TimToady .u ♄ 18:59
FROGGS ENOYOLEAUX
rjbs TimToady: Hey, buddy, you show me an Andy Capp where he said that while fighting and I'll drop my complaint.
lue
.oO(I'm telling you, say "⚂".char-name would solve our problems.)
TimToady which reminds me, we need a consistent specced way to look up unicode props 19:00
geekosaur preflex: seen yoleaux 19:01
preflex yoleaux was last seen on #perl6 4 days, 23 hours, 39 minutes and 11 seconds ago, saying: 18:23Z <diakopter> Mouq: awesome work on stdsigspace Mouq++
geekosaur :(
lue
.oO(Mouq's gotten a lot of accidental karma from people seen-ing yoleaux :P)
rjbs TimToady: Have you discussed/tackled the problem, yet, of properties on trans-Unicode codepoints?
19:02 dmol left
TimToady we're still having enough trouble with the cis-Unicode codepoints 19:02
rjbs We just recently had a discussion about it. It'll definitely be something to consider. 19:03
19:04 dmol joined
lue TimToady: my first thought is either .uni-info(:propname, :prop2) or a Codepoint object that has properties as methods. "ABCD".comb».Codepoint».cclass 19:04
TimToady we don't necessarily need another char type where integers will do 19:06
though methods do have their notational convenience, I'll admit 19:07
lue I was thinking the Codepoint object stores the unicode information (maybe even to has $.cclass visibility), but sub/methods on an integer (such as 0x2A03) would do just as fine.
TimToady methods might work better where hypers are wanted 19:09
lue almost doesn't like the idea of a bunch of methods (thus the Codepoint object to put them in one place), but reminds himself that the adverb route has just as many name definitions. 19:11
19:13 mathw_ left 19:23 jeffreykegler left 19:28 fhelmberger joined 19:37 fhelmberger left
moritz masak: qandwhat.apps.runkite.com/i-failed-...interview/ 2D version of t4 asked in an interview :-) 19:41
masak nice :) 19:42
skimming it, I don't think that solution generalizes to t4's 3D at all. 19:44
still very interesting, of course :)
19:47 fhelmberger joined 19:49 fhelmberger left
moritz agreed on both points 19:49
masak today's mini-challenge: write a nice p6 solution that passes all of the test cases at gist.github.com/igor47/7228586 19:50
I won't have time for it myself, but I wish I did.
19:51 sqirrel joined 19:54 rindolf left, fhelmberger joined 19:55 fhelmberger left
japhb__ jnthn: In designing a multi-level API (low-level operations, medium-level wrappers, high-level convenience), how do I structure the code so that I can make it as efficient as possible? Right now I'm assuming all the levels (except maybe the very top) should be in the same file/comp_unit, written as modules with exported subs, using last-expression-value returns rather than return statements. 20:01
Is that correct so far? Do I still need to add 'is pure' for purely functional routines, or can that be intuited if all the ops within are pure? 20:02
masak japhb__: how important is performance?
japhb__ masak: As fast as Rakudo can possibly go.
20:02 tobyink joined
lue masak: please tell me this t4 doesn't have a "number of solvable boards" question to go along with it :) 20:03
japhb__ masak: Let's just say hand-coded C++ versions of this code are a performance bottleneck
masak lue: nope :) 20:04
japhb__: then (my point is) it sounds like you have a vested interest to measure decisions like the above.
lue: in fact, t4 is much less of a "hah! gotcha!" task this year than the previous two years. 20:05
lue: but still pretty subtle.
japhb__ masak: I'm trying to prune the decision space. If jnthn knows from his optimizer work that a certain decision disables an entire class of optimizations, that prunes out a big chunk of a combinatorial space that would take me quite a while to explore if unpruned.
moritz japhb__: note that *only* constant folding considers 'is pure' right now 20:06
japhb__ What is necessary to do or avoid for inlining to work?
masak japhb__: ooc, are we talking Rakudo on the JVM? 20:08
if we are, then you're likely to get a lot of JIT for free.
japhb__ masak: I'll take JVM. I'd love to be prepared for Moar. (Meaning, not have to redesign the API to deal with a different set of requirements for optimization.) 20:09
.oO( Write Once, Suck Less Everywhere )
20:10 lowpro30 joined
arnsholt Inlining and other optimiser decisions sounds like a thing to add a pragma for 20:11
20:12 lizmat joined
japhb__ arnsholt: How does that help? I can certainly see a case for optionally turning them *off*, but generally you want the equivalent of -O2 to be the default, neh? 20:13
20:14 woolfy joined
arnsholt Not turning off, but tracing 20:16
I was a bit unclear about that
But logging the decisions made, so that it actually tells you what gets inlined and what doesn't
So that people don't have to guess about this kind of thing =) 20:17
20:17 [particle] left
FROGGS jaffa4: you wanna test? froggs.de/perl6/rakudo-star-2013.10-RC0.msi 20:17
jaffa4 FROGGS:ok
FROGGS jaffa4: but rename your C:\rakudo before installing it 20:18
jaffa4 FROGGS: it does not let me choose directory 20:19
moritz because parrot isn't relocatable
japhb__ arnsholt: Oh yes, that would be very helpful. 20:20
20:21 denis_boyun__ joined, denis_boyun left 20:22 bluescreen10 left
jaffa4 Failed to create directory 'blib/lib\Path/' with mode '0o777': mkdir failed: The system cannot find the path specified. 20:22
I tried to use panda 20:23
lue r: say $_ for 5^...1; # :(
camelia ( no output )
FROGGS ==> Successfully installed NativeCall 20:24
moritz \o/ # empty range is empty
FROGGS jaffa4: it works on my box
jaffa4 FROGGS: I managed to nstall it but panda does not seem to work
good for you?
20:24 bluescreen10 joined
lue any idea why ^... doesn't work? 20:25
FROGGS Failed to remove the directory '.work\1383164610_1\lib': rmdir failed: The directory is not empty. 20:26
-.-
20:26 fhelmberger joined
moritz lue: because the range is empty 20:26
lue but...
moritz oh, it's a series
that's not good
20:26 stevan_ left
moritz r: .say for 5...1 20:27
camelia rakudo-parrot d9e79f, rakudo-jvm 882e33: OUTPUT«5␤4␤3␤2␤1␤»
lue It's kinda critical for a cool-ish looking version of the Python gist masak mentioned :)
FROGGS lue: btw, your 2013.10 tar.gz lacks modules/file-find and modules/shell-command
jaffa4 panda install Text::Markov
Failed to create directory 'blib/lib\Text/' with mode '0o777': mkdir failed: The system cannot find the path specified.
masak waitwait, is ^... even an operator? 20:28
FROGGS C:\rakudo>panda install Text::Markov
==> Fetching Text::Markov
==> Building Text::Markov
masak I know ...^ is, but...
FROGGS but I am on WinXP currently
masak std: 5^...1
camelia std 8adbc60: OUTPUT«ok 00:01 122m␤»
FROGGS jaffa4: now I get the same error message like you
lue FROGGS: I have those two directories, so the makefile must've missed them :(
masak I'm wondering if that doesn't parse as 5 ^.. .1 20:29
std: 5 ^... 1
FROGGS lue: can you zip those two for me? I wonder if the mkdir errors come from outdated shell-command
camelia std 8adbc60: OUTPUT«===SORRY!===␤Bogus term at /tmp/0vx8OTR9RB line 1:␤------> 5 ^..⏏. 1␤Parse failed␤FAILED 00:01 122m␤»
masak lue: see? bogus.
lue: serves you right for not using whitespace around your infix operators :) 20:30
lue masak: either give me a ^... or a ^.. that can do the apparently arcane magic of descending numbers.
masak: I already tried whitespace and knew of the concat error.
FROGGS: where would I upload the tarball?
moritz r: .say for [5...1][1..*]
FROGGS to a place of your choice
camelia rakudo-parrot d9e79f, rakudo-jvm 882e33: OUTPUT«4␤3␤2␤1␤» 20:31
timotimo what's the current procedure to generate new bootstrap for parrot and jvm?
moritz lue: you can always use that :-)
FROGGS lue: or mail it to me :o)
moritz timotimo: aren't there make targets for that?
lue FROGGS: I don't have a good place to upload, so if you privmsg your email, that'd be good.
timotimo are there two now?
FROGGS timotimo: there are three
make <[pmj]>-bootstrap-files 20:32
timotimo excellent
moarvm already has eqat_s, so no need to regenerate the bootstrap files there
20:33 woolfy left, woolfy joined 20:34 Rotwang joined
moritz why does that even need a bootstrap update? 20:34
masak lue: re "apparently arcane magic of descending numbers" -- you apparently have never felt the pain of a range that doesn't become empty when the endpoints are in the wrong order.
lue: just do 4...1 and be done with it.
moritz the old bootstrap files that don't use eqat_s should still work, no? 20:35
lue masak: that means an ugly (*ugly*) +@elems-1 that I really don't think I should have to do :/ (moritz's solutions seems good though)
FROGGS you only need to update bootstrap-files if you want to use eqat_s in nqp 20:36
masak lue: I like to write reverse ranges like this: `reverse 1 ..^ 5`
lue: very readable -- the word 'reverse' is even in there!
20:36 sqirrel left
masak documentation as code. 20:37
FROGGS For 1 To 5 Do ' documentation as code
lue I kinda wish 1 R..^ 5 or R(1..^5) or [R] 1..^5 DWIMmed. (If you can't tell, I'm trying to prefer cooler-looking constructs here! :P) 20:38
masak lue: that's not what R means. 20:39
lue I know :) (then why do I remember a use of R that flipped a list...?) 20:40
timotimo moritz: yes, they should, but i'd like to use eqat in the nqp source code :)
masak but there's nothing to stop you from defining either of those crazy operators, I guess.
LTM will do the rest.
timotimo (actually not called eqat_s)
lue I'll workaround of course, but I'm still miffed that, for example, 5..1 doesn't work. I've never been able to understand that one, actually. (I'm also now miffed by the curious non-existence of ^... but that's another thing) 20:41
masak lue: until you understand that one, take it from people who have either felt the pain directly, or been able to extrapolate it from circumstantial evidence :) 20:42
20:42 cognominal joined
masak lue: you'll have to take up ^... with TimToady. I'm sure there's a very good reason. 20:43
timotimo lue: last time i tried to understand, it seemed like ".. is not supposed to be smart enough for that"
... will do it for you, though
masak indeed.
20:43 fhelmberger left
lue Until I understand it, I will always say "die on 5..1" it's a bogus limitation :) 20:43
timotimo :)
lue s/it's/is/
moritz or you can just say 5 - 1 ... 1
masak yeah, infix:<..> means "range *counting up*". note how we also ditched the :by(2) idea, sticking all of the magic in infix:<...> 20:44
lue moritz: but -1 is for languages that don't have ^..^ ! :)
masak this is a *good thing*, and if you disagree, do try to keep your complainin' to a minimum. :)
std: 5..1 20:45
camelia std 8adbc60: OUTPUT«ok 00:01 122m␤»
masak could do with a compile-time warning on that one, I guess.
japhb__ r: say [R,] 1..^5
camelia rakudo-parrot d9e79f: OUTPUT«Too many positional parameters passed; got 4 but expected 2␤ in block at gen/parrot/CORE.setting:16422␤ in sub at gen/parrot/CORE.setting:16529␤ in block at /tmp/MZz9SSn2al:1␤ in any at /tmp/MZz9SSn2al:1␤ in any at gen/parrot/stage2/N…»
..rakudo-jvm 882e33: OUTPUT«Too many positional parameters passed; got 4 but expected 2␤ in block at /tmp/o_031j9iTU: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…»
japhb__ awww
lue Well, at least I see how (apparently) descending simple lists is bad for .. , but I'll talk to TimToady tonight (if he'll be around) about ^...
r: .say for 1...5
camelia rakudo-parrot d9e79f, rakudo-jvm 882e33: OUTPUT«1␤2␤3␤4␤5␤»
lue r: .say for 1...^5
camelia rakudo-parrot d9e79f, rakudo-jvm 882e33: OUTPUT«1␤2␤3␤4␤»
lue r: .say for 1^...^5
camelia rakudo-jvm 882e33: OUTPUT«===SORRY!=== Error while compiling /tmp/ixENlC94xY␤Confused␤at /tmp/ixENlC94xY:1␤------> .say for 1^...^⏏5␤ expecting any of:␤ dotty method or postfix␤»
..rakudo-parrot d9e79f: OUTPUT«===SORRY!=== Error while compiling /tmp/SwcOzA_Bua␤Confused␤at /tmp/SwcOzA_Bua:1␤------> .say for 1^...^⏏5␤ expecting any of:␤ dotty method or postfix␤»
masak lue: ISTR mjd has a good comment about it in a blog post somewhere. 20:46
lue: but basically it's about programmer expectations. 20:47
timotimo i wonder how much faster nqp::eqat is than nqp::index(...) == 0
masak lue: when a coder writes 1..$end in their code, she's *thinking* "from 1 *up to* $end"
timotimo for the cases it's used in, it probably doesn't change performance much 20:48
but it still seems cleaner to me :P
masak lue: she's not thinking "from 1 up to $end *or down to it if $end < 1*"
lue: so it's a kind of conservative application of the DWIM principle, actually.
lue would like to think he'll get bit by it someday, but it won't happen in Perl 6 ;)
masak or, more precisely, it's the Principle of Least Surprise. 20:49
as in the DWIM you're proposing is not worth the WAT it introduces.
lue What doesn't help is that I see .. and I think "generate a sequence of consecutive integers from LHS to RHS"
FROGGS jaffa4 / lue: that didn't help :(
masak lue: this is why language design is interesting. your idea seems reasonable *on the surface*, but applied in an actual programming language, it's crap. 20:50
lue (and ... is for magical lists, like evens or %% 3)
masak lue: right.
lue: TimToady happens to be really, really good at anticipating the way features can come back and bite users. or the way features will interact. 20:51
20:51 ssutch joined
masak (and still, Perl 6 has changed quite a lot over the years due to things none of us could predict.) 20:51
lue r: sub infix:<.>($lhs, $rhs) { try { return $lhs...$rhs } }; say $_ for 1.5; say $_ for 5.1; # :D
camelia rakudo-parrot d9e79f, rakudo-jvm 882e33: OUTPUT«1.5␤5.1␤» 20:52
lue *headdesk*
timotimo ERROR: Unknown flag --version. USAGE: moar / just an outdated moarvm?
lue *headdesk* x2
masak lue: go back to square "how does term parsing work".
lue *headdesk* x3
r: sub infix:<.>($lhs, $rhs) { try { return $lhs...$rhs } }; say $_ for 1 . 5; say $_ for 5 . 1; # :D
camelia rakudo-jvm 882e33: OUTPUT«(timeout)»
..rakudo-parrot d9e79f: OUTPUT«===SORRY!=== Error while compiling /tmp/2jO09VzhKa␤Unsupported use of . to concatenate strings; in Perl 6 please use ~␤at /tmp/2jO09VzhKa:1␤------> return $lhs...$rhs } }; say $_ for 1 . ⏏5; say $_ for 5 . 1; # :…»
masak :) 20:53
lue (that try { } is a leftover)
Bad Perl6!
masak well, obviously you're trying to concatenate strings, you bad p5 user.
lue r: supersede sub infix:<.>($lhs, $rhs) { try { return $lhs...$rhs } }; say $_ for 1 . 5; say $_ for 5 . 1; # :D
camelia rakudo-jvm 882e33: OUTPUT«Unhandled exception: java.lang.OutOfMemoryError: Java heap space␤ in (src/Perl6/Grammar.nqp)␤ in (src/vm/jvm/ModuleLoader.nqp:62)␤ in (src/vm/jvm/ModuleLoader.nqp:52)␤ in load_module␤ in ␤␤» 20:54
..rakudo-parrot d9e79f: OUTPUT«===SORRY!=== Error while compiling /tmp/aNkEHGiyHP␤Cannot use 'supersede' with sub declaration␤at /tmp/aNkEHGiyHP:1␤------> hs, $rhs) { try { return $lhs...$rhs } }⏏; say $_ for 1 . 5; say $_ for 5 . 1; # ␤ …»
FROGGS timotimo: we have --version since 2013.10
timotimo makes sense
masak lue: clearly tonight you are trying to be the unreasonable man on which progress depends. 20:55
lue I'm sure TimToady will enlighten as to why my DWIMs and WATs on .. are inverted tonight :) 20:56
masak I thought I explained it already. :)
Tene masak: I fail to imagine a case where I'd expect to rely on a range being zero when the end is lower than the start. 20:57
Any handy intuition pumps?
jnthn japhb__: for performance: native int over Int is usually a win as you'll get operators inlined as well as the ops just being cheaper. For inlining, subs and multi-subs that are compile-time determinable will cut it, though they need to be simple. Inlining will work between compilation units; has to or we could never inline things from the setting. If creating lots of objects, write a custom BUILD rather than relying on the default thing.
20:57 BenGoldberg joined
masak Tene: I'll try to think of a real-enough-world case. 20:58
lue masak: you did, but I still can't imagine a specific case where I'd go "gee, sure is handy .. can't create a list of descending integers"
20:58 woolfy left
jnthn japhb__: Also try --optimize=3 but I know that it does at least one thing wrong that breaks spectests, even if compiling CORE.setting with it works out. 20:58
20:58 lizmat left
masak lue: the feature is not that can't deliberately create them, the feature is that you can't *accidentally* create them. 20:59
Tene lue: you're not completely anomalous; I'm with you so far.
lue 5 .. 1 :yes-i-bloody-mean-it then :)
Tene :yes-i-bloody-mean-it is perhaps spelled . ;) 21:00
lue 5 .. 1 . ?
:) 21:01
21:02 lizmat joined
lue (As a reminder, this discussion wouldn't've happened if ^... was a thing. It led to me wanting .. (and by extension ^..) to do what ... does and ^... should be doing) 21:02
masak lue: what list do you expect '10, 9, 8 ^... 5' to result in? 21:03
lue The same one ... generates, but without the 10. Clearly :P . 21:04
jaffa4 FROGGS: what will happen?
lue resorted to a ..^ list and then reversing the populated array, in any case.
timotimo i can kind of see wanting to leave out the starting value for some kinds of sequences 21:05
but i can't think of an example
Util masak: I gave some thought to that last week; 10, 9, 8 ^... 5 should produce 9,8,7,6,5; the 10 is just extra info fed to the sequence predictor routine.
masak ok, so we should have the full set of ... ...^ ^... ^...^ ? 21:06
Util Yes. Least suprise after one learns of .. ^.. ..^ ^..^
masak Util: what about 1, 2, 3 ... 10, 20, 30 ^... 100 ? 21:07
FROGGS jaffa4: well, one (like me) has to debug it
Util We have ^... now
masak no, we don't.
Util I mean ...^ - typo
lue masak: I want to say 1, 2, 3 ... 10, 20, 30 ^... 100 leaves out the 1 . But if ... does that kind of magic, we almost should have another operator that lets me do exclude-first descending numbers, without all the debug hell adding ... variants apparently would cause ;D 21:08
21:08 benabik left
masak r: .say for 1, 2, 3 ... 10, 20, 30 ... 100 21:09
21:09 benabik joined
camelia rakudo-jvm 882e33: OUTPUT«Unhandled exception: java.lang.OutOfMemoryError: Java heap space␤ in (src/Perl6/Grammar.nqp)␤ in (src/vm/jvm/ModuleLoader.nqp:62)␤ in (src/vm/jvm/ModuleLoader.nqp:52)␤ in load_module␤ in ␤␤» 21:09
..rakudo-parrot d9e79f: OUTPUT«Cannot call 'infix:<...>'; none of these signatures match:␤:($a, Mu $b)␤ in any at gen/parrot/BOOTSTRAP.nqp:1218␤ in any at gen/parrot/BOOTSTRAP.nqp:1209␤ in sub infix:<...> at gen/parrot/CORE.setting:16290␤ in block at /tmp/jMWIUYqdbt:1…»
masak huh.
masak thought that was straight out of S03
lue n: .say for 1, 2, 3 ... 10, 20, 30 ... 100 21:10
camelia niecza v24-98-g473bd20: OUTPUT«1␤2␤3␤4␤5␤6␤7␤8␤9␤10␤20␤30␤40␤50␤60␤70␤80␤90␤100␤»
lue masak: niecza++
jnthn masak: Precedence, maybe?
masak yeah.
jnthn: but niecza gets it right. 21:11
jnthn yes, that's a little confusing :)
masak submits rakudobug
21:11 benabik left
lue my attempt: gist.github.com/lue/7240356 21:12
Util My full thought from last week was that the ... design is broken w.r.t. huffmanizing. When the sequence back-end derives a code block of { $_ + some_constant } or {$_ * some_constant} then it knows that the sequence is heading toward +Inf or -Inf. In that (most common) circumstance, 1, 3, 5 ... 12 should either stop at 11, or at 13, or warn, or die; *anything* but infinitely loop. 21:13
21:14 woolfy joined
Util if someone writes 1, 3, 5 ... 12 , then they *never* mean 1, 3, 5 ... * 21:14
Landmine
21:14 jaffa4 left
arnsholt Oooh, looks like I've found a way to do CStr on JVM 21:15
timotimo Util: i disagree. we should not just say "oh they probably mean * > 12"
FROGGS arnsholt: that sounds good, ehh?
arnsholt Yeah, it's about time I do some more hacking on the NativeCall stuff 21:16
Even though CStr is a part of the API I really can't wait to kill =)
FROGGS arnsholt: yeah, the fans wanna have it :o)
masak Util: I'm trying very hard not to trigger my "halting problem" reflex. of course we could catch the simple cases, yes. 21:17
FROGGS (NativeCall, not CStr)
lue timotimo: but "they probably mean *" is also wrong, and worse IMO
I would go with a die in that case, because at that point the programmer can't sequences right.
timotimo hm. if we have one of the simpler sequences, we can probably just analytically see if the endpoint is in the sequence and carp if not 21:18
Util masak: the simple cases will be the most common, I am sure.
FROGGS well, you could eject the cd drive very hard and hope it hurts him/her
masak Util: agreed.
timotimo but only in the case where we have only numbers involved
masak .oO( Warning: the sequence is infinite, due to Fermat's Last Theorem ) 21:19
FROGGS and again, X::Insulting would help
lue timotimo: die if $cur-gen-number > $endpoint 21:20
Util Sometimes we want to code something like rosettacode.org/wiki/Hailstone_sequence , which jumps around, but most will be simple (and recognizable) "progres toward +-Inf"
lue
.oO(Dying on unsatisfactory endpoints would make ... more consistent with .. anyhow.)
 21:21
lue &
timotimo Util: i already implemented the hailstone sequence with the ... operator and it was beautiful
masak Tene: in Perl 5, `for my $index (0..$max) { send_nuclear_warhead($city[$index]) }`
Tene: if 0..$max produces a non-empty range full of decreasing negative numbers, and $city[$index] indexes from the end, then... global thermonuclear war. 21:22
Util It does not even indicate critical user dumbness. 2, { $_ + $interval } ... $endpoint , where the two variables are defined much earlier, is easily vulnerable to maintanence programmer error. 21:23
masak Tene: in Perl 6, the corresponding code dies at runtime -- unless your array is of a special (unimplemented) S09 type that allows index wraparound. but the point still stands -- when a reader sees 0..$max, the "static gestalt" (or whatever) of that expression is "0 upto $max" 21:24
Util timotimo: Yes, quite beautiful, and the thing that shaped my mind *away* from the idea that all ... should be specced as ever-increasing-or-ever-decreasing for safety. I would hate to lose that Hailstone ability, but just changin the spec to say that trailing codeblock is manditory for non-simple cases gives us the best of both worlds. 21:25
timotimo ... is not only for numbers 21:26
traverse a graph by following ".parent" pointers, for example
Util timotimo: Good point.
timotimo my @path = ($end, *.parent ... $beginning).reverse
Util FYI: p6g -e '.say for 1, 3, 5, 7 ... 12;' freezes with *no* output for me. 21:27
masak Util: that's a problem with for loops, though.
timotimo right. use map instead and it will work
Util masak: noted. Thanks! 21:28
lue proposes the en dash (–) for index ranges and .. for what he wants :P
timotimo hah, with -j3 it seems the makefile will do all the steps like /usr/bin/perl -MExtUtils::Command -e cp gen/jvm/stage2/nqpmo.jar . 21:30
three times in a row
<spam> 21:32
dalek p/eqat_op_rebased: 0b129b2 | (Timo Paulssen)++ | src/vm/ (4 files):
add nqp::eqat to jvm and parrot.
nqp/eqat_op_rebased: de1e094 | (Timo Paulssen)++ | src/vm/parrot/ops/nqp.ops:
nqp/eqat_op_rebased: special case same-encoding strings, length-1 needles.
timotimo </spam>
... didn't intend for it to be kicked :(
21:33 dalek joined, ChanServ sets mode: +v dalek
timotimo i'd like to merge that into master now, i consider it done. 21:33
am i missing something obvious? like, should i create tests?
japhb__ Tene: I'm backlogged, so not sure if anyone has mentioned this, but you want reversed ranges to become empty because of algorithms that involve pointer or index crossing.
masak japhb__: could you elaborate "pointer or index crossing"? 21:34
anyway, it sounds like my nuclear example.
jnthn japhb__: I'm robustering (hopefully) the concurrency stuff these days. Also adding tests. :) 21:35
japhb__: Promise is getting a fairly big refactor but it's largely "on the inside" so I'm hopefully it'll not regress anything you're doing. 21:36
21:36 Lorn joined, Lorn left
japhb__ masak: Any number of algorithms (of which quicksort is probably the most well known), that involve starting at opposite ends of an array or subarray, and moving the ends towards each other until the "right" one is to the left of the "left" one. 21:36
jnthn Binary search is an even simpler example than quicksort, I guess... 21:37
japhb__ jnthn: Very happy to hear that inlining works across comp units. That makes for way less ugly for me.
jnthn japhb__: I have dreams of making the optimizer have a mode where it explains what it did/didn't do to help us better understand these things.
japhb__ jnthn: Once you head to bed, I'll recompile rakudo-jvm and see if you broke my code. ;-)
jnthn OK. The earth-shattering Promise refactor may or may not land tonight. :) 21:38
japhb__ jnthn: Yeah, arnsholt mentioned that idea earlier, and I'm all in favor.
I'm using Promise pretty close to the way you used it in your talk.
jnthn I'm in favor, I'm just under-provisioned and over-assigned ;)
japhb__: Are you mostly creating them with async { ... } 21:39
japhb__: Are using Promise.new(...) explicitly?
I think my talk uses were all async { ... }
21:39 [particle] joined
jnthn s:2nd/Are/Or/ 21:39
japhb__ My very first use was 'sub async_qx($cmd,...)' which runs a qx in a separate thread, returning promises. And then a program which needs to shell out to a lot of slow things turned into a sub make_promises() which returned a hash with all values being async_qx() calls ... and then just pulling out the results in the order I wanted. 21:40
jnthn The main thing I'm doing is detangling the synchronization primtive nature of Promise from specific uses of them. 21:41
And also moving timing related stuff into the scheduler where it belongs
21:41 woolfy left, woolfy joined 21:42 mathw joined
masak japhb__: oh, lovely example. yes, my visual intuition is of little gauge arrows marking the ends of the range, and when they pass each other, the range blinks out of existence. 21:42
jnthn The main thing that's changed is Promise.new(:code({ ... })) is gonna be gone.
japhb__ jnthn: I'm using the idiom: sub async_qx { ... my $p = Promise.new; ... async { ... $p.keep($result) ... }; $p }
jnthn japhb__: Oh, that's fine.
lizmat wonders whether we wouldn't need some syntactic sugar for Promise.new 21:43
jnthn There will also be a way to get a "keeper" object from a Promise that lets you encapsulate the right to keep/break it, but it's optional
21:43 wbill left 21:44 Timbus left
lizmat
.oO( cold induced ramblings from the lobby of a hotel, somewhere in Germany )
 21:44
jnthn lizmat: Maybe... :)
lizmat: promise() # :P
21:45 Timbus joined
[Coke] timotimo: please document the eqat op. 21:45
oh, I checked in the wrong branch. 21:46
lizmat jnthn: was more thinking something like <> 21:47
timotimo the documentation is already there in that branch :)
(you documented it, even)
[Coke] timotimo++
lue
.oO(If you used UInt ranges, you wouldn't need to worry about negative indices.)
[Coke] me-- 21:48
japhb__ jnthn++ # Iterating an already pretty decent API because it wasn't good enough.
[Coke] lizmat: hio
lizmat [Coke] o/
timotimo huh, something seems to give build trouble in rakudo/eqat_op (rebased)
lizmat
.oO( there's nothing like PDD )
lue masak: I'll talk with TimToady later today to get some insight on *why* descending range is designly bad (like you mentioned earlier), but until then I'll disagree with your (and most others') definition of the .. operator :) 21:49
dalek ecs: 3150289 | (Elizabeth Mattijsen)++ | S99-glossary.pod:
Add PDD
 21:50
timotimo Error while compiling op bind: Method 'type' not found for invocant of class 'NQPMu' - that error doesn't ring a bell to me
BenGoldberg rn: .say for 1, 3, 5, 7 ... * > 12 21:51
camelia rakudo-parrot d9e79f, niecza v24-98-g473bd20: OUTPUT«1␤3␤5␤7␤9␤11␤13␤»
..rakudo-jvm 882e33: OUTPUT«Unhandled exception: java.lang.OutOfMemoryError: Java heap space␤ in (src/Perl6/Grammar.nqp)␤ in (src/vm/jvm/ModuleLoader.nqp:62)␤ in (src/vm/jvm/ModuleLoader.nqp:52)␤ in load_module␤ in ␤␤»
japhb__ lue: Did you disagree with my example of ranges used to implement index-crossing algorithms?
Gah, somebody hand that JVM some heap space. 21:52
lizmat
.oO( heap heap hurray )
lue Not really, but I don't think .. should be considered $rhs <= $lhs checker. Again, it's because I see .. as a "consecutive integer list generator", and nothing more. 21:53
BenGoldberg
.oO( heap... heap... heeaaaachooo! )
masak lue: and you do that because you've never shot yourself in the foot with your proposed semantics.
21:53 Tom_ left, wbill_ joined
masak lue: and you haven't done that because Perl (and Python, and Ruby...) chose the sensible default. 21:54
lue [ I certainly don't see it as just the "array range operator". – would evoke those sentiments from me. Not to mention, I think an array range operator should allow reverse traversal :) ]
BenGoldberg r: say 13, 11, 9 ... 1
camelia rakudo-jvm 882e33: OUTPUT«Unhandled exception: java.lang.OutOfMemoryError: Java heap space␤ in (src/Perl6/Grammar.nqp)␤ in (src/vm/jvm/ModuleLoader.nqp:62)␤ in (src/vm/jvm/ModuleLoader.nqp:52)␤ in load_module␤ in ␤␤» 21:55
..rakudo-parrot d9e79f: OUTPUT«13 11 9 7 5 3 1␤»
21:55 woolfy left
lue masak: like I said, I won't ever find out how bad it is in Perl 6 because you won't let me :) . I'll have to turn to C++ for that stuff. (I could use ..., but it doesn't have the full set of variants .. offers, which in this case I needed one of.) 21:55
21:55 lowpro30 left
timotimo why does it err out in the *grammar*?! 21:55
FROGGS timotimo: what if $needle is null here? github.com/perl6/nqp/commit/d8b49d...b29f1R2000
then nqp::istype might fail, right?
timotimo oh? 21:56
diakopter well, that's fine
timotimo i haven't thought of that
diakopter nulls are ok
21:56 xenoterracide left
FROGGS nqp-p: my $n := nqp::null(); say(nqp::istype($n, QAST::Block)) 21:57
camelia nqp-parrot: OUTPUT«0␤»
lue masak: And I semi-agree with the idea that not all of the ..^ ^.. ^..^ forms make sense in ... . Which is why it's infuriating there's no Perl6y way to generate an exclude-first descending list of consecutive integers.
FROGGS nqp-p: my $n := nqp::null_s(); say(nqp::istype($n, QAST::Block))
camelia nqp-parrot: OUTPUT«0␤»
FROGGS k
timotimo no commit i added in that range changed something with "type" in it 21:58
BenGoldberg n: say 13, 11, 9 ... 13
camelia niecza v24-98-g473bd20: OUTPUT«13␤»
BenGoldberg n: say 13, 11, 9 ... 11 21:59
camelia niecza v24-98-g473bd20: OUTPUT«13 11␤»
BenGoldberg Why exactly do you need an exclude-first descending list? Surely you can simply write the actual first element?
TimToady lue: what do you think $x ~~ $min .. $max should do if $min happens to be greater than $max? Null ranges are very useful
timotimo that's a good point i hadn't considered 22:00
TimToady degenerate cases must degenerate, not guess a 50/50 WAT vs DWIM
masak TimToady++
yes, a very good point.
lue suddenly remembers masak's words: <masak> lue: TimToady happens to be really, really good at anticipating the way features can come back and bite users. or the way features will interact.
TimToady as for ^... the problem is that the left side is a list, not a single value 22:01
the end is tested as a single value, but the initialize isn't
what would 1,3,5 ^... 7 mean?
lue
.oO(Apparently descending integers are one of those "hard things possible" things. Bah.)
TimToady also, you don't have to write +@elems-1, since @elems-1 means the same thing
lue TimToady: 1, 3, 5 ^... 7 becomes 3,5,7 obviously :) 22:02
masak 'night, #perl6 22:03
lue r: sub infix:<–>($lhs, $rhs) { return $lhs...$rhs }; say $_ for 1–5
♞ masak o/
camelia rakudo-jvm 882e33: OUTPUT«Unhandled exception: java.lang.OutOfMemoryError: Java heap space␤ in (src/Perl6/Grammar.nqp)␤ in (src/vm/jvm/ModuleLoader.nqp:62)␤ in (src/vm/jvm/ModuleLoader.nqp:52)␤ in load_module␤ in ␤␤»
..rakudo-parrot d9e79f: OUTPUT«1␤2␤3␤4␤5␤»
lizmat night masak and #perl6
timotimo r-j: say "what is going on?"
22:03 xenoterracide joined
camelia rakudo-jvm 882e33: OUTPUT«Unhandled exception: java.lang.OutOfMemoryError: Java heap space␤ in (src/Perl6/Grammar.nqp)␤ in (src/vm/jvm/ModuleLoader.nqp:62)␤ in (src/vm/jvm/ModuleLoader.nqp:52)␤ in load_module␤ in ␤␤» 22:03
22:03 lizmat left
lue I suppose I'm just dumbfounded that the difficulty of list generation rises sharply between ascending consecutive integers and descending consecutive integers (with/without excluding the start/endpoint) 22:04
BenGoldberg n: sub infix:<!> { $^a ... $^b }; .say for 1!5 22:05
camelia niecza v24-98-g473bd20: OUTPUT«===SORRY!===␤␤Negation metaoperator not followed by valid infix at /tmp/AsK4dQqZAX line 1:␤------> b infix:<!> { $^a ... $^b }; .say for 1!⏏5␤␤Parse failed␤␤»
BenGoldberg n: sub infix:<_> { $^a ... $^b }; .say for 1 _ 5
camelia niecza v24-98-g473bd20: OUTPUT«1␤2␤3␤4␤5␤»
BenGoldberg n: sub infix:<_> { $^a, $^a - 1 ... $^b }; say 1 _ 5
camelia niecza v24-98-g473bd20: OUTPUT«(timeout)» 22:06
BenGoldberg n: sub infix:<_> { $^a, $^a - 1 ... $^b }; say 5 _ 1
TimToady I suppose we could define LIST ^... END to call (LIST ... END)[1..*] or so
camelia niecza v24-98-g473bd20: OUTPUT«5 4 3 2 1␤»
timotimo there's a unicode character with three dots going from the upper left to the lower right
that sounds like the perfect character for a "descending integer list" operatotr
lue I realize there are (apparently) design decisions going against what I like to see, but I'll eternally disagree with it :)
22:07 kaare_ left
TimToady um, I just agreed with you that ^... could work 22:07
diakopter r-j: say 3
camelia rakudo-jvm 882e33: OUTPUT«3␤»
lue TimToady: I know, I was talking more about .. than anything else.
BenGoldberg r-j: say 'alive?';
TimToady that's not going to change, because .. is about intervals, not integers
camelia rakudo-jvm 882e33: OUTPUT«alive?␤»
BenGoldberg r: sub infix:<_> { $^a, $^a - 1 ... $^b }; say 5 _ 1 22:08
camelia rakudo-parrot d9e79f, rakudo-jvm 882e33: OUTPUT«5 4 3 2 1␤»
lue r: sub infix:<⋱>($lhs, $rhs) { return $lhs...$rhs }; say $_ for 5⋱1 # timotimo++
timotimo oh, intervals. that explains the decision behind it perfectly actually
camelia rakudo-parrot d9e79f, rakudo-jvm 882e33: OUTPUT«5␤4␤3␤2␤1␤»
22:08 wbill_ left
TimToady timotimo: that's why I said it :) 22:09
BenGoldberg sub infix:<^⋱>($lhs, $rhs) { return $lhs-1...$rhs }; say $_ for 5^⋱1
rn: sub infix:<^⋱>($lhs, $rhs) { return $lhs-1...$rhs }; say $_ for 5^⋱1
camelia rakudo-parrot d9e79f, rakudo-jvm 882e33, niecza v24-98-g473bd20: OUTPUT«4␤3␤2␤1␤» 22:10
lue TimToady: Yeah, it makes a lot more sense using the word "interval". (grumblegrumblereverseintervalsgrumblegrumble)^W :)
Also, I want to think some would disagree with ^... ~~ (LIST)[1..*], but I don't see another as-sensible interpretation. 22:11
TimToady if I was hard-hearted, I wouldn't even let you use 1 .. 5 to mean 1 ... 5 :)
22:12 Rotwang left
TimToady now I'm wondering how much we'll see ^... * abused to mean [1..*] 22:12
lue kinda wants to see ⋱/^⋱/⋱^/^⋱^ in Perl 6 now, if only to placate stubborn people like me :D 22:13
TimToady well, we'll have to get rid of ... then and replace it with ⋰ 22:14
timotimo not quite, ... can do any direction 22:15
including zigzag
which is totally a direction.
22:15 denis_boyun__ left
TimToady then we'd have to decide what ⋮ means 22:16
not to mention ⋯
lue Math uses all those crazy kinds of ellipses for a reason, why not the Champion of Unicode™? :) 22:17
TimToady read that as "Chameleon"
lue ... would be the only ASCII equivalent to whatever unicode ellipsis takes the role of current ... (iow, exactly the same) 22:18
TimToady howsomever, those character aren't used to mean that in math, which is a strike against them :)
lue Just because we don't have a reason to use ⋱ for matrices, doesn't mean we can't use them at all :)
TimToady anyway, you should be happy with ^... and agree with me that the others are module-worthy bloat :) 22:19
TimToady dangles a watch in front of lue pendulously and says "You're getting sleepy...sleepy...sleepy..."
lue TimToady: I was thinking the unicode ellipses should be in a (R* supplied) module anyway :) 22:20
TimToady or was that sleep^...sleepy^...sleepy^...
lue →⌚←
lue doesn't have curved arrow compose keys. He thinks. 22:21
japhb__ jnthn: re: "For inlining, subs and multi-subs that are compile-time determinable will cut it, though they need to be simple." , what does "compile-time determinable" mean in this context, and what are the limits of "simple"?
Also, why is the default BUILD slow? Is it completely generic, not type-specialized? 22:22
(I thought the generic part was BUILDALL, but it has been a *long* time since I looked at that stuff)
timotimo the default BUILD goes through a list of "operations" to perform
jnthn japhb__: Well, it is BUILDALL 22:23
japhb__: But if you have your own BUILD method the build plan is just "call that"
japhb__: Not saying it'll be slower forever, just saing how it is now
japhb__ Oh, OK, I see.
jnthn Things should/can be faster thanks to the new/bless changes, though I dunno if anybody updated new yet to be more efficient 22:24
22:24 btyler left, rurban left
jnthn About inlining, the rules are that it needs to determine that the types of the arguments being passed are going to always match the signature. 22:24
timotimo multi method new(*%attrinit) { self.bless(|%attrinit); }
jnthn And it isn't yet especially smart. 22:25
timotimo: yeah, we can do better than that :)
multi method new(|) { nqp::invokewithcapture(nqp::findmethod(self, 'bless'), nqp::usecapture()) } # or something like this
timotimo doesn't know how. just duplicate bless?
oh, mhm
jnthn uh, wait 22:26
*% instead of |
Otherwise we permit positionals.
japhb__ So to guarantee that the inliner knows types of arguments will always match, I just need to explicitly type the arguments at the caller site compatibly with either Any (if the params of the callee are untyped) or the actual type of the param. And type my return values, for expressions involving nested calls. Right? 22:29
dalek ast: d81e6df | jonathan++ | S17-concurrency/scheduler.t:
Tests for timing related scheduler methods.
 22:30
ast: 1f89cf1 | jonathan++ | S17-concurrency/promise.t:
Test for Promise.sleep.
22:31 bluescreen10 left
japhb__ r: my int64 $native64 = 40; my int32 $native32 = 2; say $native64 + $native32; # Will this be able to inline using multi infix:<+>(int, int)? 22:32
camelia rakudo-parrot d9e79f, rakudo-jvm 882e33: OUTPUT«42␤»
dalek kudo/nom: 17e9c13 | jonathan++ | src/vm/jvm/core/Threading.pm:
Re-order to better convey dependencies.

Note that due to $*SCHEDULER this isn't a hard dependency yet, but it could need to become one.
kudo/nom: dcf8e14 | jonathan++ | src/vm/jvm/core/Threading.pm:
Use schedule_with_catch in Promise.
kudo/nom: 91a1b0c | jonathan++ | src/vm/jvm/core/Threading.pm:
Add schedule_in, schedule_every to scheduler.
jnthn japhb__: Should do
And yes, it looks at "returns" when making inlining decisoins too 22:33
japhb__ Is int defined to be platform-native, or int64, or something like the C "at least 64 bits, but maybe more" kind of thing?
jnthn Platform native, but on JVM you can rely on it being 64
japhb__ And Moar too? 22:34
jnthn yes
japhb__ Yay, I get to introduce headaches in my future self!
diakopter :) 22:37
timotimo jnthn: i'm not sure why the code you pasted, the invokewithcapture stuff, is any better than the code rakudo generates for the self.bless(|%attrinit) case. could we easily do better automatically? 22:41
probably only if we can prove the hash hasn't been modified or something
japhb__ If inlining works across comp_units, does that mean that the QAST for a sub is never thrown away?
jnthn japhb__: Right. But note that the other side of this is that only some simple subs are inlinale too 22:42
timotimo .o(is that why our ram usage is so big?)
because some nodes have an $/ attached to them? 22:43
jnthn japhb__: The restrictions essentially mean that only relativley small subs get inlining data stored
japhb__: Which is generally the right choice anyway
timotimo: I think it goes to some effort to trip that out.
*strip
timotimo OK
where do i have to look to figure that out? 22:44
dalek ecs: 766f608 | larry++ | S03-operators.pod:
Allow ^... form just for lue++
ecs: b2f2915 | larry++ | S03-operators.pod:
Mention ^...^ too
lue :D
TimToady who shall we get to implement it, hmm.... 22:45
diakopter throws it at LUE 22:46
22:46 tobyink left
lue uh, I uh, need to sign the CLA first. We must maintain the integrity of TPF's copyright ownership on every single line of code! 22:46
japhb__ jnthn: OK, fair enough. 22:47
timotimo my CLA should be in the hands of the right person soon :3 22:49
jnthn ^...^ - the cat whose nose is exactly between its eyes 22:50
22:50 iSlug left
timotimo ah, i should have come up with a benchmark and measurements for object creation speed before building this 22:51
lue jnthn: a happy person straining to keep his mouth shut (lest he shout joyous excitement for a quarter-hour).
22:51 spider-mario left
jnthn nmake 22:51
um
:)
diakopter hee 22:52
/NOLOGO
timotimo i'm getting some "no subtests run" in my spectests on jvm o_O
timotimo tries again with only one TEST_JOB
diakopter I had to kill the java eval server on host07
japhb__ Next question: How do I make a 'method at_pos() is rw' for a VMArray, so that I can write to a Buf? Is it something as simple as 'nqp::setelems(self, $i) unless nqp::existspos(self, $i); nqp::atpos(self, $i)'?
Hmmm, I guess I have to nqp::decont(self) first ... 22:53
Hmmm, I guess not -- the code in Buf.pm seems to only decont in some cases, not others. 22:55
TimToady wonders how soon till someone asks for ^^... and ...^^ to exclude two values... 22:59
Tene timotimo: If you'd prefer to stop wondering, I could ask for it whenever you like.
TimToady:
TimToady which would actualy work if ^ were a meta-operator :)
timotimo jnthn: do spectests still run for you on jvm? with current rakudo and some local changes i get "no subtests run" for all tests after 22 test files or so 23:00
could it be something like that "getting a bootinterop object will get us into trouble" thing again?
TimToady with a metaop, we could say 'foo' ^xx * to exclude the first 'foo' :) 23:01
or 'foo' xx^ * to exclude the last one
diakopter what would 'foo' ^xx^ 1 do
TimToady apply the infix-meta-circumfix:<^ ^> operator to xx 23:02
and XOR everthing in the list
or was that the x^^x operator, I forget... 23:03
lue I like the sound of metaop ^ so far, but I suspect some nasty things around that syntax corner.
23:04 minimus left
TimToady yes, like preventing ^ff^ from working, and confusing the parsing of ^^ 23:04
diakopter lue: it's an event horizon in actuaity
23:04 minimus joined
lue instead of ^^...^^ there could be ^...^ :downto(2) :upto(2) or something :) 23:04
23:05 minimus is now known as Guest19106
lue
.oO(iffy, diffy, fiddly, excludy)
 23:05
r: sub infix:<RR>($a, $b) { $a + $b }; say 2 RR 3 23:06
camelia rakudo-parrot d9e79f, rakudo-jvm 882e33: OUTPUT«5␤»
timotimo hm, what kind of code would profit from the new .new method that's hopefully faster? 23:07
how many builtin objects have their own new method, i wonder 23:08
jnthn Any code with ojbects that uses the default constructor, I guess 23:09
diakopter r: Mu.new
camelia ( no output )
japhb__ timotimo: Did you figure out the jvm spectest problem?
timotimo no 23:10
jnthn japhb__: The "make Buf writable" problem is the same as the "make native arrays work" and "make int $x is rw as a parameter work" problem...
japhb__ jnthn: Care to expand that just a bit? :-) 23:11
jnthn japhb__: The problem, at its core, is that when you have an array of objects what you really have is Array of Scalar of Object. When we return the Scalar, it is thus assignable 23:12
japhb__: A Buf otoh is compact and just a load of integers
japhb__: There's no container in sight
timotimo .o(create a new proxy instance for each access!)
japhb__ wonders if I'm stuck doing all mutations on an Array, and then having a .serialize() method in my API that just does { Buf.new(@.array) }
jnthn timotimo: Well, it is another kind of container at least...
japhb__: Well, the ugly workaround is to drop to nqp:: ops :) 23:13
japhb__ Hmmm, there is that.
jnthn nqp::bindpos_i(...) should work on a buf.
japhb__ Does bindpos autoextend, or do you need to setelems first? And does setelems preserve the elements that are at lower indices than the new elems and nuke those above? 23:14
timotimo i think: autoextends, nukes elements above new element number 23:15
japhb__ timotimo: I'm wondering about doing setelems(10), setelems(5), setelems(10): How many copies happened, and how many VMArrays are now garbage? 23:16
timotimo well, calling Mu.new a few times doesn't see a dramatic improvement in performance
at least on parrot, we never shrink arrays 23:17
that's how qrpa is implemented
jnthn japhb__: auto-extend
japhb__ jnthn: Is adding *simple* types (just a single class name, no parametric roles or where clauses) ever a performance *loss*? 23:19
jnthn japhb__: In what context? 23:20
japhb__: As in, are type constraints likely to slow you down?
Most likely not, though. The type checks are done against Mu/Any anyways
And you'll hit the type cache for simple types always. 23:21
japhb__ For sub params and returns on one end, and argument and result types on the other end.
Ah, good. Then I feel better about more strictly typing things in the API.
jnthn Yeah, for sub params we're already looking for Any by default anyway
Because we need that to know if we must auto-thread a junction.
japhb__ (Might get a win from the optimizer, but won't lose because I'm now missing a fast path) 23:22
jnthn (which is implemented by noticing binding failed, and seeing a junction was to blame)
japhb__ Gotcha.
jnthn No, you *used* to get a loss way back
Not since 6model/nom landed
japhb__ Thanks for taking the time for all these questions today. :-)
jnthn np :)
japhb__ Under what circumstances do int/Int and str/Str autocoerce? And what does that do to the optimizer? 23:23
(I note that we special case int/int for a lot of infix ops, so I assume it's a win to stop the autocoercer there.) 23:24
timotimo yeah, boxing wastes time :( 23:25
japhb__ My mental model is that you can only avoid all boxing if the arg type is native and the param type is the same native, or the returns type is native and the result type is the same native. In any other combo, you'll box (and potentially unbox again on the other side) 23:27
timotimo is about to find out if the new new method improves performance 23:29
hm, yes, a tiny bit in my probably totally rubbish benchmark 23:30
time ./perl6 -e 'class Foo { has Int $.a; has Str $.b; }; for ^50000 { Foo.new(:a(1), :b<foo>) }'
took 13.55 seconds before, takes 13.3 seconds after
japhb__ Is there any difference between (--> T) and () returns T ? 23:31
(Meaning, does the optimizer see those as different?)
timotimo no, they should be identical
japhb__ Good, because I like having the returns type in the signature. It's one of the things I actually liked about Haskell's syntax. 23:33
TimToady I think --> accepts more things by spec right now, whereas returns only accepts a type
but nobody implement --> Nil yet 23:34
or --> True
dalek ast: 53799ac | jonathan++ | S17-concurrency/promise.t:
Tests for Promise.then.
TimToady std: sub foo ($a --> True) { $a.foo }
camelia std 8adbc60: OUTPUT«ok 00:01 126m␤» 23:35
TimToady std: sub foo ($a) returns True { $a.foo }
camelia std 8adbc60: OUTPUT«ok 00:01 126m␤»
timotimo should that check that $a.foo eq True?
TimToady std: my True sub foo ($a) { $a.foo }
camelia std 8adbc60: OUTPUT«ok 00:01 126m␤»
timotimo er, eqv or something
TimToady hmm
no, it should just return True, and put the final statement into sink context 23:36
timotimo OK
japhb__ r: sub foo ($a --> Nil) { } 23:37
camelia ( no output )
TimToady S06:691
synopsebot Link: perlcabal.org/syn/S06.html#line_691
japhb__ r: sub foo ($a --> Nil) { 42; }; my \a := foo(); say a.WHAT; say a; 23:38
camelia rakudo-parrot 91a1b0, rakudo-jvm 882e33: OUTPUT«===SORRY!===␤CHECK FAILED:␤Calling 'foo' requires arguments (line 1)␤ Expected: :($a)␤»
japhb__ r: sub foo ($a --> Nil) { 42; }; my \a := foo(12); say a.WHAT; say a;
camelia rakudo-jvm 882e33: OUTPUT«Type check failed for return value; expected 'Any' but got 'Int'␤ in block at /tmp/nTMN2eXD6Z:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:1086␤ in any evalfiles at gen/jvm/stage2/NQPHLL.nqp:1292␤ in any command_eval at gen/jvm/stage2/NQPHLL.nqp…»
..rakudo-parrot 91a1b0: OUTPUT«Type check failed for return value; expected 'Any' but got 'Int'␤ in block at /tmp/44wJszXHBy:1␤ in any at /tmp/44wJszXHBy:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPHLL.nqp:1133␤ in any evalfile…»
japhb__ So yeah, rakudobug still.
TimToady Nil is going to be confusing since it's currently implemented as (also) a type
std: my 42 sub foo ($a) { $a.foo } 23:39
camelia std 8adbc60: OUTPUT«===SORRY!===␤Malformed my at /tmp/qCTaEuUIKa line 1:␤------> my ⏏42 sub foo ($a) { $a.foo }␤ expecting any of:␤ name␤ scoped declarator␤Parse failed␤FAILED 00:01 120m␤»
japhb__ r: sub foo ($a --> True) { 42; }; my \a := foo(12); say a.WHAT; say a;
TimToady std: sub foo ($a --> 42) { $a.foo }
camelia rakudo-parrot 91a1b0: OUTPUT«Type check failed for return value; expected 'Bool' but got 'Int'␤ in block at /tmp/f3W2EdkXyg:1␤ in any at /tmp/f3W2EdkXyg:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPHLL.nqp:1133␤ in any evalfil…»
..rakudo-jvm 882e33: OUTPUT«Type check failed for return value; expected 'Bool' but got 'Int'␤ in block at /tmp/5s44lBDHAa:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:1086␤ in any evalfiles at gen/jvm/stage2/NQPHLL.nqp:1292␤ in any command_eval at gen/jvm/stage2/NQPHLL.nq…»
std 8adbc60: OUTPUT«ok 00:01 126m␤»
TimToady std: sub foo ($a) returns 42 { $a.foo } 23:40
camelia std 8adbc60: OUTPUT«===SORRY!===␤Invalid type name at /tmp/U9iGnY767E line 1:␤------> sub foo ($a) returns ⏏42 { $a.foo }␤ expecting typename␤Parse failed␤FAILED 00:01 122m␤»
TimToady yes, --> is special for fixed return values
japhb__ r: sub foo ($a --> Nil) { }; my \a := foo(12); say a.WHAT; say a;
camelia rakudo-parrot 91a1b0, rakudo-jvm 882e33: OUTPUT«Nil␤Nil␤»
japhb__ r: sub foo ($a --> True) { }; my \a := foo(12); say a.WHAT; say a;
camelia rakudo-jvm 882e33: OUTPUT«Type check failed for return value; expected 'Bool' but got 'Any'␤ in block at /tmp/u7i40ufUOE:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:1086␤ in any evalfiles at gen/jvm/stage2/NQPHLL.nqp:1292␤ in any command_eval at gen/jvm/stage2/NQPHLL.nq…»
..rakudo-parrot 91a1b0: OUTPUT«Type check failed for return value; expected 'Bool' but got 'Any'␤ in sub foo at /tmp/T13MYdokYo:1␤ in block at /tmp/T13MYdokYo:1␤ in any at /tmp/T13MYdokYo:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage…»
TimToady Nil only *appears* to do the right thing
japhb__ r: sub foo ($a is copy --> Nil) { repeat while $a > 0 { $a--; } }; my \a := foo(12); say a.WHAT; say a; 23:41
camelia rakudo-jvm 882e33: OUTPUT«Type check failed for return value; expected 'Any' but got 'Bool'␤ in block at /tmp/AotaN_hPf0:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:1086␤ in any evalfiles at gen/jvm/stage2/NQPHLL.nqp:1292␤ in any command_eval at gen/jvm/stage2/NQPHLL.nq…»
..rakudo-parrot 91a1b0: OUTPUT«Type check failed for return value; expected 'Any' but got 'Bool'␤ in sub foo at /tmp/Eki5_UCjlA:1␤ in block at /tmp/Eki5_UCjlA:1␤ in any at /tmp/Eki5_UCjlA:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage…»
japhb__ Yeah, clearly so. 23:42
TimToady it should neither be expecting an Any nor getting a Bool
that looks suspiciously like P5 semantics returning the result of > 23:43
japhb__ nodnod 23:46
23:48 stevan_ joined 23:50 benabik joined
BenGoldberg How odd: 23:52
r: sub foo ($a --> True) { True }; my \a := foo(12); say a.WHAT; say a;
camelia rakudo-jvm 882e33: OUTPUT«Type check failed for return value; expected 'Bool' but got 'Bool'␤ in block at /tmp/rCasMGmpyL: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.n…»
..rakudo-parrot 91a1b0: OUTPUT«Type check failed for return value; expected 'Bool' but got 'Bool'␤ in block at /tmp/5Sjmr3WHY4:1␤ in any at /tmp/5Sjmr3WHY4:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPHLL.nqp:1133␤ in any evalfi…»
BenGoldberg Expected Bool but got Bool? 23:53
jnthn Most likely it only knows how to cope with a type object there right now 23:55
japhb__ r: my $buf = buf8.new(); nqp::bindpos_i($buf, 0, 42); say $buf[0];
camelia rakudo-parrot 91a1b0: OUTPUT«42␤»
..rakudo-jvm 882e33: OUTPUT«java.lang.RuntimeException: This type does not support positional operations␤ in block at /tmp/SxLCjGJSC0:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:1086␤ in any evalfiles at gen/jvm/stage2/NQPHLL.nqp:1292␤ in any command_eval at gen/jvm/stage…»
japhb__ Damn you, JVM! :-/
BenGoldberg r: sub foo ($a --> Bool) { True }; my \a := foo(12); say a.WHAT; say a; 23:56
camelia rakudo-parrot 91a1b0, rakudo-jvm 882e33: OUTPUT«(Bool)␤True␤»
BenGoldberg r: sub foo ($a --> Bool) {}; my \a := foo(12); say a.WHAT; say a;
jnthn japhb__: Try my $buf := ...
camelia rakudo-jvm 882e33: OUTPUT«Type check failed for return value; expected 'Bool' but got 'Any'␤ in block at /tmp/A9fl1IiEjp: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.nq…»
..rakudo-parrot 91a1b0: OUTPUT«Type check failed for return value; expected 'Bool' but got 'Any'␤ in sub foo at /tmp/vKsBYctYq0:1␤ in block at /tmp/vKsBYctYq0:1␤ in any at /tmp/vKsBYctYq0:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage…»
japhb__ r: my $buf := buf8.new(); nqp::bindpos_i($buf, 0, 42); say $buf[0];
camelia rakudo-parrot 91a1b0, rakudo-jvm 882e33: OUTPUT«42␤»
japhb__ Wha...?
r: nqp::bindpos_i(buf8.new(), 0, 42); say "alive"; 23:57
camelia rakudo-parrot 91a1b0, rakudo-jvm 882e33: OUTPUT«alive␤»
BenGoldberg r: my $buf := buf8.new(); say $buf[0];
camelia rakudo-jvm 882e33: OUTPUT«(timeout)» 23:58
..rakudo-parrot 91a1b0: OUTPUT«VMArray: index out of bounds␤ in method at_pos at gen/parrot/CORE.setting:5016␤ in method at_pos at gen/parrot/CORE.setting:1654␤ in sub postcircumfix:<[ ]> at gen/parrot/CORE.setting:2322␤ in sub postcircumfix:<[ ]> at gen/parrot/CORE.settin…»
jnthn japhb__: Well, you get kinda used to Rakudo sneaking the deconts into the right places often enough for you that you don't notice how much they're needed ;)
japhb__ jnthn: So why does r-j want the binding, but r-p does not care?
jnthn japhb__: I guess r-p is doing a decont on that operation, and r-j isn't.
And nothing in spectest actually hits it either way so far
japhb__ heh, gotcha.
jnthn And it's a grey area being an nqp:: op...
japhb__ sure
dalek ast: 4623a73 | jonathan++ | S17-concurrency/promise.t:
Tests for Promise.allof and Promise.anyof.
 23:59
BenGoldberg r: my $buf := buf8.new(); ++$buf[0];
camelia rakudo-jvm 882e33: OUTPUT«(timeout)»
..rakudo-parrot 91a1b0: OUTPUT«VMArray: index out of bounds␤ in method at_pos at gen/parrot/CORE.setting:5016␤ in method at_pos at gen/parrot/CORE.setting:1654␤ in sub postcircumfix:<[ ]> at gen/parrot/CORE.setting:2322␤ in sub postcircumfix:<[ ]> at gen/parrot/CORE.settin…»
BenGoldberg What does the jvm timeout?
Err, Why?

Powered by Cro and the Raku® Programming Language.

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

Thank you!