»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.
00:01 denis_boyun left 00:04 kivutar left
dalek osystem: f31aa9e | (Andrew Egeler)++ | META.list:
Add Compress::Zlib
 00:07
00:08 jnap left
segomos raiph: if you have a back up somewhere i can throw it up on a site 00:08
building jvm p6, panda doesn't like moar yet i guess :-)
00:09 pecastro left 00:20 dayangkun left, dayangkun joined
dalek rl6-most-wanted: c98cda9 | (Andrew Egeler)++ | most-wanted/bindings.md:
Add Compress::Zlib WIP
 00:22
00:23 kbaker joined
TimToady r: my \m = 1000; say m; # oops 00:23
camelia rakudo-parrot 934c48, rakudo-jvm 934c48, rakudo-moar 934c48: OUTPUT«===SORRY!=== Error while compiling /tmp/tmpfile␤Regex not terminated␤at /tmp/tmpfile:1␤------> my \m = 1000; say m; # oops⏏<EOL>␤ expecting any of:␤ argume…»
TimToady this discussion reminds me we need to do something about that
segomos r: my \m = 1000; m.say; 00:24
camelia rakudo-jvm 934c48, rakudo-moar 934c48: OUTPUT«===SORRY!===␤Unrecognized regex metacharacter ; (must be quoted to match literally)␤at /tmp/tmpfile:1␤------> my \m = 1000; m.say⏏;␤Couldn't find terminator .␤at /tmp/tmpfile:1␤------> …»
..rakudo-parrot 934c48: OUTPUT«␀␀␀[␀␀␀3␀␀␀1␀␀␀m␀␀␀=␀␀␀=␀␀␀=␀␀␀␀␀␀[␀␀␀0␀␀␀m␀␀␀S␀␀␀O␀␀␀R␀␀␀R␀␀␀Y␀␀␀!␀␀␀␀␀␀[␀␀␀3␀␀␀1␀␀␀m␀␀␀=␀␀␀=␀␀␀=␀…»
TimToady there are a number of keywords that ought to be demoted to macro names
segomos cool error message
00:30 hoverboard left 00:46 hoverboard joined 00:56 woosley joined 01:02 raiph left 01:09 kbaker left
TimToady hi guys, you're on screen a De Anza College in Cupertino :) 01:30
01:30 fridim__ joined
JimmyZ std: my \m = 1000; m.say; 01:48
camelia std 09dda5b: OUTPUT«===SORRY!===␤Regex missing terminator (or semicolon must be quoted?) at /tmp/grUz9n_zx2 line 1 (EOF):␤------> my \m = 1000; m.say;⏏<EOL>␤Parse failed␤FAILED 00:01 124m␤»
JimmyZ std: my \m = 1000; say m; 01:49
camelia std 09dda5b: OUTPUT«===SORRY!===␤Regex not terminated at /tmp/cjk9tn3dSF line 1 (EOF):␤------> my \m = 1000; say m;⏏<EOL>␤Parse failed␤FAILED 00:01 122m␤»
[Coke] aaahaahahah. I checked the whois, found a name, checked on linkedin... twgil dude has 2 recommendations for perl. :)
01:52 cooper left, erdic left, erdic joined 01:56 iolympian1 joined 01:59 cooper joined
lue TimToady: really? 02:02
TimToady yes, really 02:03
run something to show it off :)
like maybe 02:04
m: say [*] 1..10
camelia rakudo-moar 934c48: OUTPUT«3628800␤»
lue r: sub postfix:<!>($a) { [*] 1..$a }; say 5!;
camelia rakudo-parrot 934c48, rakudo-jvm 934c48, rakudo-moar 934c48: OUTPUT«120␤»
lue hey look, we picked the same reduction op :P 02:05
TimToady why did we multiply by 1?
lue TimToady: that's just the canonical definition of factorial, I didn't make it up. :)
TimToady j: say [*] 1..10 02:06
camelia rakudo-jvm 934c48: OUTPUT«3628800␤»
TimToady p: say [*] 1..10
camelia rakudo-parrot 934c48: OUTPUT«3628800␤»
JimmyZ r: say [*] 1..30
camelia rakudo-parrot 934c48, rakudo-jvm 934c48, rakudo-moar 934c48: OUTPUT«265252859812191058636308480000000␤»
skids gist.github.com/skids/9634210 # this something that's RTed yet? 02:07
lue hopes there's a way to check for invalid var names without maintaining a list of keywords for at least part of it :/
TimToady p6: say 42
camelia rakudo-parrot 934c48, rakudo-jvm 934c48, rakudo-moar 934c48, niecza v24-109-g48a8de3: OUTPUT«42␤»
JimmyZ r: my $time = time; say [*] 1..32; say $time - time 02:08
camelia rakudo-parrot 934c48, rakudo-jvm 934c48, rakudo-moar 934c48: OUTPUT«263130836933693530167218012160000000␤0␤»
lue r: say [*] 1..32; say now - BEGIN now
camelia rakudo-moar 934c48: OUTPUT«263130836933693530167218012160000000␤0.04657962␤»
..rakudo-parrot 934c48: OUTPUT«263130836933693530167218012160000000␤0.05436981␤»
..rakudo-jvm 934c48: OUTPUT«263130836933693530167218012160000000␤0.077␤»
JimmyZ :)
lue r: say (1,2,3) ∪ (2,3,4); say (1,2,3) ∩ (2,3,4); say (1,2,3) ∖ (2,3,4); say (1,2,3) ⊖ (2,3,4) 02:12
camelia rakudo-jvm 934c48: OUTPUT«set(2, 1, 4, 3)␤set(2, 3)␤set(1)␤set(1, 4)␤»
..rakudo-parrot 934c48, rakudo-moar 934c48: OUTPUT«set(1, 2, 3, 4)␤set(2, 3)␤set(1)␤set(1, 4)␤»
TimToady r: say (1,2,3) (|) (2,3,4)
camelia rakudo-parrot 934c48, rakudo-moar 934c48: OUTPUT«set(1, 2, 3, 4)␤» 02:13
..rakudo-jvm 934c48: OUTPUT«set(2, 1, 4, 3)␤»
lue ... and just now I realize that script I wrote a couple days ago could have used ⊖ instead of putting two lists into a Bag and finding the unique ones. :P
TimToady eval: say 02:16
do we have a P5 here?
lue also, TimToady: think the set operators for regex assertions should 1) follow the (^) etc. form for the texas versions, instead of the ()-less versions they are now, and 2) have Unicode versions?
p5: say
TimToady p6: say 02:17
camelia niecza v24-109-g48a8de3: OUTPUT«===SORRY!===␤␤Unsupported use of bare 'say'; in Perl 6 please use .say if you meant $_, or use an explicit invocant or argument at /tmp/tmpfile line 1 (EOF):␤------> say⏏<EOL>␤␤Unhandled exception: Chec…»
..rakudo-parrot 934c48, rakudo-jvm 934c48, rakudo-moar 934c48: OUTPUT«␤»
lue ( that is, <rule1 ∪ rule2> and <rule1 (|) rule2> instead of what it is now, <rule1 | rule2> )
02:17 fridim__ left
TimToady p6: $_ = 'foo'; .say 02:17
camelia rakudo-parrot 934c48, rakudo-jvm 934c48, rakudo-moar 934c48, niecza v24-109-g48a8de3: OUTPUT«foo␤»
TimToady m: for my $x (@_) {...} 02:19
camelia rakudo-moar 934c48: OUTPUT«===SORRY!=== Error while compiling /tmp/2xMfoWrng7␤This appears to be Perl 5 code␤at /tmp/2xMfoWrng7:1␤------> for ⏏my $x (@_) {...}␤ expecting any of:␤ statement list␤»
TimToady m: 42.chars.say 02:21
camelia rakudo-moar 934c48: OUTPUT«2␤»
02:22 xragnar left 02:24 xragnar joined 02:25 airdisa left
TimToady m: say <a b c> ~ 1,2,3; 02:29
camelia rakudo-moar 934c48: OUTPUT«a b c123␤»
TimToady m: say <a b c> Z 1,2,3;
camelia rakudo-moar 934c48: OUTPUT«a 1 b 2 c 3␤»
02:29 iolympian1 left
TimToady m: say <a b c> Z~ 1,2,3; 02:29
camelia rakudo-moar 934c48: OUTPUT«a1 b2 c3␤»
TimToady m: say <a b c> X~ 1,2,3; 02:31
camelia rakudo-moar 934c48: OUTPUT«a1 a2 a3 b1 b2 b3 c1 c2 c3␤»
TimToady m: say 5 / 3 02:32
camelia rakudo-moar 934c48: OUTPUT«1.666667␤»
TimToady m: say 5 R/ 3
camelia rakudo-moar 934c48: OUTPUT«0.6␤»
TimToady m: say <a b c> XR~ 1,2,3; 02:34
camelia rakudo-moar 934c48: OUTPUT«1a 2a 3a 1b 2b 3b 1c 2c 3c␤»
02:52 klapperl left 02:54 klapperl joined 02:56 FROGGS_ joined 03:00 FROGGS left
BenGoldberg p6: say [R,] 1..4 03:09
camelia rakudo-parrot 934c48: OUTPUT«Too many positional parameters passed; got 4 but expected 2␤ in block at gen/parrot/CORE.setting:17009␤ in sub at gen/parrot/CORE.setting:17151␤ in block at /tmp/tmpfile:1␤␤»
..rakudo-moar 934c48: OUTPUT«Too many positional parameters passed; got 4 but expected 2␤ in block at src/gen/m-CORE.setting:16808␤ in sub at src/gen/m-CORE.setting:16950␤ in block at /tmp/tmpfile:1␤␤»
..niecza v24-109-g48a8de3: OUTPUT«Unhandled exception: Excess arguments to ANON, used 2 of 4 positionals␤ at /tmp/tmpfile line 0 (ANON @ 1) ␤ at /home/p6eval/niecza/lib/CORE.setting line 2767 (reduceop @ 12) ␤ at /tmp/tmpfile line 1 (mainline @ 3) ␤ at /home/p6eval/niecza/…»
..rakudo-jvm 934c48: OUTPUT«Too many positional parameters passed; got 4 but expected 2␤ in block at gen/jvm/CORE.setting:16754␤ in sub at gen/jvm/CORE.setting:16896␤ in block at /tmp/tmpfile:1␤␤»
03:14 SamuraiJack joined 03:23 perigrin_ joined 03:24 perigrin_ left, perigrin_ joined, perigrin_ left 03:44 SamuraiJack left 03:58 yoleaux left 03:59 yoleaux joined, ChanServ sets mode: +v yoleaux
segomos perl6 is awesome 04:13
04:40 xybre joined
xybre Whats the proper syntax to use the "continue" modifier in a rule block? I've tried a few combinations, but the only one that parsed said :c wasn't a modifier. Is it not implemented yet? 04:42
TimToady :c is mostly for bare regex; once you're in a grammar things automatically continue where they're supposed to, generally speaking 04:46
so I don't think anyone has ever tried to use :c in a rule, offhand 04:47
xybre Interesting, I'll look more into the grammar documentation then. Thanks. 04:49
TimToady well, it's documented that you can use it internally, but I guess nobody's tried it yet 04:50
everyone thinks of it as a modifier to m// or s///
I'm not sure <at()> is impleented either 04:51
m: 'foo' ~~ /...<at(3)>/
camelia ( no output )
TimToady m: say 'foo' ~~ /...<at(3)>/
camelia rakudo-moar 934c48: OUTPUT«｢foo｣␤ at => ｢｣␤␤»
TimToady I guess it is 04:52
m: say 'foo' ~~ /.*?<at(3)>/
camelia rakudo-moar 934c48: OUTPUT«｢foo｣␤ at => ｢｣␤␤»
TimToady I guess there aren't a lot of parsing problems where you have to go jumping around in the string 04:53
that feels more like SNOBOL-think to me... 04:54
certainly we have the capability of setting the current position, but :c is perhaps not the best interface for that 04:56
I don't think the current position setter is going to be found in S05 though
and now that I grep for it, it looks like primarily an nqp operation, so dunno if it's visible in P6-land 04:58
m: say 'foo' ~~ /<.cursor_pos(3)>$/ 04:59
camelia rakudo-moar 934c48: OUTPUT«No such method 'cursor_pos' for invocant of type 'Cursor'␤ in method ACCEPTS at src/gen/m-CORE.setting:12825␤ in method ACCEPTS at src/gen/m-CORE.setting:798␤ in block at /tmp/A05w87Om41:1␤␤»
TimToady it would be something resembling that
m: say Cursor.^methods
camelia rakudo-moar 934c48: OUTPUT«No such method 'gist' for invocant of type 'NQPRoutine'␤ in method gist at src/gen/m-CORE.setting:7022␤ in method gist at src/gen/m-CORE.setting:1053␤ in sub say at src/gen/m-CORE.setting:13366␤ in block at /tmp/4FJ1Kdr0np:1␤␤»
TimToady heh
grammars aren't quite HLLized completely in rakudo yet 05:00
05:00 kaare_ joined
TimToady m: say Cursor.^methods».name 05:01
camelia rakudo-moar 934c48: OUTPUT«MATCH MATCH_SAVE INTERPOLATE OTHERGRAMMAR INDRULE RECURSE orig target from pos CAPHASH !cursor_init !cursor_start_all !cursor_start !cursor_start_cur !cursor_start_fail !cursor_start_subcapture !cursor_capture !cursor_push_cstack !cursor_pass !cursor_fail …»
TimToady m: say 'foo' ~~ /<.'!cursor_pos'(3)>$/ 05:02
camelia rakudo-moar 934c48: OUTPUT«===SORRY!===␤Unrecognized regex metacharacter < (must be quoted to match literally)␤at /tmp/bVOGMsVHSu:1␤------> say 'foo' ~~ /<.⏏'!cursor_pos'(3)>$/␤Unrecognized regex metacharacter . (must be quoted to match…»
TimToady m: say 'foo' ~~ /<."!cursor_pos"(3)>$/ 05:03
camelia rakudo-moar 934c48: OUTPUT«===SORRY!===␤Unrecognized regex metacharacter < (must be quoted to match literally)␤at /tmp/PsYa6Z1Sd1:1␤------> say 'foo' ~~ /<.⏏"!cursor_pos"(3)>$/␤Unrecognized regex metacharacter . (must be quoted to match…»
TimToady well, might have to write it more nqp-ish than that, I guess 05:04
anyhoo, the functionality is certainly there somewhere hiding
TimToady hides 05:05
JimmyZ Cursor.pm needs some love, me thinks 05:06
05:10 logie_ left 05:13 BenGoldberg left 05:15 xenoterracide left 05:25 [Sno] left 05:26 [Sno] joined, [Sno] left 05:27 [Sno] joined, [Sno] left 05:36 rurban1 left
moritz \o 05:39
05:43 kaare_ left 05:48 pdcawley left 05:49 pdcawley joined 06:09 kaare_ joined 06:16 logie_ joined 06:18 kaare_ left 06:20 kaare_ joined 06:25 hoverboard left 06:27 [Sno] joined 06:34 Khisanth left 06:56 Khisanth joined 06:57 sftp left 06:59 sftp joined 07:00 Khisanth left 07:09 hoverboard joined, FROGGS_ left 07:10 arnsholt left 07:12 hoverboard left 07:15 darutoko joined 07:30 kaleem joined 07:38 FROGGS joined 07:41 Khisanth joined 07:46 rindolf joined 07:49 arnsholt joined 07:50 dwarring_ left 07:56 zakharyas joined
arnsholt nqp: say(nqp::reprname(+[])) 08:14
camelia nqp-parrot: OUTPUT«Can only use repr_name with a SixModelObject␤current instr.: '' pc 49 ((file unknown):57503018) (/tmp/tmpfile:1)␤»
..nqp-moarvm, nqp-jvm: OUTPUT«P6num␤»
arnsholt jnthn: Do we consider that a bug, or merely unexpected? 08:15
JimmyZ nqp: say(nqp::reprname((+[])).WHAT) 08:16
camelia nqp-parrot: OUTPUT«Can only use repr_name with a SixModelObject␤current instr.: '' pc 49 ((file unknown):48512506) (/tmp/tmpfile:1)␤»
..nqp-moarvm, nqp-jvm: OUTPUT«␤»
FROGGS arnsholt: it is the "that happens when 6model is not fully supported/implemented" thing 08:17
arnsholt: but when you glance at #parrot, then you see commits from rurban++ about implementing 6model
arnsholt Oh, the problem isn't that it doesn't work on Parrot 08:18
Sorry, that wasn't entirely clear
The +@thing returns the length of the array, which we'd normally expect to be a P6int
But prefix:<+> only returns nums 08:19
08:19 xinming_ joined, dmol joined
FROGGS arnsholt: that is known 08:22
and for some reason we were unable to just change it 08:23
but I can't remember why
arnsholt It's because the op driving prefix:<+> has to return a num in some other contexts, so it has to do that for everything 08:30
08:54 kivutar joined
timotimo o/ 09:03
arnsholt 'lo \o 09:15
09:23 pecastro joined 09:28 fhelmberger joined 09:29 virtualsue joined 09:35 SamuraiJack joined
arnsholt Oh fun. infix:<+> of two ints produces a num too 09:52
timotimo it always does? oops. 09:54
in nqp we already have an optimization that turns num additions into int additions if it's obvious
on rakudo itself, that may be harder, because the additions aren't always inlined (or are they?) 09:55
arnsholt Looks like from what I'm seeing 09:56
nqp: my $a := 1; my $b := 2; say(nqp::reprname($a)); say(nqp::reprname($b)); say(nqp::reprname($a+$b))
camelia nqp-moarvm, nqp-jvm: OUTPUT«P6int␤P6int␤P6num␤» 09:57
..nqp-parrot: OUTPUT«Can only use repr_name with a SixModelObject␤current instr.: '' pc 52 ((file unknown):45) (/tmp/tmpfile:1)␤»
arnsholt Yup
timotimo o_O
tadzik o:
moritz nqp: say(nqp::reprname(1+2))
camelia nqp-moarvm, nqp-jvm: OUTPUT«P6int␤»
..nqp-parrot: OUTPUT«Can only use repr_name with a SixModelObject␤current instr.: '' pc 46 ((file unknown):62151977) (/tmp/tmpfile:1)␤»
timotimo well, at least that works.
arnsholt I could fix it by doing +| instead (since that's what it actually wanted to do), thankfully
timotimo nqp: my int $a := 1; my int $b := 2; say(nqp::reprname($a+$b)) 09:58
camelia nqp-moarvm, nqp-jvm: OUTPUT«P6int␤»
..nqp-parrot: OUTPUT«Can only use repr_name with a SixModelObject␤current instr.: '' pc 53 ((file unknown):45) (/tmp/tmpfile:1)␤»
timotimo ^
arnsholt \o/ 09:59
Direct JAST compilation stage1 compiler now manages to compile the stage2 compiler
I wonder what hilarious bugs the stage2 compiler now has =D
timotimo \o/
this will decrease startup time measurably, aye? 10:00
well, at least compile-time?
arnsholt Wait, what? It didn't explode...
timotimo well, stage JAST only, really
10:00 woosley left
arnsholt Hoping it'll cut down on compile time a bit yeah 10:00
moritz arnsholt: when it gets as far as building stage2, chances are good that it's not all that messed up :-) 10:01
arnsholt Apparently =D 10:02
Let's see what happens if I make j-test 10:03
timotimo oooh, if all is well, this could go in this month's release! 10:05
moritz timotimo: you do realize that that's tomorrow already?
I'm not a fan of rushing things into releases 10:06
after all, the next one is only a month away :-)
timotimo hehe.
we've had holes in the Star releases recently, though
arnsholt I'd like to at least get jnthn to review it first
timotimo that's fine
i always get a bit too excited about these things that happen immediately before a release
arnsholt All tests passing so far though 10:07
"All tests successful." =D 10:08
timotimo well, i like the sound of that :) 10:10
10:23 telex left 10:24 telex joined 11:25 denis_boyun joined 11:40 SamuraiJack left, fhelmberger_ joined 11:44 fhelmberger left 11:54 skids left, rindolf left
timotimo how much work would it be to make rakudo-moar capable of running STD? 12:12
nwc10 what does STD currently run on?
timotimo viv
and thus, perl5
jnthn o/ 12:13
12:13 brrt joined
timotimo o/ jnthn 12:14
how's the tuit supply today?
nwc10 hey, can we have 12 sided tuits soon? :-)
jnthn Tuit supply a bit disrupted by internet access failure 12:15
But it's OK; I needed to do a run to the office sometimes today or tomorrow, so I did it today :)
timotimo arnsholt: were you able to compile a full rakudo with the new jast compiler yet? 12:17
jnthn On NQP, it basically resorts to floating point most of the time. 12:18
Unless you give it things typed as int.
arnsholt timotimo: Compiles, but for some reason weird things happen if I try to make j-test 12:19
Gonna recompile NQP to master and see what happens
timotimo okay
arnsholt O hai, jnthn! 12:20
jnthn arnsholt++ # awesome progress
I'm too risk averse to have it in the monthly :)
I'll bet it saves some off setting compilation time on JVM. :)
arnsholt Hopefully. As long as it's not broken in some subtle way 12:21
Had weirdness with the test suite
jnthn For such a big change, I expect there'll be a few things to work through :)
arnsholt make j-test gets me "no tests run" errors, but ./perl6-j t/$a_test_file seemed to work fine
Recompiling NQP right now, but once the compilations are done I'll push what I have so far 12:22
timotimo you should try it on master, see if it's supposed to work at all in a vanilla state
because "no tests run" is what i get most of the time for the j-spectest
arnsholt Aha
jnthn Yeah, that seems to happen on some JVM versions :S 12:23
arnsholt From a skim of the harness script, it looks like it might not be running ./perl6-j on JVM
jnthn Yeah, it uses the evalserver for spectest
timotimo yeah, it may be running the ... yes
hmm. a huge piece of slowness for the perl6-j benchmarks is attributed to stage parse, at least if --stagestats is to be believed 12:25
does that also contain loading and deserializing the setting?
jnthn Yes.
Happens for the fist thing you parse 12:26
uh, first
timotimo hmm. that won't get any benefit from the jast speed improvements
it would seem like loading the setting is to blame for by far the biggest chunk of time wasted on startup for rakudo-jvm benchmarks 12:27
jnthn Well, it's not waste if you gotta do it :P 12:29
Question is what exactly takes time, and what aspects of it re slow.
timotimo i'm assuming that there's something to be gained there ;)
12:31 kaleem left
jnthn Me too :) 12:31
timotimo which tool would give me help with that? jvisualvm? 12:32
jnthn maybe that, or maybe there are some profiling options that can be passed to the JVM invocation too 12:33
12:34 fhelmberger_ left
arnsholt Hmm, looks like it works with plain perl6-j 12:34
12:34 fhelmberger joined
dalek p/jastcompiler: 7e6d19f | (Arne Skjærholt)++ | src/vm/jvm/QAST/Compiler.nqp:
Make sure all handler indices are ints.

Prior to this, some indices would be nums; one because the construct "$a+$b" returns a num, even if both variables are ints (fixed by using logical or instead, since those are the semantics we actually wanted anyways), and one because "+@array" also returns a num (fixed by using nqp::elems instead).		 12:34
p/jastcompiler: fe4a4ac | (Arne Skjærholt)++ | / (18 files):
Direct compilation of JAST nodes to bytecode.
arnsholt That's the branch, at any rate 12:35
Should probably do some $dayjob things today, too =)
JimmyZ timotimo: java.dzone.com/articles/java-profiling-dtrace
timotimo thank you :) 12:36
JimmyZ I know the oracle dbas is using it
timotimo so all i need to do is install a solaris and build rakudo-j on that!
arnsholt You can get dtrace on Linux as well, I think
JimmyZ docs.oracle.com/javase/6/docs/techn...trace.html 12:37
arnsholt But it's not in the mainline kernel because the dtrace licence isn't compatible with the GPL
timotimo right
JimmyZ yeah, github.com/dtrace4linux/linux
timotimo did i mention i can't profile perl6-j because the jvm segfaults when i click the "profile" button? 12:38
12:40 kivutar left
timotimo i can only let it sample every 20ms 12:41
jnthn wow 12:42
Try a newer/different version?
timotimo it also spews a bunch of warnings before that
probably related to the way we create lots of classes at run-time
gist.github.com/timo/0c08d358a2357e098da7 12:43
jnthn wow 12:44
fail
moritz timotimo: do submit a bug report to the JVM devs :-) 12:46
timotimo i'm afraid they'll tell me it's a DIHWIDT 12:47
12:53 guru joined, guru is now known as Guest94710 12:54 Guest94710 is now known as ajr_ 12:59 kbaker_ joined
dalek kudo/nom: 4869942 | rurban++ | src/vm/parrot/ops/perl6.ops:
parrot/ops/perl6.ops: add missing prototype decl for Parrot_lib_perl6_ops_init

required for parrot-6.1.0 which fatalizes missing prototype declarations		 13:01
kudo/nom: 01d16c2 | (Tobias Leich)++ | src/vm/parrot/ops/perl6.ops:
Merge pull request #252 from rurban/nom

parrot/ops/perl6.ops: add missing prototype decl for Parrot_lib_perl6_op...
13:12 Sqirrel joined 13:22 lowpro30 joined
timotimo i'm wondering if it would be wise to let int + int give int given ints range limitations; but that might be what the user wants anyway, since nqp doesn't give you bigint semantics for int 13:23
FROGGS that sounds pretty sane for nqp me thinks 13:25
jnthn NQP does not have to be a general purpose language. It just needs to be good for writing compilers. :)
JimmyZ and for C :P
jnthn To be honest, it's floating point we should probably make cost.
And assume int in general.
Since compilers rarely need to do scientific math :) 13:26
JimmyZ there is XXX in the nqp that is nqp really wants constants
timotimo aye; but for now we can probably teach prefix:<+> before an @-variable to return an int
jnthn Yeah, that's a todo :)
timotimo: yes, that is an acceptable thing to do for now.
timotimo: In fact, we can even re-write it directly to an elems call 13:27
timotimo not 100% sure where that goes, though
13:31 lowpro30_ joined 13:32 lowpro30 left 13:39 lowpro30_ left
arnsholt timotimo: The grammar bit handling prefix:<+> probably 13:43
timotimo yeah, probably
arnsholt Then it can special-case if its argument is a variable with the @ sigil and compile to the elems op, instead of numify 13:44
Er Actions, not grammar, really
timotimo yes
13:47 AW3i joined
jnthn I think maybe in NQP::Ops, which is backend-specific. 13:48
arnsholt Oh, you mean fiddling with the AST directly in QAST::Compiler? 13:49
jnthn No, it's that the numify thing is implemented differnetly by backend.
iirc
arnsholt Oh, that 13:50
But at least on JVM the return type of the numify op is Double
I was thinking that the grammar would just emit an elems op when the argument is a literal variable whose sigil is @ 13:51
jnthn Well, it's not the grammars jot to emit anything 13:53
13:53 rurban1 joined
jnthn It'd be in the actions if anywhere 13:53
FROGGS <arnsholt> Er Actions, not grammar, really
arnsholt Yeah, I keep saying grammar, but it's the actions of course >.< 13:54
I think my brain just doesn't want to separate the actions and the grammar =)
timotimo worst case, the optimizer can do it.
because the important point is that everything downstream from the optimizer gets to know that it returns int. 13:55
FROGGS .oO( worst case, the timotimo can do it )
jnthn optimizer feels wrong
Actions can work.
FROGGS I always think "timotimo" when I read "optimizer" for some reason :o)
timotimo fair enough, i'll have a look
13:56 ajr_ left
arnsholt timotimo: I took a peek at the code, and prefix:sym<+> in NQP::Grammar sets the op via the <O()> directive 13:56
moritz IMHO it's not a job for the grammar 13:57
but rather for the code-gen or the optimizer
timotimo i see that
in theory, the prefix:<+> rule could do a lookahead for an @ character
and generate a nqp::elems instead 13:58
13:58 btyler joined, guru joined
arnsholt Don't even have to touch the grammar. The action can just make a different AST based on the type of the AST of its argument 13:58
13:59 guru is now known as ajr_
arnsholt moritz: That's true too, but then we'll have needlessly duplicated code to handle this 13:59
timotimo arnsholt: but then i'd have to replace the functionality of the O in that case :) 14:00
probably would just have to remove the :op piece, though
FROGGS - token prefix:sym<+> { <sym> <O('%symbolic_unary, :op<numify>')> } 14:01
+ token prefix:sym<+> { <sym> [ <?[@]> <O('%symbolic_unary, :op<elems>')> || <O('%symbolic_unary, :op<numify>')> ] }
won't that work?
timotimo that's pretty much the change i have ,except i put a <![@]> in there for good measure
so that failure won't be weird
arnsholt Does token have sigspace? I can't remember... 14:02
14:04 jnap joined 14:05 treehug88 joined
timotimo no, that's rule 14:05
with that change, tests still pass apparently. i'll try rakudo now.
arnsholt Right. In that case, that grammar tweak will fail on "+ @array" I think
timotimo 41 wallclock seconds on parrot, 14 on moar ♥ 14:06
that's not prefix:<+>, though
14:07 jnap left 14:10 jnap joined 14:12 AW3i left
moritz <?before \s*\@> <O('%symbolic_unary, :op<elems>')> || <!before \s*\@> <O('%symbolic_unary, :op<numify>')> 14:13
timotimo nqp: my @foo; say(+ @foo); 14:14
camelia nqp-moarvm, nqp-jvm, nqp-parrot: OUTPUT«0␤»
FROGGS nqp-p -e 'my @array; say(nqp::objprimspec(+@array))' # gives 1, which might be correct
timotimo oh, huh.
FROGGS moritz: prefix '+' has no whitespace afterwards
timotimo FROGGS: apparently it does, though 14:15
see my eval above
moritz FROGGS: the OPP eats that, iirc
FROGGS true, --target=ast shows it delegates to prefix:<+>
timotimo i'll add the whitespace then. 14:16
and now it would be nice to teach the optimizer that infix:<+> will turn a list into an int.
FROGGS but I would not add a <!before in the second branch when using a sequential alternation 14:17
timotimo OK
14:20 kaare_ left
moritz hopes that terminating LTM doens't cause any extra warts 14:21
14:21 kaleem joined
hoelzro o/ #perl6 14:22
FROGGS hi hoelzro
moritz does NQP need floating point numbers at all?
timotimo sometimes they're nice to have 14:23
FROGGS moritz: at least for --stagestats :o)
14:24 salv0 left
moritz FROGGS: well, that could be done with nqp::div_n ops etc. 14:27
FROGGS: so more on the "library" level than on the "language" level
FROGGS yeah 14:28
I don't need nums in nqp fwiw
at least not as a result of an expression
timotimo hm
14:34 Alula joined 14:41 jnap1 joined, salv0 joined
[Coke] I would like to start keeping something very vaguely resembling a project plan. We've talked about what has to be done for "6.0", I'd like to get a list, not to set anything in stone, but to help communicate to people looking for impactful things to help with, or people asking "how soon?". If you have anything you're working on that might go on such a thing, or know of tasks that are currently not claimed (like a syn that's been in draft for a while), feel fre 14:41
FROGGS diakopter created such a list 14:42
perhaps I am able to find it
14:42 jnap left
timotimo we still have the roadmap in rakudo's repository i believe 14:43
colomon item for list: ecosystem fully functional on all three backends.
timotimo now the optimizer turns 10 + @list into an add_i, too
FROGGS [Coke]: ethercalc.org/i982ooyabg.html
14:44 arnsholt left
FROGGS [Coke]: this list is not complete 14:44
[Coke]: if a synopsis is missing there, it is more likely that one needs to record issues rather than beeing implemented fully 14:45
timotimo FROGGS: do you have a date with that? 14:48
colomon FROGGS: It seems to me that list is too detailed … at least for what I'd imagined [Coke]++ meant
[Coke] yah. I could see drilling down to something like that, though. 14:49
colomon: you're a music guy: share and enjoy my son's arrangement of this song: www.youtube.com/watch?v=i6ZOAQU44es 14:50
FROGGS timotimo: around May 2013
colomon: true, but you only can record the details, and then put it in boxes labeled with a nice task description 14:51
timotimo oh, not terribly old then
colomon [Coke]: nice job he did! 14:53
14:53 ajr_ left 14:54 guru joined, guru is now known as Guest54850, Guest54850 is now known as ajr_
timotimo aye, not bad! 14:55
[Coke] colomon: danke. 14:57
pmichaud what's going on with issue #253 ("add env handling to parrot qx") ? 15:06
timotimo it seems like a non-negligible amount of people visit my blog, but none are interested in the LHF i'm posting 15:07
i wonder if i should continue digging for LHF to post on my blog
pmichaud good morning, #perl6
timotimo pmichaud: currently parrot will not propagate changes to %*ENV when using qx to execute stuff
pmichaud I understand the problem, I don't understand the proposed solutions.
timotimo reini made changes to parrot and rakudo to make it work again
pmichaud ...except that reini keeps posting failing tests
timotimo i haven't a clue what exactly is going on 15:08
that branch will have to wait another month anyway, though
FROGGS pmichaud: yes, there is some sort of confusion going on 15:09
I commented again right now
pmichaud also, reini's proposed patch (c87126f2) is going absolutely in the wrong direction... replacing nqp:: calls with pir:: calls 15:10
timotimo only inside the #?if parrot block, though
pmichaud I'm not ever a fan of putting more pir:: into the rakudo codebase 15:11
whether it's in an #?if parrot or not
beyond that, #?if parrot should be a rarity, not the norm
timotimo fair enough
FROGGS pmichaud: I think we can replace that with an nqp::open(... 'rp')
timotimo that block was there all along ;)
pmichaud well, all of these #?if thingies have been added since I was last active. I agree we'll need some... but they should be rare 15:12
FROGGS true
pmichaud having the ability to shell out with an environment in place should be an nqp-feature, not a rakudo one.
timotimo mhm
that makes sense, yeah
pmichaud or, at least, we should be solving it at the nqp level
FROGGS pmichaud: when the code in parrot works, we can make the nqp::openpipe op do all the work, like on moar and jvm 15:13
pmichaud switching VTABLE void set_pmc_keyed(...) to be METHOD void set_pmc_keyed() is almost certainly wrong. 15:14
for that matter, $pipe.set_keyed('env', $env); is almost certainly wrong also. 15:16
jnthn That looks odd to me too 15:17
15:17 hoverboard joined
FROGGS I am all ears 15:17
jnthn I thought nqp::openpipe was the thing we were putting this behind?
pmichaud jnthn: yes, I agree it should go somewhere like that
FROGGS jnthn: still, parrot should support the env handling first (unless we do that on our own like we do for shell/run)
pmichaud but I don't quite understand Parrot's patch either.... it seems odd to put the environment into the filehandle
jnthn Are we looking at something that never worked on Parrot here, or are we dealing with a regression of some kind? 15:18
FROGGS jnthn: env handling in qx{} never worked 15:19
jnthn OK.
But does on Moar/JVM?
FROGGS yes
tadzik problem solved :P
jnthn OK, then it's not a regression, so we needn't hurry on this ahead of release.
FROGGS true
pmichaud I'm more worried that a parrot misfeature is starting to turn into a rakudo misfeature
jnthn Yeah 15:20
My expectation is that April's Star release will suggest Moar as the default thing.
brrt pmichaud, i think i can think of a reason for that
pmichaud and the comments I'm seeing betray an unfamiliarity with Parrot internals
FROGGS jnthn: that seems reasonable
brrt just wait until i renember
:-)
FROGGS pmichaud: I am here! you can say that my comments prove that :o) 15:21
brrt nope, forgotten
FROGGS heh
pmichaud the whole pipe['env'] = env approach seems very wrong to me. it seems to me that opening a pipe in parrot should automatically accept the current environment... there shouldn't need to be an explicit pipe['env'] = env step 15:22
FROGGS pmichaud: the current env ist just a hash from rakudo
pmichaud FROGGS: that's an imprecise statement. 15:23
regardless, I reject the patch proposed in #253
FROGGS m: %*ENV<foo> = "bar" # we do not call setenv here
camelia ( no output )
pmichaud FROGGS: I understand what you're saying. 15:24
FROGGS (or putenv or whatever)
pmichaud FROGGS: the %*ENV hash is not the current environment... it's rakudo conception of what the current environment ought to be
so saying "the current env is just a hash from rakudo" is imprecise.
FROGGS so we need to pass %*ENV around, like here to a CreateProcess or spawn or so
15:24 xenoterracide joined
pmichaud FROGGS: yes, I know that. I'm the one that created that approach. 15:25
FROGGS pmichaud: %*ENV is our understanding of the current environment
15:25 arnsholt joined
pmichaud I'm saying that passing the environment through Parrot's FileHandle object is not the way we should be doing it. 15:25
FROGGS so I say "current environment" of the viewpoint of a Perl 6 hacker
moritz what does FileHandle have to do with it? 15:26
FROGGS yes, passing it to the thing that spawns the process might be saner
moritz: opening a pipe seems to be a method on a filehandle
moritz aaaarrg
FROGGS :o)
pmichaud again, imprecise
FROGGS meh 15:27
pmichaud well, not too bad
FROGGS meh!
jnthn As I understand it, nqp::openpipe should take the environment to use for the created process as one of its arguments. Exactly what that boils down to at the Parrot level is up for debate, but Rakudo should absolutely be using nqp::openpipe, meaning that it's NQP that should have any funky code-gen-y things.
15:27 kaare_ joined
pmichaud in Parrot, the way one opens a pipe is to create a FileHandle PMC, then invoke the 'open' method using 'rp' as one of the arguments 15:27
jnthn: I agree, except I suspect that at some level NQP ought to be environment-aware.
and should provide an API that rakudo could use that more smartly handles things like %*ENV 15:28
i.e., we need to push more of Rakudo's environment-handling down into NQP
jnthn pmichaud: It's not clear to me that'd be better, nor worse. :)
pmichaud: That is, "I think you might be right but I don't see it right away" :)
pmichaud jnthn: well, if spawning and openpipe are at the nqp level, then environment handling probably need to be there also
jnthn Since %*ENV is contextual in nature.
pmichaud well, it could be contextual in nqp, too :) 15:30
jnthn :)
Point.
15:30 rindolf joined
pmichaud anyway, we're in agreement that nqp needs to know more about the environment, whether nqp manages it or lets the hll manage it is open for discussion 15:30
jnthn I'm happy with such a factoring, anyway. Fully agree that we shouldn't be doing more pir::'ing in Rakudo.
And yes, #?if mechanism is an escape hatch. 15:31
The goal is to need them less.
JVM and Moar often have the same ideas about things.
pmichaud I just added my summary/comments to #253 15:34
FROGGS and since nqp should be the glue language for backend specificas, we should be able to factore stuff there (src/vm/...) 15:35
15:38 sqirrel_ joined
japhb__ FWIW, I believe this *is* a regression, I just don't know when it happened -- because perl6-bench has some code that depends on it, but since last June it's mostly been timotimo++ working on it, so he may have been lucky and not run into the bug before now. 15:46
FROGGS japhb: well, there was a time when we actually changed the environment of our own process 15:48
japhb__ FROGGS: Ah, interesting. That would certainly explain it. But it does mean it's a regression from the end-user's point of view, just not a regression in parrot per se. 15:50
FROGGS japhb__: yes, but not a regression happened since the last releases (I know, that does not help much) 15:51
japhb__ FROGGS: Sure -- as I said, I've about a 9 month window when it could have happened. :-/ 15:52
15:54 xinming__ joined
pmichaud if it's a regression, then that would indicate that rurban's parrot branch isn't needed. 15:56
unless something regressed in parrot also 15:57
that's pretty easy to check, actually.
I have to run an errand now but might write a short test program when I return.
15:57 xinming_ left
FROGGS it is a regression caused by rakudo 15:57
pmichaud so a parrot change isn't needed, right?
FROGGS because, IIRC, we used to have a proxy in %*ENV that does a setenv/putenv of our process 15:58
15:58 virtualsue left
FROGGS a parrot change is needed to cope with our %*ENV hash 15:58
pmichaud that's the wrong approach. 15:59
FROGGS either that or we move the entire openpipe code to nqp, as we did with run() and shell()
pmichaud I suspect that parrot's openpipe already properly handles the environment.
because it apparently used to do so. 16:00
FROGGS the question is: which environment
pmichaud *the* environment -- the process' environment.
FROGGS yeah, but we do not touch it enymore, which is a good thing
pmichaud the fact that rakudo has a separate %*ENV hash is not something that parrot should be considering.
in particular, let's just establish up front that Rakudo's %*ENV is not "the environment". 16:01
so, if I say "the environment", I'm not talking about %*ENV
(if I want to talk about %*ENV, I'll say "%*ENV")
FROGGS k
pmichaud anyway, I have to run my errand, I'm back in 20.
FROGGS k 16:02
japhb__
.oO( Do not try to unconfuse Professor Michaud. He is not confused. )
moritz so, nqp should call setenv before launching stuff?
japhb__ moritz: Yeah, I think we're generally agreed that nqp is the right layer for abstracting the VM's env handling. Unless I misunderstood the backlog. 16:04
FROGGS moritz: that might be a solution, except on machines where you are not allowed to setenv at all 16:06
moritz and there seem to be two general methods: setenv before doing stuff, and passing an env hash to operations
FROGGS moritz: I am currently searching for informations about that, since I hit that problem back in 2008 or so
segomos is there a good write up on how to write better modules using roles and "new" stuff now or is it still kind of in an infancy stage? i'm wanting to write something similar to dbix and i want to use cool new shiny stuff
16:11 lembark joined 16:12 denis_boyun left, sqirrel_ left 16:19 spider-mario joined, isBEKaml joined
pmichaud I propose we create nqp::setenvhash(hash) which updates the current process environment to match 'hash' 16:20
openpipe should continue to assume the current process environment is to be passed to the pipe process
16:22 kaleem left
rurban well, it should restore the old env on openpipe exit, I assume 16:23
and it should do nothing on empty cwd and env args 16:24
pmichaud gist.github.com/pmichaud/9645412 # parrot already properly passes its environment to child pipe subprocesses
16:28 slavik left
pmichaud I've updated Parrot issue #1065 with this gist. 16:28
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...ml?id=1065
16:29 [Sno] left 16:30 pmurias joined, pmurias left, pmurias joined
pmurias pmichaud: so shell shouldn't take the env as an argument anymore? 16:31
16:31 brrt left
pmichaud pmurias: I'm not sure about that coupling. It seems to me as though "set the environment" ought to be decoupled from shell 16:31
FROGGS pmurias: are you talking about nqp::shell or rakudo's shell() ? 16:32
pmichaud because there are many places where we might want to set the process environment -- it should be factored out.
TimToady seems like the general idea is that we either pass the proto-environment via dynamic %*ENV, or we pass it via an actual set environment, but rarely/never as an argument
pmichaud TimToady +1 16:33
FROGGS ahh, hmmm, for some reason I thought rakudo's shell() would take an optional env hash... but it does not
pmichaud passing it as an argument certainly has precedence in C, though
TimToady yup
pmichaud still, I think I prefer to keep "set environment" as its own separate operation
or, put another way: we should definitely have an nqp:: set environment operation. Given that, I'd tend to not want another way to do it via nqp::shell(). 16:34
(i.e., DRY)
I might reconsider if JVM expects an environment as an argument to however it does the shell equivalent 16:35
TimToady it's a bit like the boundary enforcement for NFG--we let our semantic model drift from what the world expects in the interests of better internal consistency, but then enforce the world's view at the edges of the process 16:36
pmichaud correct.
similarly, I don't know that nqp::shell() should have a separate $cwd argument; it should use the process cwd. But I'm less certain about that one.
TimToady there maybe trapdoor functions hiding there, if it ever tries to relocate itself relatively, or we can't find an absolute path that works 16:37
16:37 ajr_ left
TimToady the environment is a bit better behaved that way 16:38
FROGGS pmichaud: the jvm does not provide a way to chdir, so you have to keep what the user thinks is the cwd, and pass that along
pmurias pmichaud: JVM expecting an argument to the internal shell thing is the reason we pass the env as an argument currently
TimToady the chdir semantics get excercised at the boundary much more frequenetly than spawning does, I think 16:39
pmichaud FROGGS: I agree except with the "you have to pass that along" part. There are other ways to do that.
16:39 guru joined, guru is now known as Guest94161
pmichaud pmurias: what's the JVM command for a shell (i.e., where can I see the API/docs for it?) 16:39
16:39 Guest94161 is now known as ajr_
FROGGS pmichaud: "you have to feed it to the jvm when invoking a shell" to be more clear 16:39
pmichaud I'd like to see the actual code or API 16:40
FROGGS nqp/src/vm/jvm/runtime/org/perl6/nqp/runtime/Ops.java:877: public static long shell3(String cmd, String dir, SixModelObject envObj, ThreadContext tc) {
pmichaud looking.
pmurias it uses ProcessBuilder
FROGGS ProcessBuilder pb = new ProcessBuilder(args); 16:41
pb.directory(new File(dir));
Map<String, String> pbEnv = pb.environment();
[...]
TimToady well, if a VM requires it as an argument, then that's one of those process boundaries at which we have to do what the world expects 16:42
pmichaud okay, so it's not JVM that expects it as an argument -- that's an API we created.
TimToady but if, yeah, that
pmichaud I'm saying there's a different refactoring that doesn't require env/cwd as an argument. 16:43
16:43 SamuraiJack joined, jeffreykegler joined
pmichaud what JVM is doing seems much closer to POSIX than what we're using as an API 16:43
TimToady does ProcessBuild default to the current envinronment?
*er
pmichaud TimToady: it appears so
TimToady then really we can decide one way or the other based on which is more efficient or "safer" 16:44
pmichaud but ProcessBuilder appears to allow you to set the cwd of the new process
TimToady specified as a race condition^W^Wstring?
FROGGS so we should call nqp::getlexdyn('$*CWD') at that point instead of getting it passed to, say, shell()? Is that what we want? 16:45
pmichaud FROGGS: that's not exactly what I'm proposing, now. 16:46
*no
FROGGS k
pmichaud I'm proposing that creating a HLL shell operation should be some combination of: nqp::setcwd($*CWD); nqp::setenv(%*ENV); nqp::shell(cmd); 16:47
timotimo the good thing about enforcing environment and cwd at boundaries only is that we can have "my $*CWD" and friends work out naturally
pmurias nqp::setenv will have to be emulated on the jvm
pmichaud it has to be emulated pretty much anywhere 16:48
and it should be nqp::setenvhash(%*ENV)
(typo in my original -- "setenv" actually has another meaning to folks)
but it seems to me that nqp::setenvhash() is just
Map<String, String> pbEnv = pb.environment(); 16:49
pbEnv.clear();
pbEnv.putAll(env);
with some slight variation to handle the different forms that env might take
16:50 ajr_ left
pmurias but that sets the env variables for the new process but doesn't change the env variables in the current one 16:50
pmichaud it doesn't?
timotimo no, the pb will be recreated for each invocation of nqp::shell
pmichaud huh? 16:51
pmurias it seems to be impossible to change the current env variables
timotimo there has to be some internal variable that carries the data from nqp::setenvhash to nqp::shell
on the java side, that is
pmichaud that's not how I'm reading the code
isBEKaml pmichaud: that's from processbuilder. That's only if you're planning to execute any programs off the current shell. If you want to access env variables, you can use System.getenv("prop name")
timotimo i should look closer
FROGGS timotimo: that is what we currently have in the HLL :o)
isBEKaml pmichaud: That said, it's good to see you back here. :-)
pmichaud I'm looking at the code now, and the way that the environment is being set is by simply modifyinb pbEnv. I don't see anywhere that pbEnv is being linked up with the ProcessBuilder object. 16:52
isBEKaml pmichaud: Can you post link?
pmurias pbEnv = pb.environment()
timotimo ah 16:53
isBEKaml pmurias: pb --> ProcessBuilder
pmichaud github.com/perl6/nqp/blob/master/s....java#L931
timotimo now i see where your confusion comes from
pmichaud oh, I get it.
timotimo note that th ebeginning of the "private" spawn method is to create a new ProcessBuilder
pmichaud yes, I see the link now.
pbEnv is the pb's environment
okay.
isBEKaml pmichaud: yes. exactly.
FROGGS so we have exactly the same problem for cwd and env, which is not a bad thing 16:54
(on jvm)
pmichaud well, not necessarily
because pb.environment() defaults to the current process environment
FROGGS in kinda inherits from it, yes 16:55
if we had a way to change the current env...
timotimo sadly, my optimizer changes to nqp to make +@foo and 1 + @foo give integer operations caused S03-operators/identity.rakudo.moar to regress 16:56
FROGGS pmichaud: stackoverflow.com/questions/318239/...-from-java
pmichaud and yes, it appears that jvm doesn't provide a clean way to change the current process' environment. I'm not necessarily sad about that, but I do want to rethink the API a bit
16:57 hoverboard left 16:58 jeffreykegler left, myfree joined
pmichaud what's the JVM equivalent of exec(), ooc? 16:59
16:59 lembark left
pmichaud or is there no such animal? 17:00
isBEKaml awkward animals.
FROGGS www.javaworld.com/article/2071275/c...won-t.html 17:01
isBEKaml pmichaud: ProcessBuilder and Runtime.exec were meant for that. It turned out that inheriting environments isn't always quite feasible.
pmichaud: so you end up building your own set of properties to use and importing them into spawned processes. 17:02
pmichaud isBEKaml: in other words, there's not an exec() exact equivalent
i.e., one that replaces the current process with a new command
isBEKaml pmichaud: that's where we use Map<String, Object> objects.
masak Util++ # Parrot release mail quote
pmichaud isBEKaml: those all sound as though they're creating new processes, though 17:03
isBEKaml pmichaud: Nothing exactly same as system() or backticks. No.
pmichaud system() creates a new process. backticks create a new process. exec() replaces the current process image with a new image.
so, I'm not looking for a system() or backtick equivalent, I'm looking for an exec() one. 17:04
isBEKaml pmichaud: No, we don't have one on the JVM side of things.
dalek c: 8535ec3 | Mouq++ | lib/modules.pod:
Blindly copy module information from archived wiki

  (This will need to be updated)		 17:05
pmichaud okay. so that's something we'll only be able to emulate (badly) or declare is not supported
17:06 treehug8_ joined
isBEKaml pmichaud: well, if you understand that JVM by itself is a sandboxed environment, you'll see that it's meant to not execute off its current environment. So out goes exec() -- though I remember seeing something like that somewhere. Let me do some searching to see if there's one. 17:07
pmichaud isBEKaml: yes, I understand it -- just needed some confirmation that this is the case
exec() is the only place where I can think of that it'd be important to get the process environment to match %*ENV
although TimToday++ undoubtedly will know if there are others :)
timotimo and jvm doesn't have exec()
so we can probably safely emulate %*ENV for everything on jvm 17:08
pmichaud well, I think we have to emulate %*ENV everywhere
timotimo if someone choses to exec using NativeCall ...
they'll likely run into other problems :)
pmichaud well, via NativeCall we'd be able to modify the current process environment, yes?
(I'm not familiar with how NativeCall works either... I'm about a year out of date) 17:09
arnsholt jnthn: Could you try nqp/jastcompiler and Rakudo a whirl, BTW?
I suspect I've broken something, but I'm not entirely sure =)
17:09 treehug88 left
timotimo oh, yeah, with NativeCall the user probably could 17:09
wouldn't be terribly surprised if the JVM laid some more stones in our way for that, though >_>
masak feature request for Rakudo: make it impossible to build Rakudo on an outdated NQP or Moar. 17:10
timotimo masak: well, if you constantly Configure.pl like any sensible person ... ;)
pmichaud masak: ??? Doesn't rakudo already do that check?
I don't believe in making it "impossible". 17:11
I would be okay with having a check done at each 'make' invocation, instead of only at Configure.pl time 17:12
timotimo oh, whoops
FROGGS it only croaks if you reconfigure
err, yes
timotimo letting prefix:<+> look ahead for a @ can easily break down
for example in this line: return +@!BLOCKS[$i].symbol($symbol)
(this type doesn't support elems) 17:13
pmichaud doing a lookahead for syntax is almost certainly wrong
masak ok, "impossible" is a strong word.
timotimo aye.
masak I meant, used to be with Parrot the `make` invocation told me if the dependencies were too old, and refused to build.
timotimo should have realized that earlier :)
masak I'd like that.
timotimo masak: are you sure you're not thinking of having an older Makefile than Makefile-Parrot.in? 17:14
pmichaud we can easily come up with a build tool that checks dependencies
does Parrot no longer check the dependencies? 17:15
er
masak timotimo: I don't know the exact circumstances.
pmichaud do we no longer check the dependencies even for Parrot builds?
masak timotimo: but I know this: I git `git pull`, `make`, and then the build failed.
timotimo Configure.pl will loudly complain if the *_VERSION files are not met
17:15 SamuraiJack_ joined
timotimo but make alone will never do thtat 17:15
pmichaud timotimo: make used to do that
masak timotimo: it was *fixed* when I re-configured, but I think `make` should be clever enough to not waste time building something that won't build.
17:16 kaare__ joined, zakharyas left
masak yes, make used to do that. 17:16
it was nice.
timotimo oh, ok
pmichaud geez, so many absolutes in this channel now! "impossible" "never"
timotimo must have been before my time :P
masak pmichaud: :P
why *shouldn't* make do that?
isBEKaml pmichaud: we even have P6BP! :P
timotimo (which obviously doesn't mean "make will never do that" is wrong. because if it ever did, it did so before my time! which, you see ...)
masak isBEKaml: we do!? 17:17
isBEKaml masak: well, there you are. :P
17:17 SamuraiJack left
pmichaud maybe I should just crawl back under a rock for another year :) 17:17
masak no please stay
timotimo oh, btw, who am i going to see at the GPW?
pmichaud I'm still working out my schedule after having had my spring/summer completely jumbled around for a couple of weeks. 17:18
FROGGS <---
isBEKaml pmichaud: No -- please don't ! It's fun to see someone who says "OMG, where has the world gone to??!!" :D
17:19 kaare_ left
pmichaud well, I have to take a break for a bit... bbl 17:19
17:19 myfree left
FROGGS soup & 17:19
17:19 FROGGS left
[Coke] masak: yes, that should be doable. Please open a ticket. We can figure out that dependancy. 17:23
masak submits rakudobug 17:24
17:24 SamuraiJack_ left 17:25 denisboyun joined
[Coke] very excited to see progress on parrot #582 17:27
(Add 6model to Parrot)
isBEKaml [Coke]: how goes mojo port to perl6? :-) 17:31
17:39 pmurias left 17:43 isBEKaml left 17:46 guru joined, [Sno] joined, fhelmberger left, guru is now known as Guest23952, Guest23952 is now known as ajr_
[Coke] isBEKaml: work is interfering. haven't had a lot of cycles. 17:50
17:52 SamuraiJack_ joined
masak [Coke]: here you go: 🚲 17:53
[Coke] ooh, a diamond with a question mark.
sweet.
masak :P 17:54
masak .oO( clearly it's an empty rectangle, not a diamond with a question mark ) 17:55
17:55 SamuraiJack_ left 17:59 FROGGS joined 18:00 rylinaux left 18:01 rylinaux joined, rylinaux is now known as Guest10848 18:05 Guest10848 left 18:06 Ryan_ joined 18:13 Ryan_ is now known as rylinaux
[Coke] masak: as someone who is about to try to convince a PM to use stylized unicode numbers rather than going out to find an icon set of numbers, I sympathize. 18:14
timotimo a perl monger? 18:15
prime minister?
arnsholt Project Manager, probably 18:16
timotimo ah
that makes more sense
TimToady someone should rewrite irssi in Perl 6, and fix everything that's wrong with it :) 18:17
vendethiel Does somebody have a graph for perl6-jvm perfs when the JIT is hot ?
arnsholt I must confess I don't really have any things that really annoy me with irssi 18:18
TimToady yes, somebody does :)
vendethiel Trying to get somebody interested in perl6 :p.
TimToady arnsholt: well, not rendering characters that I have the fonts for, for one 18:19
timotimo vendethiel: t.h8.lv/p6bench/2014-03-09-every_perl.html
(but please mention that the microbenchmarks are not very good)
vendethiel timotimo++ # which ones are the least-micro ?
arnsholt TimToady: Oh, that does sound annoying. OTOH, it does sound like a bug in your terminal emulator rather than irssi =) 18:20
timotimo vendethiel: at the very very bottom
arnsholt rc-forestfire shows the JIT at work 18:21
TimToady arnsholt: yes, that's certainly possible
timotimo aye, the forestfire benchmark is a good one for seeing how the jit warms up
vendethiel arnsholt: yeah, I took this one :-)
arnsholt Rakudo doesn't do very well on it, but the graph for nqp-jvm is nice
TimToady perhaps the most important thing to point out is that the performance of Perl 6 will approach the performance of nqp as our optimizer gets better
arnsholt Starts out pretty slow, but as the JIT heats up it actually ends up faster than Perl 5 18:22
vendethiel timotimo: need to do it again with jdk 1.8 :-)
timotimo will have to, yeah
also point out by how much the rakudos and nqps get faster between the regular and _native variants of some of the microbenchmarks
TimToady in particular, forest-fire uses 2-D arrays that will be much, much faster someday soon
once we have shaped arrays of known, preallocated sizes 18:23
timotimo and when the type specializer (perhaps) starts removing bounds checks
TimToady of native values
timotimo when we have very fast, shaped native arrays, that'll give us something nice for Modern OpenGL™ 18:24
TimToady not so much the bounds checks as it is all the redispatch to at_pos, and the checking for element existence for autoviv
timotimo ah, yes, shaped arrays of natives won't ever do autoviv
right?
TimToady no, but we need to teach the compiler about lvalues, which it's pretty ignorant of currently 18:25
same problem as doing $a++ on a native, basically
timotimo ah, that one
that keeps cropping up more and more often 18:26
18:26 hoverboard joined
TimToady we need something like C's pointer to native types to be able to talk about lvalues 18:26
though one can fake it up to a point 18:27
and indeed JVM kinda forces you to fake it
vendethiel Also -- I'm trying to get the post of perl6advent that adds custom operators, like the exponential one, describes a macro `/* */` but my search do'nt bring anything up
TimToady so our "pointers" are likely to be array/hash refs with precalculated sublocations when we can't use C pionters 18:28
*oi
vendethiel: need a link to understand what yer askin' 18:29
vendethiel TimToady: I need a link myself, actually ;-) 18:30
TimToady: There's a really cool perl6advent post that got me interested in perl6 in the first place, creating custom operators to deal with error percents 18:31
and some other mathematical notations (basically introduces sub infix:...) and also has, IIRc, an example of a macro that could parse `/* */` comments 18:32
timotimo perlgeek.de/blog-en/perl-5-to-6/13-...ators.html ?
oh, actually notd
18:33 bjz left
vendethiel timotimo: I do believe it was a perl6advent blog post 18:33
TimToady actually, to do /* */ correctly requires not a macro, but a tweak to the definition of whitespace
vendethiel TimToady: it was maybe a "fake" example in that it didn't actually really work, but demonstrated a possibility 18:34
TimToady our current macros can only replace a term, really 18:35
once we generalize that to replacing any syntactic category, then you could put a macro into the comment: space instead
timotimo perl6advent.wordpress.com/2012/12/1...is-futile/ ? 18:36
TimToady it's vaguely possible it could be forced to work currently somehow
vendethiel timotimo++ # thanks ! I believe that's it !
Definitely :-) 18:38
I'll try to look for the macro and stuff one later, though.
TimToady m: macro comment:</* */> () { '' }; say 1 /* phooey */ + 41
camelia rakudo-moar 01d16c: OUTPUT«===SORRY!=== Error while compiling /tmp/RURkzDIp31␤Cannot add tokens of category 'comment'␤at /tmp/RURkzDIp31:1␤------> macro comment:</* */> ⏏() { '' }; say 1 /* phooey */ + 41␤ expecting any of:␤ …»
TimToady yeah, not supported yet
or augment slang might work someday too 18:39
but that all probably depends more on the HLLization of the P6 parser that I mentioned yesterday 18:40
we can't really claim that all other languages are just dialects of Perl 6 till that happens :)
vendethiel .oO(Perl6 is getting slower yet) :p
TimToady well, we can pessimize it lazily at the point where you try to make it do something fancy :) 18:42
it would be fun to try to write a yacc grammar for the subset of Perl 6 that people usually use, and switch out to the recursive descent when yacc isn't suitable, for instance 18:44
there are also probably ways to cheat on how often we have to generate a new lexer, which slows things down when you define new operators 18:46
or to remember the lexer you generated last time you were here 18:47
at some point it starts looking more like an incremental compiler if you do that 18:48
but obviously there's still lots of room for improvement, and we hope that we're not locking anything into the design that precludes such improvements 18:50
18:52 tgt joined
TimToady another way to cheat is to keep short exception lists in grammatical categories, so, for instance, a few unambiguous postfixes might just be added to an exception list rather than being immediately compiled into a new lexer, especially if it's easy to recognize the exceptions based, say, on their first character 18:55
there are no other postfixes that start with ! for instance, so it's really easy to look up a ! postfix as an exception to the current postfix lexer 18:56
likewise most unicode symbols will be unique in the first char
so basically you install a 1-level trie in front of the lexer for the category in question 18:57
18:57 logie_ left
TimToady you could add all the operators in that advent post without ever recalculating a new lexer 18:59
PerlJam hmm. 19:01
TimToady or another way to view it would be to say that the lexer is really allowed to decide when it needs to redo the FSA
PerlJam TimToady: you seem to be feeling much better; is this true? :) 19:02
TimToady or perhaps more importantly, the distinction might be whether the lexer is allowed to mutate itself in place 19:03
PerlJam: well, considering I hacked in all the Unicode stuff while I was having pneumonia, I dunno how you'd tell the difference :)
PerlJam volumetric analysis of your text + cogency ;) 19:04
nwc10 pmichaud: this blog post about how JRuby fights the JVM to do fork, exec, etc might be relevant: blog.headius.com/2013/06/the-pain-o...ocess.html 19:05
timotimo: I'm going to the German Perl Workshop
TimToady { Language A .... mutate-to-lang(B) .... Language B }; as long as there's no need for Language B to ever refer to Language A as a subset, you can just mutate the current Language A lexer into Language B 19:06
and you just throw away the mutated lexer at the end of the block anyway
but that's assuming Langauge A didn't come from outside the block
so you have to clone the langauge at least once, but after that you can just mutate in place 19:07
s:g/gauge/guage/ 19:08
timotimo nwc10: cool :) 19:09
TimToady you can do the same sort of cheating with Cursors when you know the old Cursor isn't going to be reused to backtrack with, for instance 19:10
nwc10 I assume that FROGGS is, given that he's speaking: act.yapc.eu/gpw2014/talk/5252
FROGGS nwc10: \o/ 19:11
nwc10 wendy and Liz are "confirmed" according to act 19:12
19:13 darutoko left 19:25 kivutar joined
TimToady
.oO(The JVM interprets sanity as damage and routes around it.)
 19:31
geekosaur ain't that the truth 19:34
19:35 Rotwang joined
lizmat nwc10: indeed 19:50
Sex and Violence: Social and Technical Lessons from the Perl 6 Project ( www.linux-gull.ch/GestionActiviteDo...40401.html ) 19:51
another TheDamian presentation
nwc10 PHP 353, Perl 12. We lose 19:52
www.cvedetails.com/product/128/PHP-...ndor_id=74 vs www.cvedetails.com/product/135/Larr...ndor_id=81
that really is quite a spectacular difference. :-/
19:54 brrt joined
PerlJam Why does the perl data stop at 2007? 19:57
nwc10 yes, good question.
OK, 353 vs 15: www.cvedetails.com/product/13879/Pe...or_id=1885 19:58
The[ir] data is crap
PerlJam clearly 19:59
But if people are going to use TIOBE to say that Perl is dying, we can use this to say Perl is secure ;)
20:02 btyler_ joined
masak .oO( Eskimoes interpret fifty words for snow as damage, and route around them ) 20:03
20:06 xinming__ left, btyler left
rjbs PHP interprets security as damage and routes around it? 20:07
20:07 FROGGS_ joined, FROGGS left
lizmat
.oO( 353 shades of grey? )
 20:08
20:08 treehug8_ left, treehug88 joined
nwc10 I wonder whether their data cleaned up could be made into a graph that superficially looks like usage 20:10
(with Perl really low)
but then the punchline is that it's CVEs
rjbs haha 20:11
brrt easy shots at php again :-)
i think the CVE-count is pretty low for perl6 as well
rjbs of course, the counter-snipe is that it IS about usage
"perl has so few CVEs because nobody targets it" 20:12
nwc10 Python is split between www.cvedetails.com/product/18230/Py...r_id=10210 and www.cvedetails.com/product/2147/Pyt...or_id=1238
rjbs bunk, but would get airplay
nwc10 Python and Ruby are (also) much lower than PHP
brrt php has lots of bugs, surprising no one
nwc10 so would need some data cleanup to try to be fair
PHP manages to have CVEs on its CVE fixes 20:13
rjbs hah
nwc10 and create CVEs due to shipping a version with failing tests
rjbs strikes me as useful to discredit PHP, which nobody needs to do
publicizing it would be a bit like kicking a guy while he's down
vendethiel php contributors are known not to update related tests, though ... At least they used to 20:14
nwc10 yes, agree
PerlJam Better that we not bash on any other language, but rather tout the merits of the language(s) we love. 20:15
20:15 xinming_ joined
nwc10 yes, agree. 20:16
This was partly a thought experiment, and partly shock at *how* big the difference is
masak everything about PHP is depressing. even its lack of security. 20:17
it's as if someone deliberately designed a language for people who just plain don't care. 20:18
nwc10 implemented by people who don't care (too lazy to look up "Paamayim Nekudotayim") with code review that didn't care either 20:20
perigrin masak: actually I think basedon the quotes I've heard ... that *is* the design principle behind PHP. 20:21
PerlJam That's close to factual I think.
Didn't petdance offer to help clean up the PHP codebase (by providing patches) and got a "thanks, but no thanks" ? 20:22
huf they did basically fire the one guy who seemed to care about security :D
years and years ago
perigrin "I really don't like programming. I built this tool to program less so that I could just reuse code." - Rasmus 20:23
20:23 bjz joined
masak nwc10: "Although it has been confusing to many developers who do not speak Hebrew, it is still being used in PHP 5" -- what. the. phaamayim. 20:24
to be fair, that's the first thing about PHP where I crack a reluctant smile, rather than putting my head in my hands or banging it against a wall. 20:25
huf that's honestly the least of the problems, yeah
masak it's just so... inept.
huf not even that, it's just a quirk
"0e8217321" == "0e171717" is the really inept thing.
much more "representative" of a fuckup than the ivrit token name 20:26
masak yep.
that puts me back to "depressed" again. 20:27
huf and this isnt even an implementation bug, it's a "spec" bug :D js has similar ones.
PerlJam So ... how about those awesome feed operators in Perl 6? And the neat-o asynchrony?
huf probably not *quite* as bad, but if you treat types as suggestions AND have operator overloading, you're gonna have these
masak PerlJam: yeah. we should all build awesome indespensible apps on top of that. 20:28
vendethiel huf: don't make me more sad hahaha 20:29
huf vendethiel: sorry, i'm a cloud of depression.
20:29 stevan_ left 20:32 stevan_ joined 20:34 stevan_ left 20:45 kbaker_ left, kbaker_ joined 20:55 BenGoldberg joined 20:56 rindolf left 21:01 dayangkun left 21:02 bjz left 21:04 kbaker_ left, kbaker_ joined 21:09 treehug88 left 21:17 dayangkun joined 21:32 kaare__ is now known as kaare_
moritz std: 21:35
eeks
std: my Int|Str $foo
camelia std 09dda5b: OUTPUT«===SORRY!===␤Multiple prefix constraints not yet supported at /tmp/35yFfqrnUQ line 1:␤------> my Int⏏|Str $foo␤Malformed my at /tmp/35yFfqrnUQ line 1:␤------> my Int⏏|Str $foo␤ expecting any…»
lizmat std: my IntStr $foo
camelia std 09dda5b: OUTPUT«===SORRY!===␤In my declaration, typename 'IntStr' must be predeclared (or marked as declarative with :: prefix) at /tmp/QfVsFl253U line 1:␤------> my IntStr⏏ $foo␤Malformed my at /tmp/QfVsFl253U line 1:␤------> …»
lizmat weren't there some dual types specced but not implemented yet ? 21:36
vendethiel std: subset IntStr where Int | Str; my IntStr $a;
camelia std 09dda5b: OUTPUT«ok 00:01 124m␤»
vendethiel lizmat: Int|Str and Int&Str are speced 21:37
brrt folks, help me out here 21:39
somebody claims that to sum numbers from 1 to N, you'd need O(N^2) 21:40
not sure of space or memory, but whatever
i can sum numbers from 1 to n in constant time
21:40 ajr_ left, kaare_ left
tadzik somebody is wrong :) 21:41
you can do that in constant time and space alright
brrt i'd think so
maybe he means the solution is on the order of n^2, which is true 21:42
but that is not the right way to use O
moritz tadzik: that's not true
21:42 kbaker_ left
moritz tadzik: the result needs O(log(N)) bytes, which you can't compute in O(1) 21:43
tadzik oh, heh
brrt .... such truth
brrt tends to take integer arithmetic as constant-time-always
especially with 64 bit
moritz I'm not sure about the time complexity of multiplying numbers, but it's not O(1) 21:44
brrt 64 bit ought to be enough for everybody
tadzik yeah, I wanted to say "unless you jump into bigints", but found that to be bikeshedding :Y
brrt you can do multiplication in O(1) if you have enough transistors
moritz well, that kind of bikeshedding is what complexity theory is all about
PerlJam masak: He didn't say O(1), he said O(C) ;)
er, moritz
moritz PerlJam: and C = log(N) ? :-)
tadzik I assumed int operations being all O(1) in a hypothetical tool used to compute that
which is not true in a general case :) 21:45
PerlJam If you have a quantum computer, all algorithms are O(C) :)
brrt i still don't actually believe that
flussence rakudo-m: gist.github.com/flussence/3007551
does... that not work? 21:46
perl6: gist.github.com/flussence/3007551
huh.
it used to.
camelia rakudo-parrot 01d16c, rakudo-jvm 01d16c, rakudo-moar 01d16c, niecza v24-109-g48a8de3: OUTPUT«125000250000␤»
flussence -ENOPATIENCE
moritz ECAMELIASLOW 21:47
brrt sometimes forgets p6 can be quite nice and readable 21:48
dalek ecs: 57750d0 | (Elizabeth Mattijsen)++ | S22-package-format.pod:
Some more clarifications, especially with provides
 21:49
lue Int|Str doesn't exist anymore IIRC
moritz correct
it never really existed
just as a vague, unimplementable idea in the specs
geekosaur the specs used to include a lot of pie in the sky ideas 21:50
they probably still do in places
PerlJam b: say "hi" 21:53
camelia b : OUTPUT«sh: 1: install/bin/perl6: not found␤»
PerlJam IS that meant to still work?
tadzik what's b?
PerlJam beta 21:54
tadzik oh
right, pre-nom :)
PerlJam and old incarnation of rakudo
21:58 spider-mario left
moritz bejing, actually 21:58
PerlJam oh. that's right. 21:59
moritz it's meant to work, but I don't think I can be bothered to fix it
PerlJam
.oO( so many old things )
22:07 woolfy left 22:11 woolfy joined
lue heh, on a wiki page listing reasons various languages suck, for Perl: "what's with all the $,@,%,& things in front of variables? It takes a lot of effort to type those redundancies every single time" 22:12
22:12 dmol left 22:15 clsn left 22:16 stevan_ joined
perigrin PerlJam: are you commenting on the code? or the channel's participants ... 22:18
22:19 bjz joined 22:20 kivutar left
brrt off 22:27
22:27 brrt left 22:32 stevan_ left 22:33 stevan_ joined 22:40 treehug88 joined, treehug88 left 22:41 treehug88 joined, stevan_ left 22:48 denisboyun left 22:49 FROGGS_ left 22:54 btyler_ left 22:57 rurban1 left 22:58 kurahaupo left 23:03 treehug88 left 23:04 Rotwang left 23:06 hoverboard left, skids joined
masak moritz++ # "You can't just slap on union types onto an existing type system, and have it magically work out." 23:10
as I mentioned the other day, sum types are weird.
they basically represent runtime indecision.
(whereas product types only represent combination) 23:11
'night, #perl6
23:15 tgt left
japhb__ .tell jnthn Segfault and oddness in perl6-debug-m; see irclog.perlgeek.de/moarvm/2014-03-19#i_8463775 23:25
yoleaux japhb__: I'll pass your message to jnthn.
23:37 ilogger2 joined, ChanServ sets mode: +v ilogger2
jnthn Yeah. I'll ahve to get 'em to have a look. 23:37
japhb__ Depending on the quality of your telco/ISP, they can run tests relatively easily that can tell them not just what the failure is likely to be, but where it is (how many virtual meters from the CO equipment, where the "virtual" part comes from equipment latency masquerading as distance) 23:40
23:41 pecastro joined
japhb__ One of the DSL testers at my previous company used to say "If the answer is less than 100 feet, it's inside the CO building -- which generally means the telco accidentally disconnected or misconnected the wires." 23:42
Happens surprisingly often.
23:44 slavik joined
jnthn Yeah, I had an incident like that happen to be back in Slovakia. Came to fix/install something for somebody else, and accidentally knocked mine out while there. 23:46
23:54 stevan_ joined 23:57 guest93 joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!