»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.
00:02 sidus left 00:11 arcterus_ left
BenGoldberg rn: say 1...87 00:13
camelia rakudo 5d9f71, niecza v24-88-g1f87209: OUTPUT«1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87␤»
BenGoldberg rn: say 1...88
camelia niecza v24-88-g1f87209: OUTPUT«1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88␤»…
..rakudo 5d9f71: OUTPUT«1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88␤»
BenGoldberg Camelia bug? 00:14
timotimo maybe
geekosaur someone did an rnp: earlier today and only got the rakudo and niecza responses
BenGoldberg rnp: say 1...88 00:15
camelia pugs: OUTPUT«*** ␤ Unexpected "88"␤ expecting operator, ":", "," or term postfix␤ at /tmp/n8ncdXMzhN line 1, column 9␤»
..niecza v24-88-g1f87209: OUTPUT«1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88␤»…
..rakudo 5d9f71: OUTPUT«1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88␤»
BenGoldberg rnp: print 1..88
camelia rakudo 5d9f71, niecza v24-88-g1f87209, pugs: OUTPUT«12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788»
BenGoldberg rnp: print 1..200
camelia rakudo 5d9f71: OUTPUT«1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241…
..niecza v24-88-g1f87209: OUTPUT«1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211…
..pugs: OUTPUT«1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271…
00:16 xinming left 00:18 xinming joined
BenGoldberg This might be a bit of an odd idea, but how hard would to be to create a "lazy" hash? 00:22
timotimo quite.
if you access a key, you cannot determine if it'll never be there or if you just haven't looked long enough 00:23
unless the set of keys can be known before
BenGoldberg Obviously, it wouldn't always be useful :) 00:24
But, my %h = LazyHash.new(1..Inf); .say for %h.keys; would work 00:25
timotimo would that give 1 => 2, 3 => 4, 5 => 6, ...? 00:27
BenGoldberg Right
timotimo that would be easier to solve with an accessor that generates the values on-demand if they're missing from the "backing store" hash 00:28
00:31 census joined 00:33 dmol left 00:45 census left 01:07 btyler joined 01:32 coax75ohm joined 01:39 btyler left 01:40 lizmat left, lizmat joined 01:45 cognominal left 01:52 xinming left 01:56 xinming joined 02:04 raiph left 02:23 PacoAir left 02:33 coax75ohm left 02:45 arcterus joined 02:50 Aww left 02:51 Aww joined 03:14 cognominal joined 03:17 stevan_ left 03:24 stevan_ joined 03:25 stevan_ left 03:26 stevan_ joined 03:31 BenGoldberg left 03:38 mstwu00 left
sorear good * #perl6 03:43
03:45 preflex left, preflex_ joined, ChanServ sets mode: +v preflex_, preflex_ is now known as preflex 03:52 cognominal left 03:54 strider42 joined
labster o/ sorear 03:57
sorear o/ labster 04:05
how goes
labster pretty good, busy with $work lately, trying to figure out exactly how to balance my time. 04:06
.oO (Another day, another user privilege escalation bug)
sorear in p6, or in $work? 04:08
what is $work these days?
last I heard you were looking
labster perceptyx.com 04:09
A lot of legacy Perl code driving their website, which is most of what they do. Which is employee surveys. The website is laden with managementspeak, so that might be confusing. 04:10
not in p6, RESTRICTED.setting is kind of a joke, we all know it doesn't really stop anything except accidental writes. 04:11
sorear 4? 5.00x? 5.LATEST? 6?
labster 5.12.4 04:12
flussence
.oO( my last job had stuff running on 5.006... )
 04:13
labster which is pretty good compared to other places I'd interviewed...
sorear by 5.00x I meant pre-vstring perls
flussence I took that to mean "not 5.01x" :) 04:14
sorear still isn't quite sure whether 21:06 < labster> .oO (Another day, another user privilege escalation bug) refers to the known sieveness of RESTRICTED.setting or stuff with perceptyx
and heya flussence
labster No, not in Perl 6.
flussence I've not been paying enough attention to stuff here lately, did I miss anything exciting? 04:15
labster Does P6 even have a concept of user privilege?
At some point in the future, we ought to consider making RESTRICTED.setting into a proper sandbox, but that seems pretty low priority to me. 04:16
flussence I'd like to see something using libseccomp, but that's a nasty non-portable thing... 04:20
dalek kudo-star-daily: 459e93b | coke++ | log/ (5 files):
today (automated commit)
 04:27
kudo-star-daily: a7c24f9 | coke++ | log/ (5 files):
today (automated commit)
rl6-roast-data: ee9c2ab | coke++ | / (5 files):
today (automated commit)
 04:29
04:31 lizmat left 04:33 lizmat joined
labster coke++ gaining karma 04:34
moritz \o 04:46
sorear o/ moritz 04:48
flussence: me neither
05:11 mstwu00 joined 05:12 mstwu00 left 05:14 fridim_ left 05:23 lue left 05:26 strider42 left 05:29 kaare_ joined 05:35 lue joined 05:49 FROGGS joined 05:54 FROGGS left 06:00 FROGGS joined 06:01 ilbot_nextgen joined
moritz idly wonders why pressing the home key in the browser usually doesn't bring me to the top of the document 06:02
given the name, it should take me ~ :-) 06:03
.oO( take me ~, country road )
 06:04
06:07 FROGGS left
diakopter moritz: 0_0 06:13
... "west virginia..."
decommunistize & bbiab 06:14
06:26 arcterus left 06:28 arcterus joined 06:31 crab2313 joined 06:56 ilbot_nextgen left 07:01 ecocode joined 07:05 ecocode left 07:08 FROGGS joined 07:28 FROGGS left 07:58 moritz is now known as moritz_, moritz_ is now known as moritz 08:08 Guest87971 is now known as ponbiki 08:41 arcterus left 08:42 dmol joined
masak mornin', #perl6 08:53
08:57 sidus joined
moritz \o masak 09:01
masak reviews t3 solutions 09:17
moritz \o/
09:26 FROGGS[mobile] joined
jnthn morning, all 09:32
moritz \o jnthn
nwc10 [A 09:33
jnthn tries to parse nwc10's greeting as some kind of drawing 09:36
nwc10 it's coincidence. And irssi being a bit daft
in that, if the characters ESC, [ and A arrive with the wrong timing, it *doesn't* interpret them as cursor up 09:37
masak wow, my guess was "cursor up"! 09:40
clearly I've spent too much time with computers.
09:45 spider-mario joined 09:55 rindolf joined
jnthn Well, the [ coulda been "on the phone"... :) 09:56
masak jnthn: knowyourmeme.com/memes/sidetalking-n-gage :) 09:59
FROGGS[mobile] I get a 503 10:06
10:09 arcterus joined 10:17 arcterus left 10:21 ecocode joined
jnthn github.com/rakudo/rakudo/commit/29...c482003201 # looks odd, the $?FILE and $?LINE refer to the setting, not current compile position.. 10:21
10:29 iSlug joined
jnthn Hm, it seems at some point some change hurt the optimizer's ability to inline stuff, even on Parrot. 10:40
11:08 BenGoldberg joined 11:10 arcterus joined 11:14 arcterus left
FROGGS[mobile] jnthn: how do you see that? 11:14
jnthn FROGGS[mobile]: Looking at the AST from --target=optimize 11:15
Though I've hunted it down some more now
I was wondering why infix:<+> doesn't get inlined
Turns out its proto is:
proto infix:<+>($a?, $b?) is pure { * }
And the compile time bind analysis doesn't know what to do with optionals. 11:16
FROGGS[mobile]: Why was 1541771 in nqp needed? 11:29
11:29 kaare_ left 11:31 iSlug left 11:33 rindolf left 11:49 pmurias joined
pmurias jnthn: while compiling QAST, if I encounter a QAST::Block how should I determine if I should generate the block or if the block is only refered to? 11:50
11:51 FROGGS[mobile] left
pmurias jnthn: like if it's mentioned in nqp::setcodeobj it's only used as a reference 11:52
jnthn pmurias: References are always through a QAST::BVal 11:53
The QAST::Block should really only appear in the tree once.
uh, aside from any QAST::BVal references. 11:54
And those shouldn't compile the block, just refer to it
pmurias jnthn: so a QAST::BVal always should refer to the block? 11:55
jnthn pmurias: Yes 11:56
pmurias: Anything mentioned inside a QAST::BVal should appear somewhere else in the tree
pmurias thanks 11:57
dalek kudo/nom: a04ebf2 | jnthn++ | src/Perl6/Optimizer.nqp:
Some optimizer code cleanups, commenting.
 11:59
kudo/nom: c2ef115 | jnthn++ | src/vm/parrot/guts/bind.c:
Compile-time analysis of sigs with opt params.

This means the optimizer doesn't trip up on protos like ($a?, $b?), meaning we can inline +, < and so forth for natives again.
colomon oh, lizmat++ will be happy 12:00
jnthn++ # back at it 12:01
dalek ast: 585d3d9 | jnthn++ | S02-types/nil.t:
todo => skip, due to stricter compile time checks.
 12:02
ast: 6f39558 | jnthn++ | S32-io/IO-Socket-INET.bat:
Make socket tests run less annoyingly on Win32.

They could pop up windows before; this prevents it.
jnthn I broke her streak? :)
12:07 dmol left
timotimo i thought more about optimizing perl6 last night and the massive dynamicism that can take place is kind of frustrating. a type restriction on "Foo" hardly means a thing when you can freely derive from Foo after the optimizer had a chance to do things, so even if we have a "link-time" optimization in place, it still can't look at the whole program and say "this Foo class here is never derived anywhere", because what if MOP? 12:41
sigh. i know so pitifully little about optimising :|
masak timotimo: it's much easier to optimize at the "lexical lookup" end of things, rather than the OO/MOP end. 12:49
jnthn Well, also keep in mind that a 6model-aware JIT will be able to have some fun too 12:50
timotimo yeah, this is probably JIT territory 12:51
12:51 rindolf joined
timotimo speaking of OO/MOP ... recently i mused about an optimization that could perhaps turn the creation of "super simple" objects into a super fast low-level op, so that creating an object without custom BUILD or new would be just as simple as creating a parcel or hash or something; someone mentioned that Cursor (even in nqp) could benefit from that, because cursors keep being created and destroyed all over the place. thoughts? 12:52
jnthn We already did customize Cursor creation quite a bit
timotimo ah, that's good then 12:53
although every optimization that i think of that then gets shot down as "already done" makes me a tiny bit sad, too. as in "aaw, we can't make that any faster this way? :("
12:56 xinming left 12:58 xinming joined 12:59 pupoque_ is now known as euge 13:06 euge is now known as eugene, eugene is now known as eugenius, eugenius is now known as eugeniuss, eugeniuss left
lizmat does a git pull and does not see "branch nom is up to date" :-) 13:12
jnthn lizmat: Figured you'd done enough streaking :) 13:14
lizmat jnthn++
I wouldn't have minded so much if I hadn't been blocking on so many things 13:15
jnthn: if you have time, I'd like your ear on a couple of things 13:16
jnthn lizmat: Sure.
I'm not yet really caught up on all that happend while I was gone, besides reading the GlitchMr++ post :)
lizmat adding the $.slash attribute in Variable, and passing $/ from nqp into it, how bad is that ? 13:17
I needed it to be able to get proper compile time errors
13:17 coax75ohm joined
jnthn Sounds odd 13:17
lizmat is there another way of getting at $/ in Variable, so I can use it to throw compile time exceptions properly ? 13:18
jnthn The proper way is to:
lizmat submethod throw ( |c ) { $*W.throw( self.slash, |c )}
jnthn 1) Have the exception type do X::Comp
lizmat check 13:19
jnthn oh, hm :)
jnthn sees what lizmat means...
13:19 gesh joined
jnthn hmm 13:20
Do we do treat applications through $*W.apply_trait?
'cus it does it through ex-handle
That in turn goes through rethrow method in world, which fills in the file and line
That's the "right way" to do it 13:21
And means $/ doesn't need to be passed
(into the Variable object)
lizmat it is currently in Actions.declare_variable
*not* through apply_trait 13:22
13:22 gesh left
lizmat tries to adapt code to use apply_trait 13:25
jnthn yeah, but wait
where does the $applier come from?
lizmat traitlist -> $_.ast 13:26
jnthn Right, which is a closure that does invoke the thing through apply_trait
lizmat /
jnthn will is doing it also 13:27
3664 or so
lizmat well, perhaps, I don't see that, but in either way, without my hack, it wouldn't show the line in which the error occurred
which raises one other point: exceptions thrown from trait_mod will, are not rrethrown 13:28
jnthn oh? 13:29
That's very odd
lizmat frustrating as hell more like it :-)
jnthn Given it looks like it *should* be working
Heck, it even tries to mix in X::Comp
lizmat well, I'll give it another go
jnthn Which exception type, btw? 13:30
ruoso Hmmm... The fact that ThreadContext pre-initializes a few exceptions make debugging slightly harder, because the stack trace will point to the stack that created the exception, not the point that threw it... 13:31
lizmat X::Comp::NYI
ruoso trying to fix t/spec/S04-statements/loop.t
benabik Did the advent calendar start really early this year? *stares at Planet Perl Six* 13:32
jnthn ruoso: Yeah, not doing that gets insanely costly
lizmat wonders what's wrong with loop.t
ruoso gather/take seem to be broken 13:33
jnthn Oh, is it the same failure mode as the failing gather/take tests?
benabik Huh. Showed up in my feed reader, but it's not on the page. Strange.
lizmat is doing the first build of the day and hasn't seen any problems yet
jnthn lizmat: Failure is on JVM
ruoso jnthn: yep 13:34
13:38 dmol joined
lizmat jnthn: it does throw again from within the trait_mod spec 13:39
jnthn Hm. I got the trail bind stuff ported to JVM. It seems to work, it choses the candidate at compile time, but seems unwilling to inline. Oddness...
lizmat $ perl6 -e 'my $a will init { say "Hello" }'
===SORRY!===
Variable trait "will init {...}" not yet implemented. Sorry.
at -e:1
------>
jnthn lizmat: That has a line number, no? 13:40
lizmat yes, but want more:
$ perl6 -e 'my $a is foo'
===SORRY!===
Can't use unknown trait 'is foo' in a variable declaration.
at -e:1
------> my $a is foo⏏<EOL>
expecting any of:
TypeObject
default
dynamic
jnthn The expecting any of comes through highexpect though... :) 13:41
lizmat the part after the ------> is only possibly if I have $/
jnthn That sounds fixable.
lizmat it's about "my $a is foo⏏<EOL>"
jnthn Yeah, that bit we should be able to manage
SET_FILE_LINE should also learn to take the stuff to put in the ---> line I guess
lizmat: See method locprepost 13:42
lizmat looking
13:44 xinming left 13:46 xinming joined 13:47 dmol1 joined, dmol left
jnthn oh...I know why it won't inline... 13:52
13:54 telex left 13:56 telex joined
jnthn nqp: sub foo(:$a) { say($a) }; foo(:!a) 13:57
camelia nqp: OUTPUT«0␤»
moritz and named params don't inline? 14:08
jnthn moritz: Correct, but there I was just checking if NQP supports :!a syntax :) 14:11
yay, think I've got it to do it 14:12
14:19 birdwindupbird joined 14:20 leont joined
lizmat jnthn: trying to remove Variable.slash as a first step 14:33
but it appears I also need that for registering global phasers, such as CHECK, INIT, END 14:34
suggestions?
currently: $*W.add_phaser($v.slash, 'CHECK', $block) 14:35
14:37 Psyche^_ joined
jnthn Doing stuff with $*W in trait mods sounds...odd 14:38
What on earth does "will check" even mean?
I mean, what's it operating on... 14:39
lizmat it's basically a CHECK block, with $_ as the variable on which it is defined
jnthn yeah. hm.
14:39 FROGGS joined 14:41 Psyche^ left
lizmat jnthn: there are only 2 add_phaser methods, one on Block, and on in WOrld 14:41
FROGGS jnthn: re github.com/perl6/nqp/commit/1541771 - it always returned 'obj', even when it was not... so I made it behave like nqp@JVM 14:42
lizmat CHECK INIT END clesrly do not belong on the block
FROGGS .oO( Y U SCREAM? )
lizmat and fwiw, my hack with Variable.slash mate CHECK work
*made 14:43
jnthn FROGGS: Hm, I don't see why it always did that, is all...
14:45 crab2313 left
FROGGS jnthn: well, before it checked whether var->vtable->base_type is smo_id, and returned zero otherwise ( == obj) 14:45
and since there are other base types, these need to be checked 14:46
jnthn well, we can try to pretend Parrot things are native types, but it's probably not a good idea.
FROGGS hmmm 14:47
jnthn: should I revert it? 14:48
jnthn oh, actually, that wasn't what really bothered me 14:49
This is
- $1 = ss.inlineable ? ss.boxed_primitive : STORAGE_SPEC_BP_NONE;
+ $1 = ss.boxed_primitive;
It's completely legit for boxed_primtive to contain complete junk if inlineable is false. 14:50
FROGGS ahh, okay
jnthn I'm not sure if anything tries to get away wiht that, but it's entirely possible.
oh wow...JVM port just caught a bug that probably leads to reading random memory on Parrot. 14:53
FROGGS O.o
jnthn (in a strange edge-case involving return type checks and inlining)
masak r: my @a = [1,2,3], [4,5,6], [7,8,9]; say ([Z,] @a).perl 14:54
camelia rakudo c2ef11: OUTPUT«((1, 4, 7), (2, 5, 8), (3, 6, 9)).list␤»
masak is there a way to "harden" those lists into arrays?
r: my @a = [1,2,3], [4,5,6], [7,8,9]; say ([Z,] @a).tree.perl
camelia rakudo c2ef11: OUTPUT«([1, 4, 7], [2, 5, 8], [3, 6, 9]).list␤»
masak ah.
15:03 crab2313 joined 15:04 grondilu joined
grondilu rn: my @a = 1 .. 5; say @a[0,2...*] 15:05
camelia niecza v24-88-g1f87209: OUTPUT«(timeout)»
..rakudo c2ef11: OUTPUT«1 3 5␤»
grondilu I used ^this in rosettacode.org/wiki/Fast_Fourier_t...orm#Perl_6 but I'm afraid Niecza is right here: we can't use a infinite list as subscript range and expect a finite output, can we? 15:07
moritz we can expect a lot of things to happen :-)
but yes, that's a place where rakudo's array indexing breaks the Liskov substitution principle 15:08
(Array is a subclass of List, and List truncates at the end, but Array doesn't)
15:17 ecocode` joined 15:19 benabik left
lizmat Q: is there a way to "connect" an array lazily to the return value of a method ? 15:19
*inside* the method ? 15:20
tadzik connect?
lizmat well, I have a lazy map {} pushing to an array
the array should be returned lazily
15:21 ecocode left
lizmat if I just say "return @array" it would be eager, no? 15:21
jnthn Where is the lazy map?
lizmat inside the method (trying for a naive implementation of uniq(:with) 15:22
15:22 ecocode` left
jnthn pushing sounds odd 15:22
gather/take sounds more appropriate. 15:23
masak lizmat: just return a gather?
jnthn Or just a straight map
lizmat straight map it is then
masak yeah, even a straight map'd work.
but it'd likely have side effects :)
jnthn only do gather/take when a map can't work, fwiw. :)
(mapiter is cheaper)
15:24 benabik joined
masak indeed. 15:24
15:24 cognominal joined
masak and even perf considerations aside, gather/take is the heavyweight and map is easier on the eye. 15:24
jnthn aye 15:25
benabik The PicoLisp "implementation" of FFT is just calling libfftw3?
15:26 PacoLinux joined
diakopter what exactly about gather/take makes it slow 15:28
dalek p: 8e8ae36 | (Tobias Leich)++ | src/vm/parrot/QAST/Compiler.nqp:
we cant trust boxed_primitive when inlinable is false, jnthn++
jnthn diakopter: The fact that under the hood it's a coroutine/continuation, whereas a straight map only needs closures. 15:29
diakopter right, but why is that slower
on parrot I mean 15:30
clearly it could be slower on jvm b/c of the stack saving/restoring
15:31 leont left
masak has some fun running the t3 submissions against various inputs 15:31
15:32 mtk left
jnthn wonders if masak ran any of them on Rakudo JVM :) 15:32
15:35 mtk joined
masak ooh, no I didn't :) 15:39
but that's a fine idea.
lizmat
.oO( oh how I *hate* it when something works, but not in the CORE.setting )
 15:52
moritz hopes masak++ doesn't find input where his t3 submission recurses infinitely 15:57
masak moritz: why, are you aware of such input? :> 16:03
moritz: one contestant actually does recurse infinitely... :/ 16:04
moritz masak: no, but before I got the algorithm into its current form, it would loop all the time 16:05
it felt very haskellish: either infinite loop, or correct :-)
masak heh. 16:06
16:06 FROGGS left
moritz go++ # not whining when I include trailing newlines 16:07
16:07 FROGGS joined 16:10 arcterus joined
lizmat masak: wondering about .uniq: maybe it would be handier if you could specify a :normalize parameter, possibly as well 16:11
masak lizmat: I was thinking about that yesterday too. 16:12
lizmat: in some sense it feels right, in another sense it feels like adding that one bell-and-whistle too many.
jnthn Why is it not :by?
lizmat instead of :normalize ? 16:13
good point, :by makes sense
moritz fwiw there's nothing wrong with giving uniq a transformation function
but I don't think we should give it a comparison function
because that would make the algorithm O(n²), when somebody with knowledge of the comparison function might do much better 16:14
and I don't think the setting should contain such "surprisingly" slow methods
masak moritz: you mean the comparison function I spec'd yesterday? 16:15
16:15 arcterus left
moritz masak: correct 16:15
masak yes, we can revert it.
and I think your argument carries some weight.
lizmat so :by it is going to be ?
masak though the fact remains that I *did* have a need for :with(&[eqv]) 16:16
even though it's slow.
lizmat why not have both ?
moritz masak: you should get that with :as(*.perl), no?
masak though, hm, a :by transformation function would have been enough in that case.
moritz: yes, just realized.
lizmat :as ??
masak moritz: as long as .perl promises to be canonical enough.
moritz lizmat: I'd name the transformation function :as 16:17
masak lizmat: transformation function.
moritz because :by still sounds like comparison
lizmat ah, ok
masak I'm now 100% with moritz. we should unspec :with and spec :as.
and always do === semantics.
lizmat ok, then I'll ditch my :with implementation
masak I'll reverse the spec and write some new one.
moritz feels bad for lizmat, who recently has implemented several things only to have them reverted/removed in the specs 16:18
lizmat that's life, and it's also about the travel, not just about the goal
moritz wise words 16:19
dalek ecs: f6aa7f8 | masak++ | S32-setting-library/Containers.pod:
Revert "[S32/Containers] specify :with for .uniq/.squish"

This reverts commit ec1b7c619dec5b6d7b779c97bfefc4e2285e72f6.
Conflicts: S32-setting-library/Containers.pod		 16:20
masak conflicts because of some improvements that snuck in after the reverted commit :) 16:21
16:24 PacoLinux left
dalek kudo/nom: fe0813a | jnthn++ | src/vm/jvm/runtime/org/perl6/rakudo/ (2 files):
Implement compile time bind analysis on JVM.

Gets compile time multiple dispatch analysis working. For some reason, inlining doesn't happen yet, though.		 16:26
masak does .perl guarantee that hash keys are ordered?
moritz I don't think so
masak dang, then .perl is not a good supplement for infix:<eqv> 16:27
moritz but that's about the only case where it's not
masak rn: say [{ :a(1), :b(2) }, { :b(2), :a(1) }].perl
moritz maybe we should simply make Hash.perl order the keys, and be don with it
camelia rakudo c2ef11: OUTPUT«[{"a" => 1, "b" => 2}, {"b" => 2, "a" => 1}]␤»
..niecza v24-88-g1f87209: OUTPUT«[{"a" => 1, "b" => 2}, {"a" => 1, "b" => 2}]␤»
masak rn: say [{ :a(1), :b(2) }, { :b(2), :a(1) }].gist
camelia niecza v24-88-g1f87209: OUTPUT«{"a" => 1, "b" => 2} {"a" => 1, "b" => 2}␤»
..rakudo c2ef11: OUTPUT«("a" => 1, "b" => 2).hash ("b" => 2, "a" => 1).hash␤»
masak moritz: I think so.
moritz: though it does feel "extra" compared to the original roundtripping criterion. 16:28
timotimo jnthn: is there any speed improvement to compile-time bind analysis if there's no inlining? 16:29
moritz the bind checks can be skipped at runtime? 16:30
jnthn For multi dispatch, it can choose the candidate at compile time 16:31
And then just index into the candidates list, rather than hitting the dispatch cache or, worse, the dispatcher itself
timotimo ah, good 16:32
want me to run the benchmarks? 16:33
jnthn Well, I'm still working on the inlining bit. I got it all happy on JVM
Then an on-the-side cleanup I needed to do is somehow breaking setting building on Parrot
timotimo righto, i'll wait for that then :) 16:34
moritz Ronja just plays "eating icecream" with some color pens. But she doesn't quite get the predending part yet, so the pens leave colorful traces in her face :-)
timotimo hooray for nontoxic pens! 16:36
masak I have a patch for :as, but I'm kind of put off by the fact that we don't have a replacement for the infix:<eqv> use case. 16:37
lizmat so maybe we still need both? 16:38
moritz masak: that's because some other bits and pieces are missing, not because .uniq(:as) is somehow flawed
masak: nobody yet relies on .uniq(:with), so you aren't breaking anybody's code 16:39
lizmat except for some spectests :-) 16:40
moritz :-) 16:41
masak moritz: yes, but I started out with a real-world use case, and right now we don't have a good answer to it.
moritz: I can also see :as being part of that answer; I'm just missing the other part.
FROGGS jnthn: I've played with parrot's thread while you had fun without us... gist.github.com/FROGGS/0cfae647d28c493df73e 16:43
jnthn: I just wonder why the add-i-mt.t script uses my cpus so well while the other does not
I asked in #parrot, but, well... I guess it might be better to append my questions to an existing issue... 16:44
jnthn: have an idea about something I should try? 16:45
jnthn FROGGS: Gotta run for a train now...ask me again later :)
&
moritz masak: so, what kind of data structures are you uniq()ing?
FROGGS I guess the gazillion small tasks, are waiting for the main thread b/c they access lexicals
jnthn: sure :o)
16:46 jaldhar left
dalek ecs: c313c29 | masak++ | S32-setting-library/Containers.pod:
[S32/Containers] spec :as for uniq/squish
 16:48
masak moritz: in my real-world use case? oh, just an array of hashes of simple values.
moritz masak++ # specs commit 16:49
dalek kudo/nom: 59e521e | (Elizabeth Mattijsen)++ | src/core/ (2 files):
Make List.(uniq|squish) up to spec
 16:53
lizmat 5 mins between spec and implementation change, a new record ?
moritz lizmat: we've had stuff that has first been implemented, and then specced :-) 16:55
masak depends if negative times count... right :) 16:56
&
lizmat negative times don't count
:-)
16:57 celesta joined
celesta ahoi #perl6 16:57
lizmat ahoi celesta! 16:58
dalek ast: 70ed9dc | (Elizabeth Mattijsen)++ | S32-list/squish.t:
Spectests for List.squish adapted to new :as functionality
 17:00
17:02 jaldhar joined
dalek ast: 3dd1995 | (Elizabeth Mattijsen)++ | S32-list/ (2 files):
Add tests for List.uniq(:as)
 17:04
lizmat $ perl6 -e 'say <a bb aa ccc b dd>.uniq( :as({substr($^a,0,1)} ) )' 17:05
a bb ccc dd
too bad you can't specify that as :as( substr *,0,1 ) 17:06
moritz lizmat++
lizmat: *.substr: 0, 1
lizmat ah, duh
dalek ast: f22dd02 | (Elizabeth Mattijsen)++ | S32-list/ (2 files):
Simplify :as specification, suggested by moritz++
 17:10
lizmat cycling& 17:11
17:11 ggoebel left 17:13 dmol1 left, dmol joined
lizmat moritz: about using *.perl, apart from it not working currently because .perl is not doing the right thing everywhere 17:14
is that for each comparison, the complete ..perl string would need to be buiilt
whereas with <eqv>, it could very early fail without having to traverse the whole structure being checked 17:15
17:15 birdwindupbird left
lizmat for big structures, I guess wouldn't make O(n²) so bad 17:16
especially since .perl at the moment is already *really* slow
moritz lizmat: good point that I hadn't considered so far 17:17
lizmat so I guess a case could be made for both 17:18
moritz wonders which version of rakudo he should use when he needs nativecall 17:20
2013.06 should be fine, right 17:21
arnsholt I think latest monthly should be ok 17:23
timotimo r: say "foo bar" ~~ / ' ' <?before <alpha>>/ 17:28
camelia rakudo 59e521: OUTPUT«｢ ｣␤␤»
timotimo r: say "foo bar" ~~ / ' ' <?before <foo=alpha>>/
camelia rakudo 59e521: OUTPUT«｢ ｣␤␤»
timotimo interesting
17:49 imarcusthis- joined
ruoso trying to find himself in debugging the rakudo jvm runtime... the gather/take problem seems to be caused by the control exception placed by "take" not being caught properly... 17:50
17:55 Rotwang joined
ruoso Aha! the problem is when there is no explicit block, then the gather is not creating a block for the code to be invoked... 17:56
17:58 sqirrel joined 18:01 rindolf left 18:08 sidus left 18:09 colomon left
ruoso Actually, that's not it... the problem in that test is with the "last". 18:10
18:19 jaldhar left
ruoso r: say gather for loop { take 1; last }; 18:30
camelia rakudo 59e521: OUTPUT«===SORRY!===␤Missing block␤at /tmp/YsYWgDkVdF:1␤------> say gather for loop { take 1; last }⏏;␤ expecting any of:␤ argument list␤ postfix␤ statement end␤ statement modifier␤ statement modifi…
ruoso r: say gather loop { take 1; last };
camelia rakudo 59e521: OUTPUT«1␤»
18:43 coax75ohm left
dalek kudo-star-daily: 0faf9ba | coke++ | log/ (5 files):
today (automated commit)
 18:54
19:10 sqirrel left
[Coke] hurls www.darthsanddroids.net/episodes/0918.html for the harry potter fans. 19:10
timotimo D&D is pretty excellent. 19:19
this one, i mean
19:20 iSlug joined
timotimo hm, having slurpy LoL was the blocker for turning ~ into a list precedence operator, yes? 19:25
19:25 Rotwang left 19:29 iSlug left 19:39 Rotwang joined
timotimo either implementing ~ properly or turning interpolation into calls with join("") would give a decent speed boost i think 19:43
19:44 Bzek left
diakopter it's worth trying 19:45
timotimo hm, my current implementation of trim_heredoc relies on infix:<~> calls 19:46
sorear o/ 19:50
diakopter sorear: howdy :) how goes?
sorear busy 19:57
20:09 grondilu left 20:11 bluescreen10 joined 20:21 jaldhar joined 20:22 jaldhar left, BenGoldberg left, jaldhar joined 20:24 fridim_ joined
arnsholt [Coke]: Oh, that was mean. I'd forgotten about that webcomic 20:30
Now I've got lots and lots of catching up to do =) 20:31
20:37 Rotwang left 20:43 arcterus joined 20:45 colomon joined 20:59 bruges_ joined 21:00 bruges left 21:05 crab2313 left, crab2313 joined 21:08 Ulti left
pmurias timotimo: it's possible to close classes so the optimizer can know more 21:08
21:09 Ulti joined, leont joined
timotimo pmurias: like "is closed"? 21:09
pmurias use oo :closed; 21:10
lizmat I thought all classes were closed unless use MONKEY_TYPING was seen ?
21:11 crab2313 left
pmurias assuming I don't have an old version of the spec the default is open and non-final 21:11
lizmat spec is not necessarily equal to the implementation :-) 21:12
moritz, masak, jnthn, TimToady would be able to tell you for sure, but I guess they're all away
pmurias currently we don't do much optimalisation
ruoso still kind of lost on how to debug rakudo on the jvm 21:13
colomon ruoso: "say" is your friend. ;)
pmurias lizmat: github.com/perl6/specs/blob/master...ed-classes 21:14
21:14 grondilu joined
ruoso colomon: heh... I don't mean debugging perl code in rakudo, I mean debugging rakudo itself 21:14
pmurias lizmat: so according to spec they are open and non-final by default, but it's possible to change the default for the whole up with 'use oo :final,:closed;' 21:15
colomon ruoso: then say is all the more necessary. ;)
ruoso colomon: fair enough... I guess I just need a better understanding on the transition between rakudo, nqp and jvm
lizmat pmurias: $ perl6 -e 'use oo :final,:closed' 21:16
===SORRY!===
Could not find oo in any of:
so, NYI
diakopter pmurias: where in the spec is that
colomon it's simple: if you're in rakudo, you use say. If you're in nqp, you use say(). If you're in jvm you have to use some acursed 40 character long Java think to get the effect of say.
lizmat S31:73
synopsebot Link: perlcabal.org/syn/S31.html#line_73
pmurias diakopter: github.com/perl6/specs/blob/master...ed-classes 21:17
lizmat S12:2189
synopsebot Link: perlcabal.org/syn/S12.html#line_2189
ruoso like... is there a way to dump the intermidiate states? like the parse tree, then the op tree, then the generated jvm code?
pmurias ruoso: for nqp there's --target=ast
ruoso ah... that may help
pmurias ruoso: which could work for rakudo
diakopter I thought it was --target=qast
FROGGS diakopter: that changes a few months ago 21:18
s/s\b/d/
ruoso ast is recognized, qast isn't... but ast fails with "cannot stringify this" 21:19
(at least in rakudo...)
pmurias and jast? 21:20
ruoso: nqp --target=ast -e 'say(1)'
for nqp
I should check for rakudo
ruoso I tried with rakudo, not nqp
pmurias maybe something is broken 21:21
ruoso: does --target=jast work? 21:22
grondilu me notices that '.say for open("someverylongASCIIfile.txt").lines' doesn't seem to display anything until the whole file is in memory. Which is not great.
ruoso looking 21:23
grondilu noticed that while trying to run rosettacode.org/wiki/Anagrams#Perl_6
ruoso jast does work
colomon grondilu: that's the "for isn't lazy about getting the data to loop over" thing again 21:24
grondilu r: my @a := 1 .. *; for @a { last if $_ > 10; .say } 21:25
camelia rakudo 59e521: OUTPUT«1␤2␤3␤4␤5␤6␤7␤8␤9␤10␤»
grondilu ^seems lazy enough 21:26
colomon grondilu: yes, unless it's known infinite
grondilu r: my @a := 1 .. 100000000000000; for @a { last if $_ > 10; .say }
21:26 FROGGS left
camelia rakudo 59e521: OUTPUT«(timeout)» 21:26
grondilu indeed
good to know
colomon doesn't know what that will do 21:27
ruoso jast works, but is very cryptic, any docs on how to read it?
grondilu isn't there a way to coerce lazyness in a for loop? 21:28
(just curious)
colomon it should just automatically happen, as far as I know. 21:30
ruoso std: my @a = lazy for (1..10) { say $_ }
grondilu ok
camelia std c2215f0: OUTPUT«ok 00:00 46m␤»
ruoso I think you can force "strict lazyness" with the "lazy" operator (but my reading of the spec is very very old) 21:31
r: my @a = lazy for (1..10) { say $_ }
camelia rakudo 59e521: OUTPUT«===SORRY!===␤Unexpected block in infix position (two terms in a row, or previous statement missing semicolon?)␤at /tmp/2VK44lAZwn:1␤------> my @a = lazy for (1..10) ⏏{ say $_ }␤ expecting any of:␤ postfix␤ infix…
grondilu td: my @a := gather .take for ^10; 21:33
21:34 Ulti left
grondilu ruoso: where is this documented? I don't find it in S07 21:35
21:36 Ulti joined
ruoso perlcabal.org/syn/S07.html#Levels_of_laziness , but it has DRAFT status 21:37
lizmat r: my str @a; @a.push: "foo" # huh, no push for "str" ? 21:41
camelia rakudo 59e521: OUTPUT«Cannot call 'push'; none of these signatures match:␤:(Any:U \SELF: *@values, Mu *%_)␤ in method push at src/gen/CORE.setting:1508␤ in block at /tmp/ioECPFhKAU:1␤␤»
timotimo for loops are currently always eager, iirc 21:42
due to a rakudobug
ruoso lizmat: did you mean "my Array of Str @a"?
r: my Array of Str @a; @a.push: "foo"
camelia rakudo 59e521: OUTPUT«Type check failed in .push; expected 'Array+{TypedArray}' but got 'Str'␤ in block at src/gen/CORE.setting:7198␤ in method push at src/gen/CORE.setting:7181␤ in method push at src/gen/CORE.setting:1508␤ in block at /tmp/Lvo2Zn72v_:1␤␤»
lizmat no, lowercase str
r: my Str @a; @a.push: "foo" # works fine 21:43
camelia rakudo 59e521: ( no output )
timotimo why is the signature just Any:U \SELF? there ought to be a second one i feel
21:44 BenGoldberg joined
lizmat ruoso: that syntax is wrong 21:45
it is either my Type @a, or my @a of Type
the latter doesn't work yet, is on my list though
21:59 frdmn joined 22:00 colomon left 22:10 arcterus_ joined 22:11 arcterus left
masak .oO( it's on my array though ) 22:13
timotimo .o( he's making an array, checking it twice ... ) 22:15
masak .oO( Franz Array )
lizmat
.oO( Schindler's Array )
 22:16
timotimo .o( Hans Arraiy ) 22:17
22:41 Marlene joined
ruoso hmm... so something in the ast is not stringifiable, so the entire thing blows up, now I have no idea how to discover what is failing 22:44
hmm... and when I attach the debugger it works... 22:46
actually, it's something to do with parsing file vs inline in the repl
ok, now I nailed it down, if I have an enclosing anon block, it fails... such as ./perl6 --target=ast -e '{ my @x = 1..9 }' 22:52
22:54 spider-mario left
jnthn ruoso: I can reproduce that here too. Huh... 23:01
ruoso is there some doc on the different QAST types and what they're for?
jnthn in docs/ in NQP I think is some 23:02
I think I see the problem...
23:04 jaldhar left
pmurias ruoso: docs/qast.markdown has some info on the QAST types 23:06
23:06 pmurias left
ruoso I am slightly surprised by the number of QAST::Want nodes 23:08
maybe I am just looking too early in the process 23:09
23:09 frdmn left 23:11 frdmn joined 23:12 frdmn left, frdmn joined 23:14 frdmn left 23:15 frdmn joined 23:16 Marlene left 23:21 frdmn left, frdmn joined, PacoAir joined 23:22 fgomez left 23:23 frdmn left 23:24 frdmn joined 23:31 benabik left, benabik joined 23:35 arcterus_ left
dalek p: d6f0f6f | jnthn++ | src/QAST/BVal.nqp:
Fix QAST::BVal dumping; ruoso++.
 23:36
kudo/nom: 8d851b7 | (Elizabeth Mattijsen)++ | src/core/List.pm:
Reimplement :with for List.(uniq|squish) (keep :as available)

Yes, it was removed from the spec, but the alternative for doing [eqv] based uniqueness using *.perl is 1. still not possible now (because .perl doesn't always give an accurate representation), and 2. very resource intensive, because a whole structure needs to be stringified first before being able to find out that they're different. Whereas [eqv] is very quickly able to find out whether two structures are different.
So, do I believe it belongs in the spec? Yes! :-)		 23:37
ast: 0f0ea4c | (Elizabeth Mattijsen)++ | S32-list/ (2 files):
Added tests for :with for List.(uniq|squish)
 23:38
lizmat and on that thought, calling it a day again… sleep& 23:39
jnthn 'night, lizmat 23:40
23:51 benabik left 23:57 dmol left, benabik joined, odoacre_ left 23:58 ggoebel joined, odoacre_ 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!