»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.
hoelzro I thought I had the algorithm wrong for a good 30 minutes before I figured it out =/ 00:00
jnthn hoelzro: It doesn't have to be an assignment to @!values. Declare a my @x; there and assign to it and you get the same.
hoelzro: I'm wondering if we've a $*LEFTSIGIL leak or something.
hoelzro I suspected $*LEFTSIGIL 00:01
but didn't know what I was doing =P
00:01 mtk joined
hoelzro tries some moar prints 00:01
colomon pdcawley: wonder if anyone else is building it with 10.9. 00:03
colomon is very slow to upgrade … only upgraded from 10.6 to 10.8 back in July.
00:05 xenoterracide left
pdcawley colomon: any docs anywhere on smoking the perl6 build? 00:05
colomon pdcawley: on different build platforms? I don't think we've got anything like that set up yet.
pdcawley Hmm... I'm _so_ unlikely to step up to that particular plate :) 00:06
00:06 xenoterracide joined
moritz though 'make spectest' would form a good base for that 00:06
jnthn Time for some rest...'night o/ 00:07
pdcawley g'night mate.
00:07 cognominal joined
colomon pdcawley: I've got a p6 module smoker, but so far it only tests on Parrot Rakudo on my Linux box. 00:07
japhb o/ jnthn
Is anyone around that can check out why feather.perl6.nl:3000/projects.json is timing out? 00:09
00:10 thou joined 00:12 dayangkun left 00:14 cognominal left 00:15 zakharyas left
[Coke] (smoking the perl6 build) yes, we do that daily. You could run my stuff ad hoc if you wanted. 00:15
[Coke] hurls github.com/coke/perl6-roast-data 00:16
(that tests rakudo-j, rakudo-p, niecza, and pugs every day)
er. we have no smoke server yet though. )we have the parrot one, but no one uses it.)
that's another of my backburner projects.
liz - I think we had mixbag for a day. 00:20
TimToady was in a renaming frenzy.
TimToady foams a little at the mouth. 00:23
hoelzro goes to bed too 00:24
colomon TimToady: can you quickly explain how STD handles parsing regexes? I'm trying to sort out what Niecza is doing...
itz_ how long does the moar "please wait stage" take? I'm sure its just going to core dump :) 00:25
TimToady colomon: easy peasy, just look at the P5 that code turns into :) 00:27
00:29 kivutar left
TimToady colomon: the important thing with respect to :s is that quantified_atom determined all the sigspace, in cahoots with sigmaybe 00:30
and all the constructs that enable sigspace set $*SIGOK so sigmaybe knows whether to treat the space as significant 00:31
*determines
colomon and those are all new / modified with the new sigspace rules, right? 00:32
TimToady plus anything else I said back when Mouq++ was fixing rakudo
so you could check out Mouq's changes and see what he did 00:33
or wait till he gets back online and ask him :) 00:34
the complicated bit is that the sigspace attaches at a different place in the AST, if I recall 00:40
00:42 ajr_ left 00:45 preflex left, preflex_ joined, ChanServ sets mode: +v preflex_ 00:46 preflex_ is now known as preflex
BenGoldberg So I was reading the irc log, and came across S32/Containers ... 00:50
Specifically perlcabal.org/syn/S32/Containers.html#grab
And I'm curious as to why only QuantHash has a .grab method, and not any other container 00:51
TimToady QuantHash is a role 00:54
r: say SetHash ~~ QuantHash
camelia rakudo-parrot de5c8f, rakudo-jvm de5c8f: OUTPUT«True␤»
TimToady r: say Set ~~ QuantHash 00:55
camelia rakudo-parrot de5c8f, rakudo-jvm de5c8f: OUTPUT«True␤»
TimToady hmm, that seems wrong
sets are immutable, so .grab makes no sense
r: my $s = set(<a b c>); say $s.grab 00:56
camelia rakudo-parrot de5c8f: OUTPUT«Cannot call 'grab' on an immutable 'Set'␤ in method grab at gen/parrot/CORE.setting:15300␤ in block at /tmp/j7jTJDNa3r:1␤ in any at /tmp/j7jTJDNa3r:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPHLL.…»
..rakudo-jvm de5c8f: OUTPUT«Cannot call 'grab' on an immutable 'Set'␤ in block at /tmp/309gmtPfNA:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:1086␤ in any evalfiles at gen/jvm/stage2/NQPHLL.nqp:1292␤ in any command_eval at gen/jvm/stage2/NQPHLL.nqp:1196␤ in any command…»
TimToady anyway, Set shouldn't do QuantHash in my book 00:57
(or Bag, or Mix)
so rakudobug on reporting Set ~~ QuantHash as true 00:58
00:58 pmurias left
TimToady on the compiling jvm, I've often run into the OOM that piers hit, but recompiling with the same command immediately generally works! 00:59
dunno why that happens, unless jvm is keeping some state around from run to run 01:00
(this is openjdk)
01:01 xenoterracide left
TimToady maybe it's jvm keeping process around after they've 'exited' somehow 01:01
pdcawley TimToady: Weird. 01:04
Sooner I can get off JVM and onto Moar, the better, methinks.
01:06 emma_ is now known as emma 01:09 jnap left 01:15 xenoterracide joined 01:19 pdcawley left 01:26 Entonian joined 01:28 skyheights joined 01:29 skyheights left, jlaire left
BenGoldberg Why isn't there an Array.grab method? 01:30
01:32 xenoterracide left 01:36 TimK1 left, xenoterracide joined 01:39 jnap joined
lue BenGoldberg: because Array.pick and Array.roll should do the job already :) . 01:40
01:42 xenoterracide left 01:43 thou left 01:44 jnap left 01:47 Entonian left, xenoterracide joined 01:59 jeffreykegler left
timotimo lue: except grab is supposed to remove the item at the same time as returning it, no? 01:59
02:07 lizmat left 02:09 aindilis` left, aindilis joined
lue timotimo: .pick acts as though you're removing items already. Arrays are immutable, AFAIK, so .grab really doesn't provide anything Array.pick doesn't already do. 02:13
TimToady removing items from an Array is not generally what you want to do, either
you should probably be using a SetHash in that case 02:14
lue Array.grab would only serve as a useful synonym for too-liberally type-constrained variables. :)
TimToady well, if you did implement Array.grab, you'd probably want it to do a splice, not a delete 02:17
lue I meant that as in method grab(<params>) { self.pick(<params>) } 02:18
colomon lue: Arrays are mutable. 02:20
lue oh?
r: say ++(Array.new(1,2,3)[1]);
TimToady mapping .grab to .pick is always wrong
camelia rakudo-parrot de5c8f, rakudo-jvm de5c8f: OUTPUT«3␤»
lue r: say (++Array.new(1,2,3)[1]).perl;
camelia rakudo-parrot de5c8f, rakudo-jvm de5c8f: OUTPUT«3␤»
TimToady pick is "without replacement", but the array itself is not modified 02:21
grab always modifies whatever you're grabbing from
it's not polite that way :)
lue would then contest that changing the size of an Array via randomly removing certain numbers is perhaps not the best idea. I'm with TimToady on that being a more SetHashy thing. 02:22
s/numbers/elements/ 02:23
colomon just timed his ISO 10303-21 test suite, and it's 4x faster under jakudo. 02:27
shaves 75% off the timing, I mean. 02:28
also, having perl-j available for easy testing is really nice. 02:36
02:37 kaare_ joined 02:48 filly left 02:50 jnap joined 02:57 filly joined, jnap left 03:05 hummeleBop1 left 03:08 jeff_s1 left 03:19 segomos left, segomos joined
diakopter forwarded something to p6-l 03:23
colomon p6-l? 03:25
r: say (32 < * < 128)(1000) 03:26
camelia rakudo-parrot de5c8f, rakudo-jvm de5c8f: OUTPUT«True␤»
colomon rakudobug
timotimo known :( 03:27
colomon no need to call you-know-who, then. timotimo++ 03:28
colomon mourns his beautiful code which …. no, I can do this. 03:29
actually looks better this way.
03:42 atroxaper joined 03:46 xinming_ joined 03:49 xinming left
diakopter p 03:51
colomon: perl6-language?
you know, on that thing we used before github, email 03:52
colomon diakopter: that was more along the lines of "why would a sensible person send something to perl6-language?"
diakopter heh oh 03:54
colomon I've come to associate it strongly with endless pointless chatter. :)
diakopter take a look at what I sebt 03:56
sent
colomon diakopter++ # hopefully that's safe 03:57
diakopter what's safe? 03:58
colomon your message
people won't start arguing about the color of the benchmarks, or something like that.
bikeshedding, that's the term. 04:01
04:20 atroxaper left, atroxaper joined 04:21 preflex_ joined, ChanServ sets mode: +v preflex_ 04:22 preflex left, preflex_ is now known as preflex 04:26 atroxaper left 04:27 atroxaper joined 04:32 BenGoldberg left 04:49 Entonian joined 05:01 adu joined 05:05 atroxaper left 05:09 raiph left 05:16 atroxaper joined 05:18 thou joined 05:20 araujo left, itz joined 05:21 Entonian left
diakopter well, you could build it entirely out of paint. 05:37
or paint thinner it into nothing.
05:40 Entonian joined 05:46 Entonian left 05:52 atroxaper left 06:08 rurban1 joined 06:12 rurban2 joined 06:13 rurban1 left 06:15 SamuraiJack_ joined 06:19 rindolf joined 06:22 rurban2 left 06:27 SamuraiJack_ left 06:30 SamuraiJack joined 06:38 SamuraiJack left, rurban1 joined 06:39 SamuraiJack joined 06:50 zakharyas joined 06:52 rurban1 left 07:06 darutoko joined
[Coke] diakopter: wow, if that was on a weekend, I'd make the drive. 07:08
07:11 zakharyas left 07:19 quester joined 07:23 rindolf left
dalek rl6-roast-data: 2d5a20e | coke++ | / (5 files):
today (automated commit)
 07:38
FROGGS japhb: I am working on that: irclog.perlgeek.de/perl6/2013-12-06#i_7967667 07:45
07:52 SamuraiJack_ joined 07:53 lizmat joined 07:56 SamuraiJack left 07:57 adu left, adu_ joined, adu_ is now known as adu
moritz m\orning 07:59
07:59 sqirrel joined
FROGGS morning mo/ritz 07:59
preflex: tell pdcawley that your problem seems to be that java's classpath environment variable is set, can you unset it and try again? 08:02
preflex Consider it noted. 08:03
08:03 adu left
FROGGS lee_: same for you 08:03
jnthn: rakudo-p's stage parse dropped from ~124s to ~117s # jnthn++ 08:10
moritz thought the optimizations where jvm only? 08:12
oh, there's the "Optimize QAST node creation" commit in nqp 08:13
FROGGS yeah, must be that then...
I mean, I didn't make a proper benchmark 08:14
moritz and a commit not use slurpies in the accessors of the QAST nodes
FROGGS both are likely to save seconds 08:15
08:16 quester left, xenoterracide left, quester joined 08:17 quester is now known as Guest62135 08:18 sqirrel left
Guest62135 help set 08:24
08:25 Guest62135 left 08:26 xenoterracide joined
FROGGS I have an idea about installing several modules that provide a binary that has the same name as the other modules 08:30
when installing it a wrapper is created with that file name, that would execute the binary of the "best" module [more]
like a "use Foo" would need/import the best module
and then the wrapper would provide options like --auth, --name and --ver, that does the lookup if you want to execute an older binary or a dev version 08:31
08:36 quester_ joined, quester_ left, quester_ joined, genehack_ is now known as genehack
FROGGS moritz: what is your opinion on that? 08:36
08:37 quester_ left
FROGGS it would even work with development modules, since their bin folder should be in path first (and the user is responsible for that) 08:37
08:39 lizmat left
FROGGS and btw, I am thinking about creating a wrapper that defaults to `perl6`, and a wrapper-p for perl6-p, and a wrapper-j for perl6-j 08:41
so, panda-j would work then in theory
08:51 quester__ joined, quester__ left, quester__ joined 09:06 quester__ is now known as quester 09:08 FROGGS left, quester left 09:14 lizmat joined
lizmat FROGGS: isn't that the responsability of CompUnitRepo.candidates ? 09:22
09:23 PacoAir joined
lizmat and morning #perl6! 09:23
09:28 FROGGS joined
FROGGS lizmat: the wrapper would ask the CompUnitRepos for candidates, as for "use Foo", yes 09:29
lizmat so I'm not sure you would need a wrapper ? 09:31
FROGGS lizmat: I am talking about installed scripts
how will that work without a wrapper?
say you have installed version 1 of panda and version 2 of panda in home 09:32
both provide a panda executable 09:33
lizmat that would be 2 CompUnitRepo objects, no?
FROGGS no
home is a CompUnitRepo
lizmat one for the installed version, and one for home?
FROGGS err, I meant that both are installed in home 09:34
so, when you execute `panda`, you should clearly execute the binary from version 2
but what if you want to execute `panda` from version one? 09:35
you are unable to do so
(without a wrapper)
lizmat I'm unsure whether it should be possible at all to have this situation 09:36
FROGGS why? 09:37
lizmat assuming a CompUnitRepo maintains a database of sorts in the directory for which it is responsible
FROGGS yes
lizmat how would both panda 1 and panda 2 be installed in there without either two databases clashing 09:38
or the one database knowing about both of them
FROGGS you have a single bin folder per CompUnitRepo
lizmat perhaps
that's the responsibiliuty of the CompUnitRepo 09:39
FROGGS so, if you install several versions of panda to the same CompUnitRepo, then you would be in that situation, no?
lizmat whether or not that uses a bin dir, is entirely up to the implementation of that CompUnitRepo
FROGGS you need a bin dir to execute scripts
there is no other way
lizmat hmmm 09:40
so we're extending "use" and "need" now to qx() ?
FROGGS you need somewhere a folder that contains a thing called "panda" to run panda
no
it is about executing script from your shell
09:40 SevenWolf left
FROGGS you don't run panda from a use statement 09:41
it is a command line tool
lizmat ok, I misunderstood your original statement
09:41 woolfy joined
FROGGS good 09:41
:o)
lizmat
.oO( must wake up first, get more tea )
FROGGS so the problem is:
you type "panda" or "perl6-debug" into your shell, and press enter 09:42
it looks up a binary in you PATH
and then, this script is from a specific version of its distro
it will usually do a "use" of a module of its distro 09:43
which might even be the wrong one for that script version
lizmat hmmm... I guess this touches on the more general issues of non-source files associated with a distribution
FROGGS true, but binaries are a thing on its own 09:44
locating files you might load from within your module is another problem
FROGGS .oO( File::ShareDir )
btw, when doing "panda --ver=1", I'd propose that all use-statements within that panda script that use modules from its dist will get the :ver(1) passed implicitly 09:49
lizmat TimToady: re irclog.perlgeek.de/perl6/2013-12-07#i_7969313 : there is no QuantHash role implemented just yet 09:50
because of the attribute visibility issue with roles, I didn't start with refactoring all of the Set/Bag/Mix stuff to the role structure as specced just yet 09:51
09:54 Piers_ joined 09:55 cognominal joined 09:57 Rotwang joined, kivutar joined
lizmat saved a conversation with jnthn of about 6 weeks ago 10:12
gist.github.com/lizmat/7839326/raw...tfile1.txt
and is going to look at it while at the patch -p1 10:13
10:15 rindolf joined
FROGGS nice 10:17
rindolf Hi all. 10:18
FROGGS: hi.
FROGGS hi rindolf
rindolf FROGGS: there's a pull request for the SDL_manual.
FROGGS rindolf: PR applied, thanks for reporting :o) 10:20
rindolf FROGGS: OK.
FROGGS: you're welcome.
FROGGS: and thanks for applying it.
FROGGS sure
:o)
10:23 Piers_ left 10:24 araujo joined 10:26 Timbus left 10:28 sivoais left 10:29 Rotwang left, sivoais joined, Timbus joined
jnthn morning o/ 10:34
10:34 Piers_ joined
diakopter mawrnig 10:34
lizmat jnthn, diakopter: good *! 10:36
jnthn: re "At the moment, when we see a $!foo, we immediately ask the class, "do you know of a $!foo"?"
I'm having trouble locating where that is done
do you have a pointer :-) 10:37
jnthn Well, it'll be some code-path from variable...
lizmat check_variable perhaps ? 10:38
10:40 szabgab joined
jnthn No 10:40
szabgab hi everyone,
FROGGS don't you see the location when you run that code with --ll-exception?
szabgab: hi
szabgab I wonder if someone could give me a hand with a Grammar
FROGGS sure 10:41
jnthn lizmat: In make_variable_from_parts
szabgab github.com/szabgab/perl6-xml-sax
I have that stupid XML parser
jnthn lizmat: In the elsif $twigil eq '!' { ... } branch
szabgab it passed its test in february - the last time I checked
but now, with * 2013.11 it fails
lizmat jnthn: thanks! 10:42
szabgab of course if there is a good XML parser in Perl 6 , then I can start using that instead of mine ...
hi lizmat , long time not seen!
lizmat hi szabgab, see you at FOSDEM? :-)
szabgab unlikely :( 10:43
lizmat well, you are responsible for woolfy and me having gone there for the past 4 years!
so you should come every now and then as well :-) 10:44
jnthn lizmat: Specifically, it's the call to get_attribute_meta_object that actually throws 10:46
lizmat I just found that, using FROGGS++ suggestion: called from Sub 'get_attribute_meta_object' pc 14738 (src/gen/perl6-actions.pir:5834) (src/Perl6/Actions.nqp:1776)
szabgab lizmat: I know, and I am really sorry 10:47
FROGGS szabgab: I'll check
szabgab FROGGS: thanks!
lizmat don't! although exhausting, it's always been fun!
szabgab lizmat: I am really sorry I cant' come :( 10:48
lizmat ah, ok
:-)
we too :-(
FROGGS perl6-xml-sax$ PERL6LIB=lib prove -e perl6 t
t/xml.t .. 1/39 Left over string: '<chapter id="12" name="perl" ></chapter>'
szabgab: is that what you get?
szabgab yes
this is the first test case with attributes 10:49
so I think the problem in the grammar is somewhere matching attr
the grammar does not match that string 10:50
FROGGS szabgab: gimme a few minutes 10:51
rindolf szabgab: hi.
jnthn FROGGS: Mebbe sigspace changes... 10:52
szabgab hi rindolf
FROGGS that is possible
github.com/szabgab/perl6-xml-sax/b...mar.pm#L13 10:53
maybe that should be a rule?
10:54 thou left
jnthn Well, but does it rely on backtracking semantics? 10:54
But yeah, it's an issue 'cus
\< <element> <attr>* \/\> }
And then
rule attr { [<name>\=\"<value>\"] }
Will no longer match whitespace at the start 10:55
Alternatively, turning token element { \w+ } into a rule may be the quickest fix (to match whitespace after the element name)
Oh...but that may screw up the actions :)
FROGGS szabgab: I changed all regexes to "rule", and now it fails four tests only 10:56
szabgab I guess it would be hard to find out what changes were made to the regex engine that make this fail now. 10:58
and if those changes were intentional ?
FROGGS no too hard
szabgab the question if it is important to know 10:59
FROGGS after an hour or two of bisecting you should know
szabgab :)
FROGGS and I bet these changes where intentional :o)
raydiak :w
FROGGS szabgab: after dinner I'll have a fix
szabgab then it ok
raydiak erm
FROGGS err lunch 11:00
szabgab FROGGS: appreciate
but if there is working XML parser, I could use that too
FROGGS dunno, is there? 11:01
jnthn btw, from a performance perspective it's best to avoid regex in grammars.... 11:06
diakopter heh. 11:07
szabgab I also tried to switch all the regexes to rules 11:10
it seems the test failures are now all about leaving out the spaces
from the text 11:11
11:11 Piers_ left
szabgab Expected: ' before ' # Received: 'before ' 11:11
11:12 denis_boyun joined
lizmat jnthn: what is the significance of "$past.unshift(QAST::Var.new( :name('self'), :scope('lexical') ));" in "make_variable_from_parts" ? 11:15
is that creating self ?
so, if we have multiiple $!vars, we would get multiple "self"s?
jnthn lizmat: When you look up an attribute, you need two children. 11:18
The object to do the lookup against, and then the class the attribute lives in.
lizmat ah, ok
jnthn That unshift is doing the first of those. 11:19
11:19 denis_boyun left
jnthn A QAST::Var is only declaring anything if there's also a :decl(...) in it. 11:19
Otherwise it's just a lookup
lizmat gotcha
11:21 spider-mario joined
lizmat so how would I hang something off of $*PACKAGE ? 11:22
jnthn: because that's where I would to need to hang the hash of arrays of from, right ?
create another hash with $*PACKAGE as the key ? 11:23
jnthn No
Just declare a my %*ATTR_USAGES; or so in package_def, right where my $*PACKAGE is declared. 11:24
Then you'll automatically get the right behavior if one package is nested inside another.
lizmat gotcha 11:25
jnthn github.com/rakudo/rakudo/blob/nom/....nqp#L1896 is where I mean, just to be clear
Though the comment is full of it...should say type object, not meta-object :)
lizmat will fix 11:26
jnthn Hm, my apartment is darker than usual 'cus snow has covered the skylights :) 11:28
11:29 xinming_ left
lizmat so, if I understand you correctly, we would need to keep the $past around for each occurrence, as well as $/ for error reporting, right ? 11:29
so that we can fix up the $past when we compose the class? 11:30
11:31 xinming joined
FROGGS szabgab: you've got a PR 11:31
jnthn Yeah, though if you do $past.node($/) then you just can keep the $past around :)
And then just do .node if you need to look it up.
lizmat ah, gotcha again 11:32
11:33 szabgab left 11:35 szabgab joined
lizmat jnthn: github.com/rakudo/rakudo/blob/nom/....nqp#L3689 contains a call to get_attribute_meta_object 11:36
but apparently it is not interested in the side effect
only to throw if it doesn't exist
this implies there is nothing to fix up 11:37
so, should I just remove the test?
jnthn lizmat: Well, it's *only* interested in the side-effect. :)
lizmat (as I intend to resolve *all* $!private at compose time)
jnthn: duh, must wake up some more 11:38
so, get_attribute_meta_object will never throw
jnthn Well, today it does throw
That's why it calls it.
lizmat package_def will throw later 11:39
jnthn It's for the method foo($!bar) { } case.
Right, but you need to make sure that happens.
szabgab FROGGS: thanks!
colomon jnthn++ # Stage parse : 84.557
lizmat but there is no $past there to fix up?
jnthn Right, we could well just create a QAST::Var there with an appropriate name and $/. It'll be throw-away, but good enough for detecting the error. 11:40
lizmat ok
jnthn So I'd just turn that line into a %*ATTR_USAGES{$name}.push(QAST::Var.new( :node($/) )) or so
(Won't be quite that simple as no auto-viv in NQP. Maybe factor that out into an add_attr_to_check sub...) 11:41
lizmat ok 11:42
FROGGS colomon: what backend is that? 11:45
11:46 tgt joined
colomon FROGGS: parrot 11:46
FROGGS wow, nice machine :o)
colomon FROGGS: it's been routinely 90 for the last month or two, and was 100 for much of the summer 11:47
jnthn Wow, indeed. What's it do JVM in? :)
colomon give me a mine...
minuget
minute
two early
jnthn hands colomon a mine and runs away
colomon silver or gold? 11:48
jnthn Oh, I thought you meant the explosive kind
My bad P
:P
colomon compiling now. 11:49
11:51 markov left 11:53 Baggio left 11:54 szabgab left 11:59 SamuraiJack_ left
lizmat jnthn: if I understand this correctly, I should do the check *after* the pkg_compose, but before the create_code_object, right ? 12:17
jnthn lizmat: Certainly after pkg_compose
Where is the crate_code_object you're thinking about? 12:18
lizmat Actions, 1914
well, in my version, a few lines less in yours, I guess
jnthn lizmat: I think you may want to do it outside of that if/else 12:19
e.g. put it above the # Document
lizmat also for a role?
jnthn Sure, we need to check attribute declarations in roles too...
lizmat ah, duh
ok
jnthn That's why I'm suggesting putting out outside of the if/else 12:20
Note that both branches call compose.
lizmat ok 12:21
do we have .kv in nqp ? 12:22
jnthn No 12:23
Just do for %hash { ... }
And then you can do $_.key and $_.value
lizmat yeah, found that meme in the code :-) 12:33
12:34 markov joined
lizmat jnthn: Stage start : 0.000 12:39
Error while compiling op p6decontrv: No scope specified or locatable in the symbol table for '$!WHY'
jnthn: the diff so far: gist.github.com/lizmat/7840724 12:40
jnthn yeah, I can see what's up 12:41
You can't move as much as you have 12:42
These:
- my $attr := get_attribute_meta_object($/, $past.name());
- $past.returns($attr.type);
Are correctly moved, and should be.
But the other lines...
- $past.scope('attribute');
- $past.unshift(instantiated_type(['$?CLASS'], $/));
lizmat need to be at the original location 12:43
jnthn - $past.unshift(QAST::Var.new( :name('self'), :scope('lexical') ));
lizmat ok, gotcha
jnthn Noeed to stay put.
Otherwise role compilation likely explodes.
12:46 colomon left 12:52 tipdbmp left 12:55 tipdbmp joined
lizmat hmmm... seems I can't build rakudo now because it dies on the delayed get_attribute_meta_object 12:55
13:14 denisboyun joined
lizmat found thinko 13:17
FROGGS uhh 13:21
jnthn++ # blog post!
lizmat ah, yes, jnthn++ (was too focused on my patching) 13:29
good to read 6guts.wordpress.com/2013/12/07/a-fe...k-updates/
13:29 clkao_ joined
lizmat Cannot assign to a non-container 13:30
current instr.: 'BUILD' pc 42 ((file unknown):-2113359804) (gen/parrot/CORE.setting:10024)
jnthn: I think I need to add the (original) $*PACKAGE to get_attribute_meta_object 13:31
because it's now running at another time, so it may be wrong? 13:32
jnthn lizmat: No, $*PACKAGE should be the same thing...you're in the action method for package_def, and that is called at the end of package_def in the grammar, before it returns. 13:36
So I don't think that's your problem.
lizmat hmmm....... still, it seems to die on the first private attribute it encounters in the setting
ah! 13:37
has num $.re;
$!re = re;
so the attribute is defined as public
but referenced as private
could that be the problem?
jnthn Well, that's perfectly allowed, since has num $.re means "declare has num $!re, then ccreate an accessor" 13:38
13:38 darutoko left
jnthn Oh...hmm 13:39
This could be a nasty problem.
lizmat gist.github.com/lizmat/7841431
is my diff so far
jnthn We're now attaching the type later. But the = operator looks at the type of the lhs to decide what to emit. 13:40
Whih, of course, isn't set yet...
*which
lizmat ah, yuck, so we go down deeper into the rabbit hole
jnthn + add_attr_to_check($/, $_<variable_name>, QAST::Var.new); 13:41
That wants a :node passed to new so we can report the error properly, I think...
lizmat the node is added inside the sub 13:42
jnthn oh, so it is..
Well, yes, yuck.
We'll have to solve the deeper issue here at some point I guess.
For now there's a cheat that'll get us most of the way. 13:43
Which is to *try* and find the attribute as we compile, and only build up the %*ATTR_CHECK entries for those unknown at the point we encounter them.
lizmat well, that would build the setting 13:44
but would not fix the problem I'm trying to fix, afaik
jnthn That way, everything that works today works the same, and it's only those from roles that get this new handling.
Well, it'd fix it for talking about attributes from roles unless those attributes are natively typed...
Which most attributes aren't. 13:45
lizmat ah, ok
ok, will change it like that
jnthn So it gets us much of the way. I'm quite sure it covers the stuff you wanted it for in the set/bag factoring.
13:50 kivutar left 13:52 darutoko joined
woolfy jnthn++ 6guts.wordpress.com/2013/12/07/a-fe...k-updates/ typo: "Rakudo on MoarVM progress" in line 2/3: "The first step was the get the core of the compiler itself building." It says "was the get the" and I think should be "was to get the". 13:53
13:53 jnap joined 13:56 denisboyun left
woolfy jnthn: nice blog (& I found no more typos) 14:00
lizmat $ perl6-p --ll-exception -e 'role A { has $!a }; class B does A { method b { say $!a } }' 14:06
yeah!
ok, so you will not be able to have a native private attribute in a role be visible in the class 14:12
moritz it seems the Dancer adevent calendar is stuck at day 2 14:14
14:14 SamuraiJack_ joined
jnthn woolfy: Fixed; thanks! 14:15
lizmat: yay!
lizmat: yeah, we'll have to fix the native private attr thing up later, but that's a trickier fix
lizmat which I will gladly leave to you 14:16
anyway, I learned a lot today :-)
jnthn :-) 14:17
*and* got us an important improvement.
14:18 atroxaper joined
woolfy Christmas gets closer! 14:18
14:23 hummeleB1 joined, kaare__ joined 14:24 kaare_ left 14:30 tgt left, tgt joined 14:35 kivutar joined
dalek kudo/nom: 56ee1a5 | (Elizabeth Mattijsen)++ | src/Perl6/ (2 files):
Private role attributes are now visible in class

Thanks to extensive support by jnthn++. Unfortunately, getting this to work for private *native* role attributes, proved to be needing to go deeper into the rabbit hole. Which will be for another day.		 14:44
14:45 tadzik left 14:47 xinming_ joined 14:48 dalek left 14:49 xinming left, dalek joined, ChanServ sets mode: +v dalek 14:51 dalek left, Juerd left 14:52 sergot left, dalek joined, ChanServ sets mode: +v dalek, Juerd joined 14:53 Util left, dalek left, Util joined 14:54 dalek joined, ChanServ sets mode: +v dalek, PerlJam left, PerlJam joined 14:58 masak left, masak joined 14:59 Juerd left 15:00 Juerd joined 15:04 sergot joined 15:07 telex left 15:08 telex joined
lizmat hmmm... seems dalek quit? 15:10
anyway, I added tests for role attribute visibility
jnthn lizmat++ 15:12
15:12 sergot left 15:13 rurban1 joined, sergot joined 15:14 dalek left, hugme left 15:15 dalek joined, ChanServ sets mode: +v dalek 15:16 masak left 15:17 masak joined 15:24 hugme joined, ChanServ sets mode: +v hugme 15:28 PerlJam left
nwc10 sorry if it's obvious, but it's not obvious to me - why would having one shared NFG across the interpreter be potential for a DOS, but having it one-per string would not? 15:33
15:33 dalek left 15:34 Juerd left 15:35 dalek joined, ChanServ sets mode: +v dalek
diakopter nwc10: b/c you could run out of negative synthetics if the mapping were shared 15:35
nwc10 I think you'd need at least 8Gb of input to do that
2 billion negative synthetics, and each takes at least 4 bytes of input to define 15:36
(I think.)
15:36 Juerd joined
diakopter 1 billion? 15:37
nwc10 jnthn: does this commit actually generate "better" code, or is it just useful in deferring something to help get bootstrapping: github.com/rakudo/rakudo/commit/84e430f241
diakopter er
nwc10 I thought we had 2147483648 negative synthetics to play with
jnthn nwc10: It eliminates a check that will always come out as "true" 15:38
nwc10 jnthn: aha
15:38 synopsebot left
jnthn nwc10: So yes, better code. 15:38
15:38 masak left
diakopter bye bye feather? 15:38
15:39 sergot left, d4l3k_ joined, Util left, dalek left
diakopter nwc10: well yes a huge input. but it would also require at least that much memory on the system 15:39
15:39 masak joined
diakopter which is too much 15:39
15:39 Util joined 15:40 d4l3k_ left, PerlJam joined 15:41 dalek joined, ChanServ sets mode: +v dalek 15:42 atroxaper left, SamuraiJack_ left, markov left, sivoais left, cognominal left, woolfy left, awwaiid_ left, labster left, kst left, segomos left, [particle] left, sjn left, ribasushi left, Exodist left, Juerd left 15:43 dalek left, dalek joined, ChanServ sets mode: +v dalek 15:45 sergot joined, synopsebot joined 15:46 masak left, hugme left 15:47 dalek left 15:48 Juerd joined, dalek joined, ChanServ sets mode: +v dalek 15:50 sergot left 15:51 dalek left, sergot joined, masak joined 15:53 dalek joined, ChanServ sets mode: +v dalek 15:56 jeffreykegler joined
masak good UGT, #perl6 15:57
Woodi o/
diakopter slutations
salutations
o/ 15:58
15:59 Psyche^ joined
lizmat masak! 15:59
Woodi so what is CompUnitRepo ? 16:00
FROGGS Woodi: it is a repository of installed modules 16:01
Woodi: and a piece of code that tells you which candidates are installed for a given use-statement for example 16:02
16:02 Psyche^_ left
lizmat actually, a CompUnitRepo could represent remotely installed modules as well 16:02
FROGGS true
Woodi so it is code internal to Perl6-interpreter/compiler ? or sits in panda/packager ? 16:03
lizmat Woodi: the spec describes 2 implementations of that interface
but any developer is free to develop another implementation and sitck it in @INC 16:04
Woodi looking...
FROGGS the compiler should already have some sort of CompUnitRepo built in, but it is intended that one can odd custom ones
panda should just talk to the CompUnitRepo 16:05
lizmat jnthn: it appears that my private role attribute visibility fix does not work in settings :-( 16:06
16:06 hugme joined, ChanServ sets mode: +v hugme
jnthn lizmat: There's no reason why it shouldn't... 16:07
masak diakopter: slutations to you too :P
16:08 atroxaper joined, SamuraiJack_ joined, markov joined, sivoais joined, cognominal joined, woolfy joined, awwaiid_ joined, labster joined, kst joined, segomos joined, [particle] joined, sjn joined, ribasushi joined, Exodist joined 16:10 xenoterracide left
Woodi 404s in S11 for links to CompUnitRepo::Local::... 16:15
wiki: The classic combination of a user account number or name and a secret password is a widely used example of IT credentials. # do we need passwords on compiled code ? :) 16:20
FROGGS huh? 16:21
Woodi TGT maybe ? :) but nvm... 16:22
how compunitrepo looks in @INC ? path name ? 16:23
lizmat CompUnitRepo objects should be in @INC, it's not implemented like that yet 16:24
Woodi object ?
ok, got it 16:25
16:25 c1sung left
FROGGS p: class A::B { }; sub foo(A::B $bar) { } 16:26
camelia ( no output )
lizmat std: class A::B { }; sub foo(A::B $bar) { } 16:31
camelia std 3b262af: OUTPUT«Potential difficulties:␤ $bar is declared but not used at /tmp/vL_lrB7Ddr line 1:␤------> class A::B { }; sub foo(A::B ⏏$bar) { }␤ok 00:02 126m␤»
FROGGS p: class Panda::Project { }; class Panda::Ecosystem { method add-project(Panda::Project $p) { } } 16:32
camelia ( no output )
FROGGS this blows up in panda and I don't know why
lizmat how? 16:34
FROGGS ===SORRY!=== Error while compiling /home/froggs/dev/panda/lib/Panda/Ecosystem.pm 16:35
Invalid typename 'Panda::Project' in parameter declaration.
moritz did you use Panda::Project?
FROGGS moritz: yes 16:36
p: class Panda::Ecosystem { method add-project(Panda::Project $p) { } } 16:37
camelia rakudo-parrot 56ee1a: OUTPUT«===SORRY!=== Error while compiling /tmp/k2b18w192W␤Invalid typename 'Panda::Project' in parameter declaration.␤at /tmp/k2b18w192W:1␤------> stem { method add-project(Panda::Project⏏ $p) { } }␤»
FROGGS okay, at least it would fail that way
hmmm, then it must be something with my CompUnitRepo stuff 16:38
16:42 c1sung joined
moritz star: use Panda::Project; class Panda::Ecosystem { method add-project(Panda::Project $p) { } } 16:46
camelia ( no output )
moritz huh, works here..
FROGGS: maybe leave out the type constraint as a workaround
16:46 Piers_ joined 16:53 tadzik joined
lizmat *sigh* 16:55
gist.github.com/lizmat/7845244
trying to make a proper QuantHash role with %!elems private role attribute
but apparently, it can not be found: package-kind: %!elems 16:56
jnthn lizmat: Oh, but there you're referring to the attribute of one role inside another... 16:57
16:57 pmurias joined
lizmat similar situation outside of settings seems to work 16:57
jnthn lizmat: But since composition is flattening...
lizmat I have a test for that, and that works
jnthn Really?
pmurias the use of STM for concurency in Perl6 was abandoned?
lizmat let me double check
pmurias: not as far as I'm concerned, but from the spec at the moment: yes
masak pmurias: I think TimToady expressed it as not being as firmly believed in as it used to.
jnthn pmurias: Yes, so far as I'm concerned. I won't get in the way of anybody trying to do it, but I don't think it'll work out too well. 16:58
lizmat jnthn: role AA { has Int $!aa }
role BB does AA {}
class Baz does BB { method baz { $!aa++ } }
my $baz = Baz.new;
works in test 16:59
pmurias it might be best to leave it to modules
jnthn lizmat: But that's not your scenario.
lizmat: Move the method baz into role BB
lizmat no? am I missing something ?
jnthn I think that's what you have? 17:00
my role Setty does QuantHash { 17:01
- has %!elems; # key.WHICH => key
method BUILD (:%!elems) {}
lizmat hmmm... so why would a role not see a private attribute in another role it does ?
FROGGS maybe the role that has the method would need some sort of "has $!aa = ..." ?
jnthn Because role composition is flattening. A role doing a role simply means "and include this at composition time" 17:02
FROGGS like when we mixin a role that depends on a method being implemented by the class
this scenario here seems related
jnthn Not so sure it's related... 17:03
lizmat jnthn: that feels like an implementation restriction, rather than a language feature
jnthn I can see why we have it. I agree its inconvenient...
17:04 jnap left
lizmat I would think that when we compose the role, it would know which attributes it got from the roles it is doing 17:04
17:04 rindolf left
lizmat just like the class will at the end 17:04
is that a strange thought? 17:05
jnthn What happens at each of those composition times is fundementally different, is all.
In the class composition, we actually concretize all the roles.
We need the target class to do so. 17:06
lizmat not disagreeing with that
but this feels like a big drawback on code reuse to me
jnthn Here's the gory bit
17:07 V_S_C joined
jnthn role A[Numeric ::T] { has $!a }; role A[::T] { }; role B[::T] does A[T] { method m() { $!a } } # That both is and isn't valid depending on the class we compose it in to 17:08
17:10 V_S_C left
lizmat ok, am too tired now to think this through 17:10
and patch -p1 is about to fold for today
so I'll see if I can wrap my head around it tomorrow 17:11
jnthn OK. I don't know an answer, I'm afraid. :( 17:12
jnthn will ponder it too
lizmat maybe TimToady / masak / moritz would like to chime in 17:13
:-)
17:14 lizmat left
jnthn I wonder if we can just store the things to check in the meta-object somewhere and do it at composition time... 17:14
Class comp, that is...
Will be "fun" go make error reporting not suck
*to
TimToady is running a quiz meet today, so won't have much brane 17:15
FROGGS I still think that "Attribute $.foo must be implemented by class Bar because a role requires it" is how it could be by stubbing the attribute in the role 17:16
17:22 atroxaper left
Woodi but we can have roles with implementation (right?) so would be nice to put attrs into r-class 17:23
FROGGS Woodi: you can basically use a role like a class, yes
Woodi and two roles can use the same attr: class User does Admin does Gamer {... 17:24
eg. password
FROGGS sure 17:25
would be weird to do it in another way
Woodi on the other hand two roles can have attrs with same names for totally other purposes :) so kind of roles groupping needed. but roles can inherit roles ? 17:29
FROGGS well, the user is responsible to choose roles that play well together 17:30
and yes, roles an do other roles
17:31 cognominal left
Woodi but what was orginal, lizmat++ problem ? :) 17:32
17:33 woolfy left 17:35 jnap joined, SamuraiJack_ left 17:36 SamuraiJack_ joined 17:39 jnap left 17:47 SamuraiJack_ left
japhb Something very weird with perl6-j (built last night): When I do `which perl6 perl6-j`, they're in the same place, .../rakudo-jvm/install/bin/ . And when I do `perl6 -v`, it works fine. But when I do `perl6-j -v`, I get bash: .../rakudo-jvm/install-jvm/bin/perl6: No such file or directory 17:48
Known issue?
(FWIW, I can't figure out where it's still looking for .../install-jvm/... . My path is clean of that, and so are the launchers, as far as I can see.
)
17:49 denis_boyun joined
raydiak perl6advent.wordpress.com/?p=1945 # anyone feel free to offer corrections or improvements 17:52
moritz japhb: sounds like something weird is going on, like you being in a working directory that has been delete, or something funny
japhb: or there's a dangling symlink involved, or something 17:53
raydiak: say $vec.length; # 1 17:55
raydiak: shouldn't that be $vec.magnitude?
japhb moritz: The only symlinks in that tree are the top level (rakudo-jvm -> rakudo-jvm-de5c8f874e6ec2f560a4538fee00661e529098f4) and libparrot.so in the parrot version of the tree
raydiak moritz: oh yeah, I renamed it...thanks :) 17:56
moritz raydiak: also minor point, the other posts so far used 4 spaces for indentation 17:57
raydiak moritz: good to know, will fix
moritz raydiak: as for the methods x, y, z: you already made them 'is rw', so there's no need for the optional argument
you can simply use them as $vec.x = 42 17:58
r: sub f($x) { say VAR($x).defined }; f 42;
raydiak moritz: but the default autogen accessors support passing the value as a param as well, don't they?
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«True␤»
moritz raydiak: no 17:59
r: sub f($x?) { say VAR($x).defined }; f
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«False␤»
raydiak moritz: hoory! I hated how it made them look when I added that param garbage
moritz r: sub f($x?) { say VAR($x).defined }; f Any
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«False␤»
moritz raydiak: also it looks to me like VAR($x).defined and $x.defined mostly do the same thing, so I would leave out the VAR() part 18:00
japhb Any other ideas on where to look for my .../install-jvm/... problem? This is driving me nutters. 18:01
raydiak moritz: from what I read, VAR.defined will return true if the value was passed, even if it's an undefined value, which is a big difference as opposed to not passing it at all
moritz p: sub f($x?) { say VAR($x).defined }; f Any 18:02
camelia rakudo-parrot 56ee1a: OUTPUT«False␤»
moritz raydiak: ^^ doesn't look like it
raydiak: otherwise very nice post. Thank you!
raydiak moritz: quite right, will make changes to that part. and thanks for the help! 18:03
moritz raydiak: you're welcome
18:10 rindolf joined
japhb AAAUUUGGGHHH I found it. PEBCAK, where the P happened a month ago. Gah. 18:14
raydiak japhb: where was it? 18:16
18:18 hummeleB1 left 18:20 hummeleBop1 joined 18:22 raiph joined
japhb raydiak: I had a month ago set an alias for perl6-j, and with the new install location (install/bin/perl6-j, rather than install-jvm/...), that alias was out of date. :-( 18:22
The bash builtin `type` is, btw, your friend in these cases.
raydiak ah ha...tricky 18:24
moritz ok, the dangling thing wasn't a symlink, but an alias :-)
raydiak tends to prefer single-line scripts in ~/bin over alias definitions 18:28
moritz tends to prefer $PATH changes over aliases and single-line scripts in ~/bin 18:30
japhb moritz: The path in question is on a network filesystem. Having every command-name tab completion go to sleep when the network FS is having trouble, because something on the network FS appears in your PATH, sucks. 18:33
raydiak we should collaborate on a project based around writing an alias to a script in ~/bin which modifies $PATH :) 18:35
18:36 jnap joined, ajr joined
japhb yipes. 18:36
18:36 ajr is now known as Guest29520
japhb Oh, raydiak, did you release your source? I see you were discussing your post with moritz .... 18:36
18:37 Guest29520 is now known as ajr_ 18:39 kivutar left 18:40 jnap left
raydiak japhb: no, this was just a small unrelated advent post. now that it's done, I can get back to pray. I was thinking about extracting Pray::Math::Vector3D into Math::Vector3D, an optimized (or will be) vector class specifically for 3D...would you be interested in digging in to that before the rest of pray is out? 18:40
I am hoping to have another advent post this month for the Pray release, though, if that's the one you mean 18:42
japhb Hmmm, how optimized a class are you planning to write? In particular, are you going to be doing all native nums and packed arrays? 18:45
raydiak it's definitely on the table when those things are a little more complete 18:46
japhb Hmmm, why does shell("echo foo") work in both r-j and r-p, but run(< echo foo >) only works in r-p?
You can already build the classes with native num attributes IIRC. 18:47
I wonder if packing them in 'is repr(CStruct)' is a net win or loss performance- and memory-wise right now .... 18:48
FROGGS japhb: run(< echo foo >) should not work at all
japhb japhb: Oh? Did it get renamed or somesuch?
18:48 darutoko left
FROGGS japhb: my bad, it should work 18:48
japhb :-)
FROGGS maybe perl6-j does not search for echo in /bin? 18:49
18:49 Mouq joined
FROGGS does that work? run(< /bin/echo foo >) 18:49
japhb Hmmm, plausible. In fact, perhaps it doesn't search the path at all?
japhb checks
raydiak iirc, when I last tried, native nums couldn't be assigned to or operated on or coerced to, or something along those lines...
FROGGS japhb: that can well be
japhb Hmmm, nope /bin/echo doesn't work either
FROGGS run() should fake the searching in path that shell does... 18:50
hmmm
japhb r: class Vec3D { has num $.x is rw; has num $.y is rw; has num $.z is rw; }; my Vec3d $v .= new(:x(0e0), :y(1e0), :x(2e0)); say $v; $v.x = 4e0; say $v; 18:52
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«===SORRY!===␤Cannot create rw-accessors for natively typed attribute '$!x'␤»
japhb Ah, I see the problem
r: class Vec3D { has num $.x; has num $.y; has num $.z; }; my Vec3d $v .= new(:x(0e0), :y(1e0), :x(2e0)); say $v; $v.x = 4e0; say $v; 18:53
FROGGS japhb: I have no idea why run(< /bin/echo foo >) fails :/
camelia rakudo-parrot 56ee1a: OUTPUT«===SORRY!===␤Type 'Vec3d' is not declared. Did you mean 'Vec3D'?␤at /tmp/3BlTY8skhq:1␤------> x; has num $.y; has num $.z; }; my Vec3d⏏ $v .= new(:x(0e0), :y(1e0), :x(2e0)); s␤Malformed my␤at /tmp/3BlTY8sk…»
..rakudo-jvm 56ee1a: OUTPUT«===SORRY!===␤Type 'Vec3d' is not declared. Did you mean 'Vec3D'?␤at /tmp/C57Za9JAyV:1␤------> x; has num $.y; has num $.z; }; my Vec3d⏏ $v .= new(:x(0e0), :y(1e0), :x(2e0)); s␤Malformed my␤at /tmp/C57Za9JAyV:…»
japhb FROGGS: Dang it. :-(
raydiak r: say int(1).perl
camelia rakudo-parrot 56ee1a: OUTPUT«invoke() not implemented in class 'int'␤ in block at /tmp/Ghrvn5ep2c:1␤ in any at /tmp/Ghrvn5ep2c:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPHLL.nqp:1133␤ in any evalfiles at gen/parrot/stage2/NQ…»
..rakudo-jvm 56ee1a: OUTPUT«Can not invoke this object␤ in block at /tmp/IszYDeWtJ3:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:1086␤ in any evalfiles at gen/jvm/stage2/NQPHLL.nqp:1292␤ in any command_eval at gen/jvm/stage2/NQPHLL.nqp:1196␤ in any command_eval at src/P…»
japhb r: class Vec3D { has num $.x; has num $.y; has num $.z; }; my Vec3D $v .= new(:x(0e0), :y(1e0), :x(2e0)); say $v; $v.x = 4e0; say $v;
camelia rakudo-jvm 56ee1a: OUTPUT«Vec3D.new(x => 2e0, y => 1e0, z => 0e0)␤Cannot modify an immutable value␤ in block at /tmp/_pvt4yGPJj:1␤ in any eval at gen/jvm/stage2/NQPHLL.nqp:1086␤ in any evalfiles at gen/jvm/stage2/NQPHLL.nqp:1292␤ in any command_eval at gen/jvm/stage2/…»
..rakudo-parrot 56ee1a: OUTPUT«Vec3D.new(x => 2e0, y => 1e0, z => NaN)␤Cannot modify an immutable value␤ in block at /tmp/43hiqshbNZ:1␤ in any at /tmp/43hiqshbNZ:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPHLL.nqp:1133␤ in an…»
japhb p: class Vec3D { has num $.x; has num $.y; has num $.z; }; my Vec3D $v .= new(:x(0e0), :y(1e0), :z(2e0)); say $v; $v.x = 4e0; say $v; 18:54
camelia rakudo-parrot 56ee1a: OUTPUT«Vec3D.new(x => 0e0, y => 1e0, z => 2e0)␤Cannot modify an immutable value␤ in block at /tmp/ATF6p_KHs3:1␤ in any at /tmp/ATF6p_KHs3:1␤ in any at gen/parrot/stage2/NQPHLL.nqp:1146␤ in any eval at gen/parrot/stage2/NQPHLL.nqp:1133␤ in an…»
japhb Yup, that's a killer all right.
raydiak I suspect this sort of thing will become higher-priority as native call for jvm comes online
japhb I would say just doing everything in immutable/functional style, but that doesn't really work for high-performance rendering work (too much GC churn) 18:56
moritz yes, natives can't be rw yet
you'd have to write separate setters 18:57
japhb OOC, what is the difficulty with making it work? Is there an architectural issue, or just a lack of tuits? 18:58
moritz well, a native by defition doesn't have a container
and rw-ness works through containers
so you'd basically have to invent a separate calling convention, where a pointer to the native is passed instead 18:59
and then try to know at compile time whether that's the case, and guard at runtime for cases where you can't know
jnthn The idea is that we'll have a kind of container that points to the object and its attribute, and its assign handling will do the right thing... 19:01
timotimo Can't use unknown trait 'is equiv' in a sub+{precedence} declaration. o_O
jnthn And then rely on inlining to optimize that cost away.
In the common case, anyways. 19:02
japhb nodnod
timotimo jnthn: do you think it makes sense to rewrite some cases of junctions into QAST-constructed for-loops? 19:10
like $n %% any(@primes), for example
jnthn timotimo: No 19:18
raydiak japhb: to give you a more detailed answer to your original question (and kinda half making notes to myself), I've been agonizing over that advent post for a couple days, so I haven't touched Pray since the last image you saw except to get scaling half working, but iirc everything I want done before release is only 2 or 3 days worth of work...it's just shoehorning 2 or 3 full days of work into X number of RL 19:21
days that gets a little sticky
wow, that didn't look so long when I could only see one line at a time :)
FROGGS weird, with my CompUnitRepo stuff, I can do "use Test; ok 1", but not "use Test; say Test" 19:23
timotimo jnthn: happens too rarely? analysis too expensive? code-size increase not worth it?
jnthn timotimo: Yeah, I didn't see that pattern show up so much in real code. 19:25
With the optimizer, we have to remember that it takes time to run.
timotimo sd.jtimothyking.com/2013/09/17/how-...is-perl-6/ - search for %%-any here :P
jnthn So if we make it do more work, we need to be fairly confident it will pay off
timotimo but yeah.
i *think* we already have analysis that goes there partially 19:26
like, we'll already look at that piece of code and say "it's in bool context, it has a junction on one side. oh, the junction has an array argument. i guess i'll give up then"
that either means that doing that loop transformation is not much more costly, or that the existing analysis is too costly as well :) 19:27
19:28 nebuchadnezzar left 19:29 Piers_ left 19:32 Piers_ joined 19:36 jnap joined 19:41 jnap left 19:47 nebuchadnezzar joined 19:49 Piers_ left 19:53 raiph left, sqirrel joined 19:56 rom1504_ is now known as rom1504 20:01 thou joined 20:17 rindolf left 20:19 sqirrel_ joined, sqirrel left 20:20 Piers_ joined 20:23 sqirrel_ left 20:24 benabik joined 20:28 jeffreykegler left
timotimo not convinced, jnthn? 20:30
jnthn timotimo: I think we need to look deeper into why junction auto-threading is so slow. 20:32
timotimo but that's hard :) 20:34
20:36 sqirrel joined 20:37 jnap joined 20:39 colomon joined 20:42 rurban2 joined, jnap left, sqirrel_ joined, sqirrel left
colomon jnthn: answering your question from this morning very late, JVM time: Stage parse : 48.800 20:43
20:44 rurban1 left
timotimo is munch not specced? 20:45
r: my @a = 1, 2, 3, 4, 5; say @a.munch(3); say @a; 20:46
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«1 2 3 4 5␤␤»
FROGGS here is how my CompUnitRepo explodes: gist.github.com/FROGGS/392471225039d60749d7 20:47
TimToady raydiak: subscripts are no longer methods 20:50
20:51 kaare__ left, thou left
TimToady your article thinks they're inheritable, and they aren't 20:51
(they turned out to be too hard to optimize that way)
20:52 raiph joined
raydiak TimToady: what would be more accurate to say there? or should I just remove the offending statement entirely? 20:52
and why does my class support subscripts? where *do* they come from?
TimToady how old is your rakudo? 20:53
raydiak hrm
well, I can't know that for my readers
TimToady I think I'd just remove that statement, since operators aren't inheritable in that sense either 20:54
operators are all lexically scoped, since they define the language
some operators will delegate to the object's method underlyingly though
raydiak ah, I see...I had it backwards and inside-out 20:55
ooc, assuming a recent pull of nom, where do the subscripts come from? 20:56
TimToady on the binding vs assignment, I missed seeing a statement that the Array would end up containing a single element, which is the Vector you made
they come from the setting
as do all other operators
raydiak got it...I think 20:57
timotimo TimToady: there's no mention of "munch" in the specs, is it deliberately unspecced?
(also, i don't think i really understand what it's supposed to do)
raydiak r: class Foo is Array {}; my @foo = Foo.new(1,2,3); say @foo.perl; 20:58
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«Array.new(1, 2, 3)␤»
raydiak ^ no Foo at all
TimToady well, we're waiting for pmichaud++ to finish respeccing that, but he's been busy with his kids, and rightfully so
timotimo ah, right
20:58 sqirrel_ left 20:59 sqirrel joined
TimToady hmm 20:59
raydiak it's working like it says in the article, the part that shows what not to do
20:59 sqirrel left
TimToady r: class Foo is Array {}; say Foo.new(1,2,3).WHAT; 21:00
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«(Foo)␤»
raydiak create an array, then assigns the list of values entirely ignroing the class on the right
TimToady r: class Foo is Array {}; my @foo = Foo.new(1,2,3); say @foo.elems 21:01
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«3␤»
TimToady hmm, I guess it flattens anyway
timotimo yeah, the .perl method just writes "Array.new" regardless wether or not it's actually an array
timotimo whips up a quick fix patch
Mouq colomon: what was it you were asking about parsing grammars?
timotimo oh, what is this
TimToady yes, it's just interpolating the Foo into the list flat, which is okay, I guess
raydiak b/c it does Positional? 21:02
timotimo someone else needs to look at this :\
colomon Mouq: I'm trying to figure out how to fix sigspace in Niecza. and the first step is just trying to figure out how it works at all... 21:04
21:05 kivutar joined
timotimo is always happy to hear work on niecza happening 21:05
raydiak not sure if this is at all related to the other behavior, or just wishful thinking, but it'd be nice if "my Vector @vec = 1, 2, 3" DWIMmed
TimToady that would be an array of Vectors 21:06
you want my @vec is Vector = 1,2,3 for the other thing
21:06 dwarring joined
raydiak oh right...maybe I should have been trying my @vec is Vector :P 21:06
TimToady the return type of a composite is always the individual type
raydiak I find myself making that mistake a lot 21:07
Mouq colomon: well, basically, space is only significant after a construct that actually does matching. '{1} ', ':my $a = 5; ', ':i' aren't. '<a> ', 'abc ' is. Spaces can come before and/or after a quantifier, in which case they mean different things: 'a +' is ':!s [a <.ws>]+', 'a+ ' is 'a+<.ws>'
21:08 Fatalnix joined
Fatalnix hey guys 21:08
diakopter howdy
Mouq *':i '
colomon Mouq: yes, I more or less understand that. I'm wondering how Niecza implemented the old method of doing it.
Mouq Oh.
21:09 benabik_ joined
Mouq Rakudo's method was weird. I *think* it captured spaces as a seperate metachar before, but I really didn't understand what it was doing nor why it worked 21:10
Fatalnix gonna try getting the JVM backend to work again
never compiled last time heh
is there a document on the features Parrot vs JVM differ? 21:11
for rakdo
Mouq The way it does it now is based on parse-time variables, and then all the actions have to do is if $<sigspace> or whatever is defined, attach it to the appropriate thing
Fatalnix rakudo*
21:11 zakalwe_ joined, zakalwe_ left, zakalwe_ joined 21:12 ggoebel112 left, benabik left, ggoebel112 joined, pmurias left, PerlJam left, benabik_ is now known as benabik
Mouq Although I want to change that to be entirely in the actions, which will allow for :s($not-known-at-parse) to work 21:12
raydiak 'is TypeObject' NYI...now I remember why that pattern didn't sink in when I was studying
21:12 PerlPilot joined
Mouq r: say 'AbCd' ~~ /:i abcd/ 21:13
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«｢AbCd｣␤␤»
Mouq r: say 'AbCd' ~~ /:i(True) abcd/ # :(
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«Nil␤»
Mouq r: say 'AbCd' ~~ /:i(1) abcd/ # ? 21:14
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«｢AbCd｣␤␤»
Mouq Ahhh, compile time value
jnthn Fatalnix: Not really; the big difference is Rakudo on JVM has support for concurrency stuff. Differences besides that are minor.
Fatalnix ah 21:15
is there a way do you know to do annotations on the JVM? Like Java's annotations
I wanna try and see if I can get Forge for Minecraft to just load a Perl6 program and say something
as a proof of concept
Forge requires annotations to setup the mod. though I can always make a Java main class if I had to. 21:16
jnthn You'll probably be the first person to try and call Perl 6 from Java. While the other way around (call class libraries from Rakudo) has had some work, the other way really hasn't yet. 21:17
Fatalnix oh really? I figured that the JVM took care of most of the bumps 21:18
dwarring I've written a demo grammar to parse and rank poker hands - gist.github.com/dwarring/7848868 21:20
lue r: sub foo (:$bar, :$baz) { say $bar, $baz }; foo(:bar, :baz); foo(:bar :baz); # shouldn't the second call fail with TTIAR, or alternatively work like the first call? 21:22
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«TrueTrue␤True(Any)␤»
FROGGS dwarring: nice!
lue Or is that supposed to be understood by Perl 6 as an adverb applied to an adverb? :)
jnthn lue: Work like the first, but multi-colon thingies are NYI
dwarring just for fun 21:23
lue jnthn: OK. I guess I'm surprised then that it doesn't fail with TTIAR or NYI in the meantime.
Mouq &
dwarring i wrote a prolog version quite a few years ago 21:24
Fatalnix lol prolog
the almost useless Erlang
:D
dwarring this is better 21:25
Fatalnix how long do you think it would take to compile this JVM bild with make -j 5 (if it even usesthat) and a dual core hyperthreaded i5 @ 2.4 Ghz?
jnthn dwarring: Nice 21:26
Fatalnix I almost said Mhz lol
oh, looks like its done already
dwarring could write it up for advent 21:28
dwarring but not for tomorrow
jnthn dwarring: I think there's some open days a bit further ahead 21:29
dwarring i'll start on a write up 21:30
21:30 raiph left
Fatalnix all tests pass! 21:31
21:33 raiph joined
Fatalnix ls 21:36
oops
moritz no such file or directory :-)
Fatalnix haha
21:36 benabik left
Fatalnix trying to find out if there is a way to compile the perl6 code into a .class 21:37
21:38 jnap joined, benabik joined
jnthn Fatalnix: perl6-j --target=classfile --classname=Foo --output=Foo.class something.p6 21:39
or .pm
Fatalnix ohhh
hmm
no wonder I couldnt find perl6
:D
its perl6-j 21:40
lets see if this works
raydiak TimToady: made changes to text and code to skirt around the blurry parts and not subvert expectations if the flattening behavior changes in the future; any more feedback welcomed, as I might be leaving soon 21:41
Fatalnix classname doesnt exist but lemme look at the usage here
raydiak ^^ same for anyone else with suggestions :)
FROGGS raydiak++ # very nice post btw 21:42
21:42 jnap left
raydiak thanks! glad you like it 21:44
21:51 Mouq left, Rotwang joined, ajr_ left 21:53 woolfy joined 21:54 lizmat joined 21:55 rurban2 left 21:58 thou joined 22:00 colomon left 22:04 colomon joined, rurban1 joined 22:13 telex left
raydiak timotimo: sorry, I missed what you were telling me about .perl earlier, just got it reviewing the log 22:13
r: class Foo is Array {}; my @foo := Foo.new; say @foo.perl;
camelia rakudo-parrot 56ee1a, rakudo-jvm 56ee1a: OUTPUT«Foo.new()␤»
raydiak timotimo: ^
22:17 telex joined, FROGGS left
timotimo k 22:18
22:18 FROGGS joined
lue Just added some Unicode information functions and a storage-method query to Str in my provisional S15, thought it was important enough to mention: gist.github.com/lue/7761244 22:19
22:19 thou left 22:20 spider-mario left 22:30 denis_boyun left 22:36 FROGGS left 22:37 pmurias joined 22:38 FROGGS joined 22:39 jnap joined, rurban1 left 22:43 jnap left 22:44 rurban1 joined 22:48 cognominal joined 23:04 FROGGS left 23:06 FROGGS joined 23:12 xenoterracide joined 23:14 pmurias left 23:16 pmurias joined 23:18 thou joined 23:19 jnap joined 23:21 Mouq joined 23:22 FROGGS left 23:23 FROGGS joined 23:25 xenoterracide left, Mouq left 23:30 hummeleBop1 left 23:31 FROGGS left, woolfy left
jnthn sleep & 23:33
23:38 dmol left 23:55 Piers_ left 23:57 FROGGS 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!