»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg p6eval perl6: ... | irclog: irc.perl6.org/ | UTF-8 is our friend!
Set by sorear on 4 February 2011.
masak 'night, #perl6 00:15
TimToady 'night
quietfanatic nught 00:16
00:21 skids joined
jnthn 'night o/ 00:32
00:50 replore joined
clkao /wnWW/WIN21 00:52
01:00 zhutingting joined 01:10 Guest65764 left 01:21 fgomez joined 01:22 drbean left 01:25 drbean joined 01:35 zhutingting left, replore left 02:03 orafu left 02:04 orafu joined 02:09 replore joined 02:16 tokuhiro_ left 02:32 replore left 02:37 Psyche^ joined 02:41 Patterner left, Psyche^ is now known as Patterner
dalek ast: e8c9ac5 | coke++ | S10-packages/scope.t:
pugs fudge
 02:55
03:00 xinming left
takadonet good job coke :) 03:00
03:02 xinming joined 03:19 crab2313 joined 03:24 jaldhar joined
[Coke] takd? 03:25
takadonet yes?
03:38 jaldhar left 03:41 jaldhar joined
TimToady nr: say ['a','b'] X~ 1 .. 3; 04:00
p6eval rakudo f23a45: OUTPUT«a1 a2 a3 b1 b2 b3␤»
..niecza v19-15-g051783d: OUTPUT«a b1 a b2 a b3␤»
TimToady niecza++
rakudobug
nr: say ['a','b'],() X~ 1 .. 3; 04:03
p6eval rakudo f23a45, niecza v19-15-g051783d: OUTPUT«a b1 a b2 a b3␤»
TimToady nr: say ['a','b'] Z~ 1 .. 3;
p6eval rakudo f23a45: OUTPUT«a1 b2␤»
..niecza v19-15-g051783d: OUTPUT«a b1␤»
TimToady same deal, rakudobug
04:04 ironcamel joined
TimToady nr: my $a = ['a','b']; say $a X~ 1..3 04:04
p6eval rakudo f23a45: OUTPUT«a1 a2 a3 b1 b2 b3␤» 04:05
..niecza v19-15-g051783d: OUTPUT«a b1 a b2 a b3␤»
TimToady same deal
TimToady is kinda surprised we haven't noticed this before 04:06
nr: say 1..3 X~ ['a','b'] 04:09
p6eval rakudo f23a45: OUTPUT«1a 1b 2a 2b 3a 3b␤»
..niecza v19-15-g051783d: OUTPUT«1a b 2a b 3a b␤»
TimToady nr: say ([1,2] X [3,4]).perl 04:13
p6eval niecza v19-15-g051783d: OUTPUT«([1, 2], [3, 4]).list␤»
..rakudo f23a45: OUTPUT«((ListIter.new(),), (ListIter.new(),), (ListIter.new(),), (ListIter.new(),)).list␤»
TimToady seems to be a generic problem with list infixes not binding both sides like slurpy lists 04:14
TimToady suspects this may be another manifestation of the problem that pmichaud++ is already working on 04:15
04:25 benabik left 04:31 benabik joined 04:37 quietfanatic left 04:39 provider joined, provider left 04:42 crab2313 left 04:45 kaare_ joined 04:47 tokuhiro_ joined 04:58 skids left 05:37 adu joined
moritz \o 05:43
sorear o/ 05:49
TimToady backlogged! 06:42
'course, now I'm about to go to bed and earn some more backlog...
moritz sleep well, dream of short backlogs 06:52
TimToady thanks, o/
06:56 am0c joined
sjohnson hi 07:23
07:25 kaxing joined
dalek ar: 12b921c | pmichaud++ | skel/tools/build/ (3 files):
Refactor .msi building process a fair bit.
 07:39
ar: a280cf5 | pmichaud++ | skel/tools/build/star-product.wxs:
Change INSTALLLOCATION -> INSTALLROOT
ar: f2057b0 | pmichaud++ | skel/tools/build/Makefile.in:
Fix directory argument for building source .wxs file.
07:56 GlitchMr joined 08:01 benabik left
pmichaud I'm about to publish the 2012.07 star release; anything I should look at or check before doing so? 08:04
(bbi10) 08:05
08:10 sergot joined
sergot hi o/ 08:10
08:12 cognominal_ joined, plobsing left 08:14 cognominal left
pmichaud back again 08:23
moritz \o sergot, pmichaud
pmichaud okay, time to upload the release files 08:25
moritz pmichaud: does the MSI support installing to arbitrary paths? 08:26
pmichaud moritz: no.
(because rakudo/parrot can't support it) 08:27
moritz ok
thought so :(
pmichaud indeed, the msi would've likely been a bit easier if we could've done that 08:28
(easier to build, that is)
moritz another reason to investigate alternative backends 08:40
pmichaud release published, announcements made to rakudo.org, p6c, p6u, and parrot-dev
moritz pmichaud++ 08:41
pmichaud: do you think it would be beneficial to upload a binary linux build?
pmichaud I think if we do that, we'd want to package it as a .deb and/or .rpm 08:42
felher If i want to have a enumeration of JAN/FEB/... to Date.new("2012-01-01"), is this how i'm supposed to to it:
r: enum Months (JAN => BEGIN {Date.new("2012-01-01")}); JAN.year.say;
p6eval rakudo f23a45: OUTPUT«2012␤»
felher pmichaud++ :)
*to do it
moritz r: enum Months (Jan => Date.new("2012-01-01"), Feb => Date.new("2012-02-01")); 08:43
p6eval rakudo f23a45: OUTPUT«===SORRY!===␤Enumeration value must be known at compile time␤at /tmp/9UjunzXE62:1␤»
moritz I guess I should fix that
r: try eval 'enum Months (Jan => Date.new("2012-01-01");'; say $!.WHAT
p6eval rakudo f23a45: OUTPUT«X::AdHoc()␤»
pmichaud so, I've been considering publishing our own .deb and/or .rpm distros... but wanted to get the windows world handled first. 08:44
we'd also have to figure out how we'd want to coexist with the existing debian packaging, if at all.
moritz which is why I'd simply go with tar'ing up /opt/rakudo-star-2012.07
pmichaud I fear that if we did that people would then say "...why aren't you providing it as a .deb/.rpm ?" 08:45
moritz then we can answer "because we haven't figured it out yet" 08:46
pmichaud oh, but someone's figured it out, because rakudo .debs are already available.
they just tend to lag our release cycle by a lot 08:47
moritz well, I'm trying to be pragmatic 08:49
pmichaud sure
moritz some people have trouble building rakudo due to memory requirements
and we're not currently in a position to provide polished .deb/.rpm packages 08:50
pmichaud /opt/rakudo-star-2012.07 would include the parrot binaries?
moritz hence my proposal to give them *something*
pmichaud: yes
pmichaud I'll think about it a bit more. 08:52
masak pmichaud++ # R*
pmichaud having done the .msi stuff, and having done a few .debs and .rpms before (for other packages, not rakudo), I suspect we can come up with a way to handle .deb/.rpm without too much difficulty.
so, let's revisit it for august (perhaps around yapc::eu) 08:53
masak (morning, #perl6)
pmichaud since none of us are apparently giving talks, we may have time to work on other things :-/
masak pmichaud: is it now final that none of us are giving talks?
if so, that sucks.
pmichaud masak: I've not heard anything final, no. 08:54
masak writes to the organizers with some nice but very precise wordings
pmichaud but the longer things go without hearing anything changing, the less likely I think it is
moritz masak: the last response from the organizers was that they'll try to figure out how to enable more talks
masak: please do, and set me CC
masak will do.
will cc pmichaud and jnthn too.
moritz you might also mention that people in here have pondered not going to YAPC::EU due to absense of Perl 6 talks 08:55
and I'm not talking about the core hackers here
pmichaud I can entirely understand that it may have been a statistical or logistical fluke that caused us to end up with no talks there. And I know that my talk proposal wasn't very solid or concrete. 08:56
I can still update it and make it more solid or concrete; but things got a little hectic here as the talk deadline approached. 08:57
masak pmichaud: "Currently it seems as if some talk votes were counted differently from the expectations." blogs.perl.org/users/moritz/2012/07...ent-179251
pmichaud masak: yes, I saw that note. I'm not exactly sure what that meant. :-)
masak pmichaud: i.e. "we screwed up and accepted some talks before the ones we should've accepted, and now we can't take it back and we have a problem" 08:58
pmichaud well, 49 talks seems like a fairly smallish number relative to yapc's past 08:59
(too)
I'm guessing they only have three tracks/rooms
masak email sent. 09:10
moritz masak++
I haven't reached my CC yet, but email isn't real-time
s/reached/received/
moritz should learn English one of these days
huf did i wake up in another language again? :) 09:11
masak oops, sent to questions@yapceu2012, not [email@hidden.address] 09:12
re-sending, re-cc:ing. sorry 'bout that.
pmichaud fwiw, yapc::na will be in Austin next year, and here's the first topic request that came on their mailing list: github.com/rakudo/star/blob/master...ce/2012.07 09:15
Essentially, "Anyone have an idea on the topics or panel? Would love to see Perl 6
discussions."
oops, wrong url
mail.pm.org/pipermail/austin/2012-J...01807.html
anyway, I should get some sleep here -- bbiaw 09:16
dalek ast: 17c51d7 | moritz++ | S32-exceptions/misc.t:
remove an exception test where we do not strictly need to die
 09:21
kudo/nom: 7d101ef | moritz++ | src/core/List.pm:
simplfy List.classify, make it return a Hash as per spec
kudo/nom: b87659a | moritz++ | src/Perl6/Actions.pm:
compile-time evaluate enums

instead of complaining that they are not known at compile time		 09:22
moritz felher: ok, no more need for BEGIN magic in enums :-)
Coleoid pmichaud++ # MSIs help me tempt friends... 09:24
felher moritz++ # thanks a lot :) 09:25
jnthn morning, #perl6 09:29
moritz \o jnthn
09:30 hoelzro|away is now known as hoelzro
jnthn pmichaud++ # Rakudo release 09:31
*Star
hoelzro morning perl6 folk
oh hey, release!
hooray
pmichaud++ 09:32
masak: when you built that R* bundle for me, you just used the pristine R* 2012.06 archive, right? Did you provide any special options to Parrot's Configure.PL? 09:33
masak no. 09:36
I did not configure Parrot specifically. just R* itself.
au ponders making a rakuto-star-2012.07.pkg 09:42
hoelzro masak: any special options to Configure.PL there? other than --prefix?
I ask because when I compile Parrot myself, I get the crash 09:43
but when I install your parrot and compile everything else, it works fine
I just want a working Rakudo =(
09:47 MayDaniel joined
masak I did `$ perl Configure.pl --gen-parrot`, just like the README says. actually, I just followed the README the whole time. 09:49
moritz star: say 1 09:50
hoelzro hmm
p6eval star 2012.06: OUTPUT«1␤»
hoelzro why does it crash for me?! 09:51
even the Parrot from Arch's repo causes the crash
moritz star: say 1
p6eval star 2012.07: OUTPUT«1␤»
09:52 whiteknight joined, whiteknight is now known as Guest74836 10:00 cognominal_ left 10:06 Patterner left 10:07 tokuhiro_ left
hoelzro wtf, now it works?! 10:15
10:18 spider-mario joined 10:20 crab2313 joined
masak oh, good. 10:23
10:24 Psyche^ joined, Psyche^ is now known as Patterner
masak lol, I blogged! \o/ strangelyconsistent.org/blog/july-2...-and-dying 10:25
hoelzro masak: I think I figured out the issue on my Arch box =) 10:26
we'll know in about 20 minutes
10:28 cognominal joined
masak people, this is an excellent opportunity to find bugs in the game -- after I finished the whole storyline, but before I've started general cleanup and refactoring. 10:28
download the game at github.com/masak/crypt/ and start finding deficiencies. I wrote it, so I can guarantee that there are a few! 10:29
hoelzro goes to try it 10:30
masak \o/
hoelzro masak: I noticed it uses the executable 'nom'; can I run this with R*? 10:31
10:33 benabik joined
moritz masak: is it normal that the game takes ages to load? 10:35
hoelzro is pmichaud still around? or past his bed time =) 10:36
moritz 11:16 < pmichaud> anyway, I should get some sleep here -- bbiaw
hoelzro moritz: thanks
masak moritz: yes.
hoelzro what time zone is pmichaud in, anyway?
masak moritz: this will get lots better once I break the 3kLOC file into modules. 10:37
hoelzro: you can. I have a /usr/bin/nom on my system, but it's running latest Rakudo.
hoelzro masak: ok, cool
masak er, /usr/local/bin/nom 10:38
10:38 crab2313 left
hoelzro should I just symlink it to $(which prel6)? 10:38
s/re/er/
moritz hoelzro: you can just run perl6 bin/crypt
hoelzro oh, good point =)
masak one of the talk proposals I submitted for YAPC::EU is about how I wrote this game and what I learned writing it. 10:40
hoelzro I was hoping to do the same with my chat bot 10:42
moritz ETOOLATE this time 10:43
hoelzro ok, here's what is causing the crash on my Arch box 10:57
parrot is being built without --optimize
and that causes problems...for some reason. 10:58
11:00 zby_home_ joined, adu left
Woodi afternoon 11:07
11:10 crab2313 joined, GlitchMr left
jnthn hi Woodi 11:11
masak Woodi! \o/
Woodi pmichaud: maybe we can create (deb|rpm)s into /usr/local ? probably Debian/Ubuntu/Fedora/... package maintainers want to compile their own packages to keep it under their control... 11:12
hallo jnthn++ & masak++ :)
btw. it is nice your boss(es) officially support Perl 6 :) 11:13
11:16 GlitchMr joined
masak that is nice. 11:16
11:16 crab2313 left
jnthn Aye :) 11:19
11:20 spider-mario left 11:24 jferrero left 11:26 hoelzro is now known as hoelzro|away
moritz they do? 11:28
ah yes, it says Perl 6 on www.edument.se/konsulter/
star: use LWP::Simple; say LWP::Simple.get('perl6.org/').substr(0, 20) 11:29
p6eval star 2012.07: OUTPUT«<?xml version="1.0" ␤»
moritz star: use LWP::Simple; say LWP::Simple.get('perl6.org/').match(rx{'<title>' <( .*? )> '</title>' }).Str
p6eval star 2012.07: OUTPUT«Perl 6␤» 11:30
moritz star: use LWP::Simple; say LWP::Simple.get('rakudo.org/').match(rx{'<title>' <( .*? )> '</title>' }).Str
p6eval star 2012.07: OUTPUT«rakudo.org | Rakudo Perl 6␤»
moritz jnthn: it's funny that module precompilation is so sensitive, and still it works to load a module that was precompiled with a different setting 11:31
I guess that's because the "different" setting is still being loaded for p6eval 11:32
masak moritz: our bosses are also fine with us working about 80% so that we'll have time to do Perl 6 the remaining 20%.
jnthn Not to mention disappearing for conferences... ;-) 11:34
moritz: Different source or different compilation? 11:35
moritz: It uses a sha-1 of the source code to determine "does it match" 11:36
mhasch masak: you do what you do for Perl 6 in just 20% of your time? Or is ist 80% of your leisure time and 20 % of your working time?
masak mhasch: I try to allocate that time for the macro grant. 11:38
with varying success. :)
mhasch: the 20% fall inside the working week. in order to get the velocity right, I like to think of another 20% as falling in the weekend. I'm a bachelor, so that tends to work. 11:39
moritz jnthn: well, p6eval runs with --setting=RESTRICTED
jnthn: but around it is still CORE, with which the modules were compiled 11:40
but of course the modules still see CORE
jnthn moritz: Right.
moritz which is why it probably works
jnthn Yes. :)
mhasch masak: I see. It still takes a lot of focus to switch between different projects so frequently. 11:42
11:43 UncleFester6 joined
UncleFester6 my "make install" has module precompilation failures for lots of modules under both cygwin and low memory linux (using parrot -G) 11:45
Missing or wrong version of dependency 'src/gen/BOOTSTRAP.pm' (cygwin)
Missing or wrong version of dependency 'src/gen/CORE.setting' (linux) 11:46
masak mhasch: yes, but in a sense the work I do is already switch-intensive in that way.
and mixing a bit of Perl 6 into my regular work seems to make my regular work more interesting, too. :)
jnthn actually can't concentrate on just one thing all the time for prolonged periods, so the variety works well :) 11:47
arnsholt jnthn: Is there anything special I should know about the hash() not-quite-literal in NQP?
jnthn arnsholt: As in nqp::hash?
arnsholt: I guess that it expects string key/object value/string key/object value/... 11:48
arnsholt: Other than that, I don't think there's anything particularly special.
UncleFester6: Are you sure that it's not older .pir files?
arnsholt Hmm. In World.pm there's a sub called just as hash() is that something else? 11:49
jnthn arnsholt: Hmm :)
arnsholt: Oh...
yes, and I just saw its implementation and wtf...
arnsholt: Anyway, it's correct, just inefficient 11:50
arnsholt: But unlike nqp::hash it expects you to pass named args.
dalek kudo/macros-d2: c7c118d | masak++ | src/ (2 files):
can parse unquotes in quasis

This now works:
   $ ./perl6 -e 'say "OH HAI"; my $a; {{{$a}}}'
   ===SORRY!===
   Can only use unquote syntax inside a quasi block
   at -e:1
   $ ./perl6 -e 'say "OH HAI"; my $a; quasi { say "woo!"; {{{$a}}} }'
   OH HAI		 11:52
UncleFester6 I unpacked rakudo star into a fresh directory with nothing else in both cases ...
masak requests feedback on that initial commit, especially from moritz, jnthn, pmichaud but also from others
jnthn star: say @*INC
p6eval star 2012.07: OUTPUT«/home/p6eval/.perl6/lib /home/p6eval/star/lib/parrot/4.6.0/languages/perl6/lib␤»
arnsholt Right. 'Cause for some reason my latest set of changes in Rakudo causes a segfault when compiling the setting, and the changes are pretty much just changing World.create_signature to take a hash and some hash literals 11:53
jnthn UncleFester6: ^ shows that it will look in your home directory's .perl6 before the place that I guess modules are installed to.
UncleFester6 that could be a problem - checking 11:54
11:58 crab2313 joined
arnsholt Arg barg! Why U segfault? 11:59
jnthn arnsholt: Hm, then what do you do with that data? 12:03
arnsholt Not much ATM 12:04
Just unpack the parameters key into a variable in World.create_signature 12:05
jnthn arnsholt: If you gist me a diff I can take a look at it and see if I spot anything. 12:06
12:07 Chillance joined
arnsholt jnthn: gist.github.com/3193047 12:08
moritz masak: looks fine to me
masak huh. 12:12
ok, then. :)
I'll just carry on in that direction.
jnthn arnsholt: Hmm, nothing jumps out at me, other than that I guess I'm not looking at the full patch?
arnsholt: e.g. I see a "+ $*W.create_signature(%signature_info)", but not where %signature_info is being set up. 12:13
arnsholt Yeah, there are two more commits that lead up to this
jnthn arnsholt: OK, and this is the "switchover" one? 12:14
arnsholt Yeah
jnthn I'm guessing it's possible that the issue is in a previous patch that is building up the data structure...that or I miss the issue in this one.
arnsholt: How quickly does it segfault?
arnsholt Pretty quickly. In the parse bit 12:15
12:15 spider-mario joined, dbr left
jnthn arnsholt: I mean, I can imagine it getting very upset if it thinks it is about to get a parameter and then doesn't. 12:16
That is, if the @parameters somehow ends up containingg something that's not a Parameter.
12:16 dbr joined
UncleFester6 star: use LWP::Simple; say LWP::Simple.get('rakudo.org/').substr(0, 200) 12:17
p6eval star 2012.07: OUTPUT«<!DOCTYPE html>␤<html dir="ltr" lang="en-US">␤<head>␤<meta charset="␤UTF-8" />␤<title>␤»
UncleFester6 star: use LWP::Simple; say LWP::Simple.get('rakudo.org/').substr(0, 2000)
p6eval star 2012.07: OUTPUT«<!DOCTYPE html>␤<html ␤dir="ltr" lang="en-US">␤<head>␤<meta charset="␤»
arnsholt Hmm. Troo
jnthn: Refresh and you'll have git diff HEAD^^
moritz took a while to not parse that as a smily :-) 12:19
but it's the same as HEAD~2 12:20
jnthn arnsholt: + my $p_sig := $*W.create_signature(hash(nqp::hash('parameters', [$*W.create_parameter(@p_params[0])]))); 12:21
arnsholt: That looks like a hash in a hash where it shouldn't be
arnsholt: ah, think I see something serious. 12:23
+ sub create_signature_object($/, %signature_info, $lexpad, :$no_attr_check) {
Here we get %signature_info, and then
for %signature_info<parameters> {
We loop over what I guess are the hashes
For each one we create a parameter object and then @parameters.push($param_obj); 12:24
But then:
- $*W.create_signature(@parameters)
+ $*W.create_signature(%signature_info)
That's being passed to create_signature in the hash there is not the list of parameter objects; it's the list of hashes.
arnsholt Oh derp. Thanks! 12:25
jnthn masak: c7c118d looks good to me. 12:27
masak \o/
$*IN_QUASI is not a STD-ism. but STD derives a grammar Quasi there, and I'm not prepared to do that right now. 12:28
UncleFester6 jnthn: thanks that was it - think I came across an LWP::Simple bug but will deal with that later 12:32
moritz star: use LWP::Simple; say LWP::Simple.get("perl6.org/").chars 12:35
p6eval star 2012.07: OUTPUT«1755␤»
moritz std: {{{1}}} 12:37
p6eval std e52e3ca: OUTPUT«ok 00:00 40m␤»
UncleFester6 moritz: looking at the LWP::Simple code it seems pretty clear it is only pulling in the first packet of the reply and not later packets. Will double check and plan to file later. 12:38
12:38 cognominal_ joined
masak moritz: it probably parses as three blocks outside of Quasi. 12:40
moritz: personally, I'm fine with not allowing it.
moritz masak: +1
masak it's a bit like a `return` outside of a routine.
12:41 cognominal left
moritz masak: but getting STD's behavior with your approach wouldn't be too hard either 12:41
just '{{{' <?{ !$*IN_QUASI }> <EXPR> '}}}' 12:42
masak: speaking of which, are you sure that there need to be an <EXPR> in there, and not something more general like a <statementlist>? 12:43
12:44 cognominal_ left
masak <statementlist> sounds more apt. 12:48
I guess the latter allows `my $blabla = 5; $foo` and not just `$foo`.
then yes.
12:49 grondilu joined
grondilu is trying to compile rakudo-star on his new fedora system 12:50
12:51 MayDaniel left
moritz masak: exactly 12:52
12:58 drbean left
masak moritz: well, <?{ $*IN_QUASI }>, but yes ;) 12:58
13:00 drbean joined
moritz masak: erm, yes 13:01
dalek kudo/macros-d2: 8c2db63 | masak++ | src/ (2 files):
backpedal on throwing an exception

as moritz++ points out, this is closer to how STD does it.		 13:03
kudo/macros-d2: 163c626 | masak++ | src/Perl6/Grammar.pm:
<statementlist>, not <EXPR>

Much better. moritz++
   $ ./perl6 -e 'say "OH HAI"; my $a; {{{say "woot!"; $a}}}'
   OH HAI
   woot!
   $ ./perl6 -e 'say "OH HAI"; my $a; quasi { say "woo!"; {{{$a}}} }'
   OH HAI
moritz masak: fwiw I didn't mean to say throwing an exception is wrong, just pointing out that STD's behavior can easily be emulated too 13:04
grondilu failed to compile :( 13:05
moritz grondilu: -v
grondilu « invoke() not implemented in class '1257060743' » 13:06
moritz :( 13:07
grondilu pretty much the same as lately
and it's a brand new linux system
(Fedora 17)
moritz wonders if shipping parrot 4.5.0 would have helped
grondilu (on i686)
takadonet morning all 13:10
masak moritz: still, I prefer this behavior. 13:14
13:19 crab2313 left 13:28 uvtc joined
uvtc The README in the R* release says it's the 2012-06 version. 13:30
13:30 fgomez left
masak :( 13:31
uvtc Also it doesn't mention using "--gen-nqp" in the build instructions. Is that no longer required?
(In the "Configure.pl" step.)
GlitchMr github.com/rakudo/rakudo 13:32
$ perl Configure.pl --gen-parrot --gen-nqp
I still see it
uvtc Turn that frown upsidown, masak.
I mean in the R* readme.
jnthn It's not required, afaik.
uvtc jnthn: thanks. 13:33
13:33 birdwindupbird joined, sergot left
masak jnthn: about to create QAST::Unquote in nqp. `class QAST::Unquote is QAST::Node {}` in its own file in src/QAST -- anything else? 13:33
jnthn masak: Update Makefile.in? 13:34
masak good call. I will.
arnsholt jnthn: Looks like you unbrokeded it. Thanks! 13:35
jnthn arnsholt: \o/ 13:36
masak what's a BlockMemo? sounds like a three-dimensional corporate document to me. ;)
jnthn masak: Used by QRegex, but endangered. :) 13:37
masak ah.
jnthn masak: It's used for attaching extra stuff to a block.
masak I see.
jnthn masak: All its usages should be refactored to use serialization.
masak: But I didn't want to block toqast on that.
13:38 grondilu left
arnsholt jnthn++ # Unbrokededness 13:38
uvtc Attempted to the 2012.07 R*, but it fails for me. Here's some of the leavings in my terminal window: gist.github.com/3193446
This is on Xubuntu 12.04, using a Perl 5.16 that I built in my ~/opt. 13:39
Please let me know if that's enough of the terminal window output, or if more is needed. 13:40
13:40 tokuhiro_ joined
uvtc Or if I should file a bug. I emailed rakudobug with the last one, but it may have been a duplicate. 13:40
masak cognominal++ tweeted it, but it's worth posting here, for those who want to read it: blog.peterdonis.com/opinions/still-...rlude.html -- briefly, inventor of Unix pipes Doug McIlroy is asked to do literary criticism of a literate program Knuth wrote. interesting stuff. 13:45
critique*, I guess. 13:46
I believe I referred vaguely to this exchange a few weeks back.
"He has fashioned a sort of industrial-strength Fabergé egg—intricate, wonderfully worked, refined beyond all ordinary desires, a museum piece from the start." 13:47
13:49 SamuraiJack joined 13:53 leprevost joined
masak "I need to put that on a Post-it note as an example of how to explain a script. The best part? It would fit on a Post-it note." :) 14:00
yes, those two posts are worth reading. 14:02
uvtc :)
14:02 Celelibi left 14:04 Celelibi joined 14:13 birdwindupbird left
dalek p: 1f54496 | masak++ | / (2 files):
added QAST::Unquote
 14:16
masak ideally, I would write tests for the new node type, as well.
they would do something like "here's an AST with Unquote notes in it; here's an AST fragment. I apply the latter to the former, and get an AST with the fragment stitched in." 14:17
(i.e. test the operational aspect of QAST::Unquote and stitching.) 14:18
it's interesting, by the way, that "stitching in" actually happens twice during the macro call. once when the static quasi is incarnated into a run-time object (conceptually) on block entry. 14:19
once when the AST from the macro call is stitched into the mainline code.
14:19 cognominal joined
masak both of these cases are "closure-like", in the sense that the stitching-in inside the macro has things carrying the mainline's context, and vice versa. 14:20
it's really beautiful if you think about it. :)
masak .oO( hey, look. I'm philosophizing about macros again! ) 14:21
anyway, given what's in t/qast/qast.t, I'm fairly confident I'll be able to write such a test. 14:22
14:23 cognominal left
masak at least if the stitched-in bit doesn't do lexical lookups of free variables. 14:23
14:24 cognominal joined
timotimo is it possible to stitch in a function in the local scope of the macro into the ast that's returned? 14:25
i'm imagining something like macro yoink() { sub frob($a) { $a + 1 }; return quasi { say {{{&frob}}}(99) } }; yoink;
r: macro yoink() { sub frob($a) { $a + 1 }; return quasi { say {{{&frob}}}(99) } }; yoink;
p6eval rakudo b87659: OUTPUT«sub frob($a) { ... }␤» 14:26
timotimo that's ... interesting 14:27
moritz masak: the double-closure thingy is what had us so confused during our early discussions, no?
masak: we tried to reduce it to one closure-like step, and it never worked out 14:28
uvtc: that bug is already in RT
14:28 skids joined
uvtc moritz: thanks. 14:28
Anyone else having the same problem trying to build R*? 14:29
moritz: is it the same "gc-gone-wild" problem I was previously having? It looks slightly different, because this time it's "invoke() not implemented in class '6778597130223229316'" rather than "invoke() not implemented in class ' 14:31
'".
14:31 tokuhiro_ left
jnthn It'll be the same underlying issue. 14:32
uvtc Anything I can do about it on my system?
jnthn I haven't managed to track down what on earth is going on there yet :(
masak moritz: that might well be it. it's a bit tricky to analyze confusion, though ;)
jnthn uvtc: Adding -G to the Parrot flags on that invocation is known to avoid the issue. 14:33
uvtc: I think it's documented in the ticket
Or some ticket.
UncleFester6 rt.perl.org/rt3/Ticket/Display.html?id=113660 - towards the end 14:34
uvtc UncleFester6++ thanks. 14:35
Forgot to mention (in case it's important), I'm on 64-bit. 14:36
jnthn uvtc: How much RAM?
uvtc jnthn: that's a good question... 14:37
masak moritz: `macro foo($y) { my $x = 1; quasi { say $x; say {{{$y}}} } }; my $x = 2; foo $x; # 1\n2\n` -- this captures it, I guess. nothing new to either of us at this point.
moritz: though I'm guessing things can be done with it that neither of us thought about yet ;)
uvtc jnthn: 1927MiB 14:39
jnthn Thanks 14:40
uvtc Is 2 GB good for building the whole enchilada?
jnthn (It's enough, it's just that memory size determines how often GC runs, and it seems that soemthing is getting eaten too early.)
(Thus having more memory reduces the probability of hitting the bug.)
moritz uvtc: it's enough, yes 14:41
uvtc Hm. Possible debugging technique: open hardware, remove a simm/dimm, try again. :)
jnthn Oh, I don't think it's a memory/hardware issue. 14:42
It's very likely something wrong in NQP.
uvtc jnthn: No, I meant, if it might have something to do with how much memory is installed, removing some could help make the bug happen for you to make it easier to track down. 14:43
More likely to bump it's head.
jnthn ah, I see 14:45
UncleFester6 on linux/bash ulimit maybe
jnthn I may also try a VM.
moritz UncleFester6: I think parrot ignores ulimits :( 14:46
geekosaur VM is how I'd do it, fwiw
ummm
ulimits are enforced by the kernel
uvtc jnthn: You're going to test it on a Vending Machine? 14:47
;)
geekosaur but they don't work the way one might naively think, so you need to be careful using them for this
jnthn ;-)
arnsholt Huh. For some reason my patch doesn't work
tadzik oh oh, macros!
14:49 UncleFester6 left
masak tadzik: they're really not that complicated. 14:51
IF YOU'RE ALREADY INSANE MWHAHAHAHA
moritz geekosaur: it's not about enforcing
tadzik SORRY I CAN'T HEAR YOU OVER THE SOUND OF HOW INSANE I AM 14:52
moritz geekosaur: parrot uses the available memory for determining how big a pool can grow before running the GC
geekosaur: and parrot ignores ulimits for that purpose
geekosaur right. because ulimits are solely enforcement.
moritz so if you set a strict ulimit and try to reproduce the bug, it'll simply OOM 14:53
geekosaur again, "don't work the way one might naively expect"
(we're saying the same thing differently)
15:00 fgomez joined 15:09 fgomez left 15:16 mucker joined 15:27 fgomez joined 15:28 fgomez left
arnsholt jnthn: Could you have a look at gist.github.com/3193047 again? 15:30
It runs fine now, but for some reason &func.signature.returns is still Mu for sub func(Int --> Int) { ... }
15:31 adu joined
jnthn + if %signature_info<returns> { 15:32
A type object is always false.
I guess you wanted if nqp::existskey(%signature_info, 'returns') { ... } 15:33
arnsholt Durrr. Of course it is =)
dalek kudo/macros-d2: a908aa4 | masak++ | tools/build/NQP_REVISION:
[tools/build/NQP_REVISION] bumped

For QAST::Unquote.
15:35 fgomez joined
arnsholt jnthn++ # $ ./perl6 -e 'sub test(Int --> Int) { say "hi" }; say &test.signature.returns'\nInt() 15:38
masak \o/ 15:39
jnthn \o/
arnsholt++
15:39 simcop2387 left 15:40 simcop2387 joined 15:41 SamuraiJack left, SamuraiJack_ joined
arnsholt And even $ ./perl6 -e 'sub test(&cb(Int --> Int)) { say "hi" }; say &test.signature.params[0].sub_signature.returns'\nInt() 15:42
Time for a full spectest
flussence has that always been broken? I remember nativecall flipping out when I used the "-->" syntax instead of "returns". 15:44
15:45 cognominal left
arnsholt Yeah. Never been implemented 'till now, apparently =) 15:46
Or rather, implemented in the grammar but not passed on to the AST 15:48
15:48 adu left 15:51 zhutingting joined 15:52 Guest74836 left
masak \o/ 15:53
zhutingting hi~ I installed rakudo-star-2012.07.msi~ An error here: PARROT VM: Could not load bytecode 15:55
Could not load oplib `nqp_ops'
16:00 kaxing left, kaxing joined
masak hm, that error message is very familiar. 16:00
could it have something to do with previously installed versions of Parrot/
?
zhutingting I installed rakudo2011.4 parrot 3.3 before 16:03
jnthn Hmm 16:04
zhutingting uninstall the old version first?
jnthn zhutingting: Where did it install it to, ooc?
zhutingting rakudo2011.4 install in c:\ 16:05
16:05 Guest74836 joined
zhutingting 2012.7.msi auto install in e:\ 16:05
jnthn hmmm. 16:07
And on mine it install it in d:\
I'm almsot certain it's *meant* to install it on c:\ 16:08
[Coke] (parrot not relocatable) is there at least a ticket about parrot not being relocatable?
moritz I think so 16:09
github.com/parrot/parrot/issues/se...elocatable seems like "no" 16:10
16:22 MayDaniel joined 16:25 birdwindupbird joined 16:32 birdwindupbird left 16:44 johnshoe joined 16:45 johnshoe left 16:52 uvtc left
pmichaud good morning, #perl6 16:53
16:53 zhutingting left
pmichaud problems with the .msi? 16:53
16:56 seldon joined
jnthn pmichaud: My best guess is that it isn't forcing the installation to happen on drive C 16:57
pmichaud I'm not sure how to do that.
jnthn pmichaud: When I tested it on my laptop yesterday, there sin't anyway option. So it seemed fine.
dalek p/pirt: c6b0eee | jnthn++ | src/QAST/ (2 files):
Consistently use :name for labels, not a mix of :result and :name.
 17:00
p/pirt: 544a992 | jnthn++ | / (3 files):
Add initial sketch of PIRT, which is to be a hopefully more memory efficient replacement for POST.
p/pirt: e74e38c | jnthn++ | / (2 files):
Give PIRT::Sub support for loadlibs, name, subid and pirflags; also emit calls.
p/pirt: 2534c85 | jnthn++ | t/qast/qast.t:
Get QAST tests to expect QAST::Compiler to produce PIRT.
p/pirt: 6bc5c6c | jnthn++ | src/QAST/Compiler.nqp:
Start updating QAST::Compiler for PIRT. Gets some QAST tests passing again.
p/pirt: d4ed168 | jnthn++ | src/QAST/Operations.nqp:
Update QAST::Operations for using PIRT, which gets many more of the QAST tests passing again.
17:00 unixms joined
pmichaud aha. apprently wix/msi installs to the drive with the most free space by default. 17:00
masak huh. 17:01
jnthn pmichaud: Hm. A user-friendly default apart from when it's not :)
pmichaud so far googling the issue gives me a lot of messages from wix folks saying that we really don't want to force installation to c:\ :-( 17:02
jnthn pmichaud: In the long run, no. In the shorter term, it's better than giving people an installation that looks like it's gonna work, then doesn't. 17:03
17:03 unixms left 17:04 UncleFester6 joined
UncleFester6 pmichaud: .msi wfm 17:05
pmichaud UncleFester6: excellent
16:08 <[Coke]> (parrot not relocatable) is there at least a ticket about parrot not being relocatable? 17:12
I know that parrot isn't very relocatable on unix systems... but do we really know that it's not relocatable on windows? 17:13
maybe I should test that theory first
jnthn Think it isn't, but I certainly haven't tried that in a long while. 17:14
moritz pmichaud: jnthn tried for the first star release
jnthn Oh, actually...
moritz is currently writing the ticket
jnthn A while back I spent some time doing a disgusting attempt at binary patching the executable..
So yeah, I guess it doesn't work.
(I didn't manage to binary patch the executable either :P)
pmichaud jnthn: do you remember if the binary actually had C:\ (or equivalent) in it, or did it just depend on a specific directory? 17:15
jnthn pmichaud: I *think* it had the drive letter in. 17:17
pmichaud: Certainly all the paths I get from parrot_config --dump have the C:\ in
pmichaud: Maybe if you configure with a --prefix=\rakudo though... :) 17:18
moritz github.com/parrot/parrot/issues/800 17:20
pmichaud jnthn: good idea -- I'll try that. 17:21
TimToady pmichaud: is irclog.perlgeek.de/perl6/2012-07-28#i_5853362 something new or the same old? 17:25
pmichaud TimToady: same old; you're correct that it's one of those things I'm working on. 17:26
TimToady good, just didn't want it to fall through the cracks if it was different
[Coke] +# 07/28/2012 - rakudo++ (22930); niecza (89.91%); pugs (40.82%) 17:27
pmichaud if you have any thoughts on github.com/perl6/specs/issues/18 it might help a bit :-)
(that's not directly related to the X/Z issues... but indirectly related things often have a habit of affecting other parts of list handling :) 17:28
[Coke] moritz: I have a ticket open to tell parrot "this is your max memory, go no further" (just like java does) and it has never really gotten traction. perhaps upvoting it will help.
TimToady maybe push on a list doesn't flatten, unlike push on an array
er .plan
pmichaud I could live with that; would non-flattening be true also for push/unshift, then? 17:29
(on List vs Array)
TimToady I still think there are two sets of shifters, depending on whether you're binding slurpy vs non-slurpy 17:30
give me the next positional arg vs give me the next slurpy arg 17:31
17:31 jaldhar left
pmichaud I follow what you mean by the different shifters; I don't see how they quite apply in this case. 17:31
right now slurpiness is a feature of the parameter list 17:32
17:32 leprevost left
pmichaud (and thus the method signature for plan/push/unshift) 17:33
TimToady non-slurpiness can either be positional (single "arg") or lol (list of "arg") where arg is a subparcel 17:34
pmichaud okay, I'll start with a more basic question: what's the signature for List.plan ? 17:36
TimToady but each arg can be its own lazy parcel, so that X can somehow have two sublists that know not to look inside [] or $
I'd suspect it's **@parcels 17:37
pmichaud oh, was your "two sets of shifters" comment aimed at X/Z, and not .plan/.push ?
17:38 fgomez left
TimToady I'm just thinking of the "shifters" as the primitive difference between the two binding styles that do or do not unpack the next parcel in the arglist 17:39
and it seems that such a difference should be accessible at a low level apart from only binding
pmichaud okay. I don't see that as a big issue (yet); to me, it's essentially the difference between **@slurpy and *@slurpy
TimToady don't care whether they're actually called "shift" 17:40
17:40 SamuraiJack__ joined 17:41 SamuraiJack_ left
TimToady but it's still basically the .getarg vs .getitem from the old S07 I'm thinking of 17:41
.getitem being the slurpy one
pmichaud s/slurpy/flattening ?
TimToady yes
pmichaud let's suppose I have my $l = ( (1,2,3), (4,5,6) ).list; 17:43
and I do my $x = $l.getitem
what's left in $l ? 17:44
TimToady (2,3),(4,5,6)
pmichaud okay, so it "flattens" the first element but only long enough to get an element
i.e., it's not 2,3,(4,5,6) 17:45
TimToady that seems ickier for some reason
though for sure people don't usually switch from slurpy back to item
pmichaud and if I have $l = ( (3,), (4,5,6) ).list
TimToady er, arg
pmichaud and then do $x = $l.getitem
$l ends up with... ? 17:46
TimToady (4,5,6)
pmichaud okay. I haven't come up with a case where a list needs to be able to support both item-based and arg-based shifting 17:47
(at the same time)
and I suspect not having that case is where the item-based and arg-based shifting distinction escapes me. 17:48
17:48 unixms joined
TimToady I could imagine a signature that says (*$item, @arg, *@items), where $item would get 1, @arg would get (2,3), and @items would get (4,5,6) 17:49
currently we'd generally do that with a subsig 17:50
pmichaud also, the spec currently says that slurpy parameters follow any required/optional parameters 17:51
so "*$item, @arg" seems illegal.
(if you're speculating that the spec might change to enable something like "*$item, @arg", I'm okay with that :) 17:52
TimToady well, I guess I don't care about that so much as understand how "list X list" loses its don't look inside [] policy 17:53
17:53 SamuraiJack__ left, SamuraiJack joined
pmichaud so, just to confirm, you're expecting the correct answer of ['a', 'b'] X~ 1 .. 3 to be ... ? 17:54
TimToady what niecza produces "a b1 a b2 a b3"
pmichaud 'a b1', 'a b2', 'a b3' is my expectation
TimToady X basically takes two parcels that need to each be treated as a slurpy 17:55
pmichaud as a flattening slurpy?
17:55 unixms left
TimToady yes, list infixes flatten 17:56
unlike hypers
pmichaud "as a slurpy" is ambiguous to me; both **@lol and *slurpy are "slurpies"
*@slurpy
"slurpy" to me simply means variadic in nature (perhaps I need an adjustment) 17:57
TimToady sure, in my mind those are slurp args vs slurp items, but when I say "slurpy" in isolation it's usually to distinguish from lol
so maybe we need a term for *
pmichaud slurp items vs slurp flat is what I tend to think of
TimToady well, it's the flattening slurpy 17:58
pmichaud right
I should say slurp lists vs. slurp flat
17:58 daftdude joined
pmichaud **@lol is the list slurpy, while *@slurpy is the item/flattening slurpy 17:58
TimToady I think of ** as arg slurp because it's the same as binding to a positional 17:59
...to multiple positionals
pmichaud I'm fine if we use "slurpy" to mean the "list slurpy" case.... although almost everything that has been written about p6 until now tends to use "slurpy" in conjunction with the flattening / *@slurpy case.
17:59 daftdude left 18:00 snearch joined
pmichaud I can live with "arg slurp" if that works better for you. 18:00
lol slurp might make more sense
TimToady from the S09 viewpoint it's more like a dimensional slurp: "Don't lose dimensional info" 18:01
18:01 snearch left
pmichaud anyway, the problem with Rakudo's X has to do with its mishandling of **@lol arguments. 18:02
I basically know what the problem is there, and how to fix it.
TimToady well, I guess X is a vararg anyway, so ** kinda makes sense
pmichaud correct
the issue with Rakudo's X is that **@lol itemizes everything in it 18:03
so Rakudo's X defaults to processing everything in the item, which means it looks inside of [ ]
dalek p/pirt: 6ff18e2 | jnthn++ | src/QAST/PIRT.nqp:
Implement method calling. Down to six tests for t/qast/qast.t that fail.
p/pirt: 01a918f | jnthn++ | src/QAST/Operations.nqp:
Be consistent in how we call infer_type. Leaves just two tests in qast.t failing.
p/pirt: 6568e10 | jnthn++ | src/QAST/ (2 files):
Fix inline, which tests the remaining qast.t tests passing when using PIRT instead of POST.
TimToady I guess it's confusing because the lol operator *does* itemize everything, but ** doesn't
18:03 mucker left
[Coke] pirt? 18:03
pmichaud correct, that's why Rakudo's **@lol itemizes stuff. 18:04
jnthn [Coke]: Replacing POST.
pmichaud I was folloiwng the .lol pattern.
TimToady so we should definitely not call ** a "lol"
jnthn [Coke]: A much, much smaller job than replacing PAST. :)
[Coke]: With hopeful memory benefits.
pmichaud I'm fine with that. how about "slurpy args", then?
and we write **@args
TimToady seems an improvement 18:05
[Coke] no relation to PIR?
TimToady does this address the other non-lazy-parcel problem we were talking about earlier?
don't remember the example offhand...
pmichaud not directly, unless you feel that .plan is indeed the way to concatenate lists
TimToady well, add args to a list 18:06
pmichaud or are you referring to some other problem we were talking about before today?
TimToady something triggered this list rethink a couple three weeks ago
pmichaud oh, sure
it was the issue with how gather/take works with parcels
TimToady my brane has been topsy turvy since then
pmichaud yes, this starts to address that as well
TimToady right, that were it 18:07
pmichaud fixing gather/take is what led me to #18
but the gather/take problem isn't so much an issue of how we deconstruct args, but more a problem of how we construct return results
TimToady it seems natural to me that list treat .plan as args rather than items
jnthn [Coke]: POST is the thing that gets turned into PIR
[Coke]: PIRT (PIR Tree) fills the same role.
pmichaud but .push would still be items?
18:08 fgomez joined
pmichaud (having .plan == args and .push == items works for me, fwiw) 18:08
TimToady I don't see that; parcels still have to live through the reification
flussence
.oO( coming soon: nqp/partridge )
TimToady but maybe that's orthogonal to .plan vs .push 18:09
pmichaud none of these really affect Array; Array forces flattening anyway.
TimToady right; just seems odd to me to supply a flattening method to pushing list when lists are supposed to be agnostic to flattening 18:10
can always .plan: flat @stuff
pmichaud I'm fine if .push does args instead of items; I just want to know where the parallels are(n't)
18:10 spider-mario left
TimToady I think I'd be more comfy with list.push pushing args 18:10
pmichaud and .unshift also? 18:11
TimToady presumably
pmichaud wfm
so, the major difference between .push and .plan is the eagerness involved
(as it's been up to now, iiuc) 18:12
TimToady well, eagerness at the arg level, not at the stuff inside the arg
pmichaud correct.
(and agreed)
TimToady it seems to be fitting together better now 18:13
pmichaud anyway, I still need a week or so for resolving X/Z stuff
there's still some fragility in the list implementation where small changes result in surprising breakages. But I think that can be all resolved this go-round
TimToady sure, just trying to see the big picture again after involunary brane shrinkage 18:14
*tary
pmichaud switching **@lol to **@args interpretation eliminates one of the problematic constraints
at least, I think that's what it does :)
I'll update #18 with our discussions here. 18:15
TimToady on lazy warnings, I'm still of two minds, after several weeks of background think 18:17
moritz jnthn: can I remove the QAST milestone from ROADMAP? or do the remainng nqp migration and inliner work make it worth keeping it?
jnthn moritz: I was thinking we'd remove it after migrating NQP to QAST. 18:18
TimToady the unification with failure seems like a good thing in the case of +«@foo and such
18:18 jaldhar joined
TimToady but I wonder if the laziness will eventually bite us; I suppose I should worry more about lazy fatals biting us than lazy warnings, and I don't worry about that (much) :) 18:19
also, it's a bit odd to have an undefined failure that yet has a Cool value, but I suppose that could be worked out somehow 18:20
probably means we would need a Failure analog that is spelled differently to easily distinguish the has-a-value-anyway behavior 18:21
there's also the fact that warnings are currently considered control exceptions, iirc 18:22
pmichaud yes, warn is listed as a control exception 18:23
TimToady call 'em Bogons or some such :)
and instead of fail() we have bogus(), except that's not a verb 18:24
bogey() :)
18:24 snearch joined
TimToady which is both noun and verb 18:24
pmichaud well, I wasn't hoping for a new Failure class. :-/
(I wasn't hoping for anything specific... but a new class wasn't it :)
TimToady maybe it's a Failure that contains a Warning
instead of an Exceptoin 18:25
moritz maybe Warning isa Exception
or manybe Exception has a fatality attribute
TimToady it's just resumable (and control)
but something would need to introspect that to return 0 or '' or False
unless we do something mixiny 18:26
or maybe it's just fail($mywarning, :return(Nil)) 18:27
or maybe it's just fail($mywarning, :value(Nil))
hmm, except that would produce another warning when Nil is evaluated... 18:28
:value('') maybe
pmichaud r: say +Nil
p6eval rakudo b87659: OUTPUT«use of uninitialized value of type Nil in numeric context in block at /tmp/Zb4T2bNjHV:1␤␤0␤»
pmichaud yup.
TimToady r: say +''
p6eval rakudo b87659: OUTPUT«0␤»
TimToady n: say +'' 18:29
p6eval niecza v19-15-g051783d: OUTPUT«0␤»
TimToady looks like '' works
seldon n: say +'abc' 18:31
p6eval niecza v19-15-g051783d: OUTPUT«Unhandled exception: Cannot parse number: abc␤ at /home/p6eval/niecza/lib/CORE.setting line 1414 (die @ 5) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3492 (ANON @ 11) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3494 (NumSyntax.str2num @ 5) ␤ at …
seldon Okay, so it's not just fed to atoi.
18:31 Guest74836 left
moritz I'm currently brainstorming which spec changes rakudo still needs to chase 18:31
currently I've come up with: sigilless variables and parameters
18:31 GlitchMr left
moritz the changes that ? makes to match objects 18:31
TimToady oh, you mean x? -> x or Nil 18:32
moritz right
pmichaud the List stuff I'm working on (although there Rakudo is leading the spec change as much as chasing it :)
moritz any other major-ish things that need updating?
pmichaud leading whitespace under :sigspace
skids Type1(Type2) in parameter lists?
moritz skids++
TimToady there's irclog.perlgeek.de/perl6/2012-07-27#i_5850929 18:33
pmichaud aye.
so, use of $_ in a block prevents it from hashing
TimToady yes, you have to use a temp var to do that
moritz last we discussed this, there was a spec issue with that 18:34
but I can't remember what exactly it was
skids moritz: handles trait?
moritz skids: that's implemented 18:35
(at least nearlly all of it)
pmichaud if we remember the spec issue, then file a github ticket for it :)
TimToady I think getting map to DWIM on { .key => transform .value } is pretty important
pmichaud TimToady: +1
it shouldn't be that hard; we just flag a block as using $_ when we encounter it.
TimToady just have to notice that .foo uses $_
pmichaud we do that already; it desugars to a lookup of $_ in the ast, iirc 18:36
moritz that looks like something I might be able to do
pmichaud then if a block is using $_, we don't let it de-hashify.
alternatively, we may want to see if the block is treating $_ as a parameter and key on that.
skids {}.assuming() ? 18:37
moritz r: say { $^a => $^b}.WHAT
p6eval rakudo b87659: OUTPUT«Block()␤»
TimToady r: say { $_ => $_ }.WHAT 18:38
p6eval rakudo b87659: OUTPUT«use of uninitialized variable $!key of type Any in string context in block at /tmp/u_Cst6Q1fC:1␤␤Hash()␤»
TimToady it's just not paying attention to $_, seems
dalek kudo/nom: bfb50b2 | moritz++ | docs/ROADMAP:
[ROADMAP] add missing features and spec chases
pmichaud well, that's partially it.
in the case of { $^a => $^b }, the block has non-zero arity, so we know it's not a hash constructor
moritz pmichaud, jnthn: feel free up correct my commit if you don't share my point of view about priorities and effort 18:39
pmichaud in the case of { $_ => $_ }, the arity is zero
TimToady whereas $_ blocks have <-> $_ = CALLER::<$_> or some such
pmichaud (and I don't know if it's been decided at that point as to whether $_ is a parameter to the block or not, in the ast)
dalek kudo/nom: 8435346 | moritz++ | src/core/ (2 files):
implement Str.codes and Cool.codes

currently an alias for .chars
pmichaud moritz++ # .codes 18:40
TimToady well, if it's decided by defaulting at call time, then you just have to guess at compile time, and in this case it should guess "parameter", I think
masak moritz: you mean '.chrs' ?
pmichaud masak: no, .chars 18:41
masak: .codes returns the number of codepoints in a string
masak oh!
right.
moritz pmichaud: my current plan for detecting $_ usage is to have a dynamic variable $*USES_DOLLAR_UNDERSCORE or so per block, and set it to 1 in blocks that use $_
TimToady darn overloaded plurals in English...
masak we have both .chrs and .chars. yet another reason to scrap .chrs
moritz pmichaud: and if that's the case, we don't execute the block to produce a hash 18:42
TimToady and then you get the opposite failure mode
moritz pmichaud: does that sound sane?
TimToady "why didn't .chars return the chars?"
pmichaud moritz: I'm not sure a dynamic works there. At the point where the hash constructor decision has to be made, you're already out of the block parse (I think)
masak nr: say <79 72 32 72 65 73>>>.chr.join
p6eval rakudo b87659, niecza v19-15-g051783d: OUTPUT«OH HAI␤»
pmichaud we decide whether to change a block to a hash constructor in circumfix:sym<{ }> 18:43
TimToady unless we make .chars lazy enough to optimize +.chars on the fly
jnthn pmichaud, moritz: We may want to start recording variable accesses that a block makes in .symbol somehow.
skids r: class C { method a handles Int { 3 } }; my C $c; # This form of the handles trait
p6eval rakudo b87659: OUTPUT«===SORRY!===␤Cannot call 'trait_mod:<handles>'; none of these signatures match:␤:(Attribute:D $target, $thunk)␤␤»
masak TimToady: one can't map words optimally to their effect. at some point people will have to learn what the methods do.
pmichaud jnthn: yeah, that was more of what I was thinking.
masak TimToady: but having .chrs and .chars isn't helping matters.
jnthn That would let us implement this but also:
18:44 quietfanatic joined
jnthn std: my $a; { say $a; my $a } 18:44
p6eval std e52e3ca: OUTPUT«===SORRY!===␤Lexical symbol '$a' is already bound to an outer symbol (see line 1);␤ the implicit outer binding at line 1 must be rewritten as OUTER::<$a>␤ before you can unambiguously declare a new '$a' in this scope at /tmp/j1B9tTyUEs line 1:␤---…
jnthn That one.
TimToady .unords :)
masak TimToady: also -- see p6eval above -- .chrs is redundant in a way which "it's diagonal" doesn't excuse, IMO.
pmichaud jnthn: I'm thinking it'd be fine for .symbol to record "seen but not declared in this blook"
PAST left a lot of the details of .symbol up to the user, so I'm hoping QAST does the same :)
jnthn pmichaud: It's only used as a fallback for finding :scope 18:45
pmichaud: Otherwise it's just a hash of hash.
pmichaud actually, just "seen in this block" would be worthwhile.
quietfanatic .chrs currently returns a list of characters, right?
jnthn pmichaud: Anyway, I'm +1 to this. 18:46
TimToady well, yes, and .ords is just .comb».ord, so if we remove one, we should remove the other
moritz quietfanatic: no, chrs applies chr() to a list of integers, and joins them and returns a string
quietfanatic ah
so, .comb gives you a list of characters
moritz correct
quietfanatic but what if I want a list of, say, codepoints? 18:47
TimToady but it seems like that the optimizer will end up optimizing both of those to some internal .ords/.chrs method, so why not expose it
quietfanatic because that takes up one more name.
I mean two morw names
moritz quietfanatic: in theory, .comb/rx:codes/./)
pmichaud expose can be good; it's the names that are confusing, I think
moritz quietfanatic: but I don't really believe that it's a good idea to expose strings at different Unicode levels through a single type 18:48
that's old p5 think, where binary strings and text strings are also intermingled
TimToady well, we may well end up with CodeStr or some such as an intermediate between Buf and Str 18:49
pmichaud jnthn: changing --prefix to '\rakudo' seems to have eliminated a lot of the C:'s from the config hash. Now I'll see if it works at all when on a drive other than c: :-)
TimToady or maybe they're called NFC, NFD, etc to keep those ideas straight too.
quietfanatic How about just a list of X where X is whatever? 18:50
pmichaud although all that does for us is mean that the install still works when it ends up on drive 'd' or whatever. People will then ask "why did the installer put it on 'd'? I want it on 'c'!"
quietfanatic If I want to think of it as a continuous string, I'll use a Str. If I want to think of it as a list of X, why not a real list of X?
pmichaud ....on the plus side, if it works on any drive, that would mean it could work on a flash drive :-) testing that now. 18:51
moritz quietfanatic: and what is X? words?
TimToady well, Buf is already positional
quietfanatic characters, codepoints, bytes, whatever strings could be composed of
TimToady but compact, whereas "real list" means high overhead to me
moritz quietfanatic: that's how I think of text most of the time (list of words and punctuation chars)
masak TimToady: I'd much rather see .comb».ord in my code than .ords, simply because I can mix and match .comb, », .ord and many other constructs to create a nice flexible result. .ords is a one-trick pony, like PHP's array_slice.
TimToady so Buf[Codes] is already more or less specced to be listy 18:52
quietfanatic Ah, I guess I'm still used to langugaes where lists are not very high-overhead.
pmichaud idiomatically, I've been wondering lately if we should encourage a lot more use of », if only because it implies parallelization/optimization potential.
quietfanatic like C, where most are arrays or intrusive linked lists
TimToady masak: but we huffmanize for common one-trick ponies all the time 18:53
see "say"
moritz pmichaud: if we want that, we shoudl first benchmark if » is at least somewhat fast right now
quietfanatic say now does something different though :)
masak this is not an important huffminaztion. it's the relative lack of merit which bothers me.
we should focus on overpopulating CORE with useful stuff :) 18:54
quietfanatic I don't think .ords is very common, personally
masak right.
pmichaud moritz: I suspect it's about as fast as .map; if not, it can certainly be made so. In fact, I bet it can be made a lot faster than .map because we don't have to worry about exceptions or laziness.
masak I could argue for .abss as well, because sometimes I just want to do .abs on all those list elements.
moritz pmichaud: iirc » uses .pick for non-determinstic order 18:55
pmichaud moritz: we've been doing that simply to prevent people from misapplying » as map where order matters.
I can still come up with a faster » that does things in a non-deterministic order, I think. 18:56
TimToady well, I guess I might argue that you really want .encode(NFG) instead of .ords, assuming a Buf[NFGChar] can listify appropriately
moritz pmichaud: that's fine, but I just suspect it won't be fast right now
pmichaud moritz: anyway, I was thinking primarily idiomatically. We should probably first agree that encouraging » use is a good thing; and if so I'll happily update the hyper code in rakudo to be a lot faster than it is :) 18:57
quietfanatic Bufs can act like lists, can't they? They do Positional.
pmichaud yeah, I think that .ords is just some form of Buf :-)
moritz r: say 'OH HAI'.encode.list 18:58
p6eval rakudo b87659: OUTPUT«79 72 32 72 65 73␤»
TimToady nr: my $x = Buf.new("foobar"); .say for @$x;
p6eval niecza v19-15-g051783d: OUTPUT«===SORRY!===␤␤Undeclared name:␤ 'Buf' used at line 1␤␤Unhandled exception: Check failed␤␤ at /home/p6eval/niecza/boot/lib/CORE.setting line 1402 (die @ 5) ␤ at /home/p6eval/niecza/src/STD.pm6 line 1147 (P6.comp_unit @ 37) ␤ at /home/p6…
..rakudo b87659: OUTPUT«This type cannot unbox to a native integer␤ in method set_codes at src/gen/CORE.setting:6758␤ in method new at src/gen/CORE.setting:6750␤ in block at /tmp/QN9yo49rod:1␤␤»
pmichaud nr: my $x = "foobar".decode; .say for @($x);
p6eval niecza v19-15-g051783d: OUTPUT«Unhandled exception: Unable to resolve method decode in type Str␤ at /tmp/PwbqtDhOn5 line 1 (mainline @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3929 (ANON @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3930 (module-CORE @ 564) ␤ at /home…
..rakudo b87659: OUTPUT«No such method 'decode' for invocant of type 'Str'␤ in block at /tmp/wxi0PLd_SK:1␤␤»
TimToady nr: my $x = Buf.new(65..70); .say for @$x;
p6eval rakudo b87659: OUTPUT«65␤66␤67␤68␤69␤70␤»
..niecza v19-15-g051783d: OUTPUT«===SORRY!===␤␤Undeclared name:␤ 'Buf' used at line 1␤␤Unhandled exception: Check failed␤␤ at /home/p6eval/niecza/boot/lib/CORE.setting line 1402 (die @ 5) ␤ at /home/p6eval/niecza/src/STD.pm6 line 1147 (P6.comp_unit @ 37) ␤ at /home/p6…
moritz TimToady: there's no Buf.new(Str) constructor
pmichaud: it's Buf.encode :-) 18:59
pmichaud moritz: what's the method for going the other way?
TimToady there should be one that handles ASCII, I suspect
moritz pmichaud: erm sorry, Str.encode, Buf.decode
TimToady and "foo" as a literal should allomorphically be both Buf and Str somehow, if the world were perfect
pmichaud r: .say for "foobar".encode 19:00
p6eval rakudo b87659: OUTPUT«Buf:0x<66 6f 6f 62 61 72>␤»
pmichaud r: .say for "foobar".encode.list
p6eval rakudo b87659: OUTPUT«102␤111␤111␤98␤97␤114␤»
pmichaud r: .say for "foobar".encode[]
p6eval rakudo b87659: OUTPUT«102␤111␤111␤98␤97␤114␤»
quietfanatic sounds like you want Haskell-style overloaded literals :)
TimToady I'm sure I wanted them from before Haskell was on my radar
TimToady and allomorphism go way back 19:01
diakopter i'd love for the regex engine to be able to operate on buf8 in utf8 mode
while Im dreaming 19:02
quietfanatic I'd personally prefer an object that coerces two ways than a type-classed polymorphic thing that is two objects simultaneously.
s/than/over
pmichaud jnthn: perhaps I should have the MSI always compile/install to "Program Files\Rakudo\..." ? 19:04
(and yes, it appears that I have Rakudo running from E:\rakudo just fine :) 19:05
diakopter make shre to install to the 64-bit one as appropiate
moritz pmichaud: and if you put it in %PATH%, can you call it from a different drive?
pmichaud moritz: checking
moritz: ah, sadly no. 19:06
so, it's not really relocatable to a different drive. 19:07
jnthn pmichaud: iirc, we have problems with paths with spaces in somewhere.
Sure there's a ticket on that
pmichaud jnthn: okay. 19:08
moritz iirc that's our fault, parrot can handle it
pmichaud So, I need to figure out how to get the installer to always go to C:\rakudo, no matter what.
I'll work on that.
19:09 spider-mario joined
jnthn moritz: I'm fairly sure t's an RT ticket I was thinking of. :) 19:09
*it's
dalek p/pirt: 16694d7 | jnthn++ | src/QAST/ (2 files):
Get hll and namespace bits added.
p/pirt: de2b269 | jnthn++ | src/QAST/PIRT.nqp:
Oops, forgot this important memory-freeing line!
p/pirt: dd40b44 | jnthn++ | src/QAST/PIRT.nqp:
Stub in node support for annotations, though not emitting them yet.
p/pirt: 7fcb20b | jnthn++ | src/HLL/Compiler.pm:
Update HLL::Compiler.pir so it knows about PIRT as well as POST.
TimToady quietfanatic: sure, that's the direction we went overall with P6's Cool, rather than the P5 approach, but the problem is how you cache to avoid reconversion of heavy things like strings
mhasch re allomorphism: for literals, it is not unusual for a language to make them compatible to different types at once. We did that in the 80's with Wirth languages all the time. 19:10
TimToady also, Bufs, unlike Strs, are mutable
moritz don't we have immutable bufs too?
we could use those instead of Buf
quietfanatic mutability vs. immutability is something I'd like to see abstracted out 19:12
rather than having two types for everything, one mutable and one immutable
with names that don't convey that difference.
moritz that would be nice, yes 19:13
pmichaud I don't know that we have many of the "two types for everything" situation now
(so I agree we shouldn't introduce/extend it)
quietfanatic yeah, it's not actually that bad 19:14
but do we still have Seq, for instance? 19:15
pmichaud not currently
moritz it seems Blob is the immutable version of Buf
pmichaud and I've been hoping to not reintroduce Seq
(and thus far, quite successful)
moritz ok, we have Buf/Blob, Pair/Enum, Hash/EnumMap
pmichaud I still don't think that Enum should be immutable; it's just a binding instead of a container relationship, like List/Array 19:16
same for Hash/EnumMap. And we already remarked a couple of days ago that EnumMap is the wrong basis for Hash.
masak aye.
quietfanatic Is Enum just a single pair-like thing? 19:18
moritz yes
quietfanatic I'd have thought of an Enum as the whole listing
moritz an immutable pair
and it's a terrible name
quietfanatic An enum should be the whole thing, not a piece
pmichaud I could be wrong, but I think Enum was trying to introduce a parallelism to List/Seq (that we've since eliminated) 19:19
quietfanatic enumeration
masak quietfanatic: you'll have to take that up with S12.
quietfanatic: it's quite clear on the enum/enumeration distinction, to the point where it actually defines it.
quietfanatic well 19:20
it's all names anyway
but it seems to me that singular of enumeration is like
number
:) 19:21
mhasch I agree with moritz that the name "enum" is not very intuitive.
masak quietfanatic: sometimes it's a number.
quietfanatic numeral
masak quietfanatic: of course, it's a name, too.
masak gets out of the bikeshed before things get "interesting"
quietfanatic right, I think 'enum' has too far evolved from its etymological origins to be compared to it. 19:22
timotimo how about "symbol" instead? :|
moritz timotimo: too generic 19:23
timotimo true. and "icon" makes one think about graphical things too much
quietfanatic well, it's a mapping of one symbol to a thing
pmichaud catalog. dictionary. mapping.
index.
timotimo in't it the point that it isn't a catalog? just one thing inside a catalog?
quietfanatic how about just a constant Pair, hm?
moritz lets rename EnuMap to Mapping, and Enum to Mappy
because it's a mini Map
masak :) 19:24
+1
quietfanatic Maplet?
Mappy sounds like a role
masak or Maplet.
moritz +1 to Maplet
pmichaud careful with "Map" named things; they pun with ".map"
jnthn Who has ever used the fact that you can assign to a Pair's value, ooc?
masak Mapka :)
jnthn Mapka! \o/
masak \o/
pmichaud jnthn: TimToady was remarking that he wanted => to produce mutable things.
(at yapc::na)
quietfanatic like a cons pair, yes 19:25
pmichaud and certainly when iterating over a hash's pairs, you expect to be able to assign to them.
jnthn r: my $a = foo => 42; say $a.value; $a.value = 100; say $a
p6eval rakudo 843534: OUTPUT«42␤"foo" => 100␤»
jnthn Hm, seems we already implement it
pmichaud correct.
jnthn Oh...the iteration thing is a decent use case I guess.
pmichaud r: my %h = a=>1, b=>2, c=>3; for %h { .value *= 2; }; say %h.perl 19:26
p6eval rakudo 843534: OUTPUT«("a" => 1, "b" => 2, "c" => 3).hash␤»
pmichaud hmmm
r: my %h = a=>1, b=>2, c=>3; for %h.pairs { .value *= 2; }; say %h.perl
p6eval rakudo 843534: OUTPUT«("a" => 1, "b" => 2, "c" => 3).hash␤»
pmichaud hmmmm 19:27
that looks like a bug to me.
moritz it is
quietfanatic r: my %h = a=>1, b=>2, c=>3; for %h.pairs <-> $p { $p.value *= 2; }; say %h.perl
p6eval rakudo 843534: OUTPUT«===SORRY!===␤Missing block␤at /tmp/ZH3gTo9iFY:1␤»
moritz it seems we recontainerize at some point
pmichaud I think that .pairs must not be returning the correct binding.
moritz might be the .pairs
TimToady n: my %h = a=>1, b=>2, c=>3; for %h.pairs { .value *= 2; }; say %h.perl 19:28
p6eval niecza v19-15-g051783d: OUTPUT«{"a" => 2, "b" => 4, "c" => 6}.hash␤»
moritz niecza++
pmichaud i.e., .pairs is constructing a new Pair instead of creating a Pair that is properly bound.
TimToady p: my %h = a=>1, b=>2, c=>3; for %h.pairs { .value *= 2; }; say %h.perl
p6eval pugs: OUTPUT«\(\("a", 2), \("b", 4), \("c", 6))␤»
moritz r: my %h = a => 1; %h.pairs[0].value = 5; say %h.perl
p6eval rakudo 843534: OUTPUT«("a" => 1).hash␤»
pmichaud ...which is why I continue to say that the Enum/Pair relationship is wrong.
(in Rakudo, at least.)
pmichaud reads S12 to find out why Enum is "Enum" 19:30
19:30 hoelzro|away is now known as hoelzro
hoelzro good evening P6 folk 19:30
timotimo maybe it shouldn't be Enum, but ANum instead 19:31
quietfanatic heheh 19:32
pmichaud speculating a bit, I'd prefer to eliminate "Enum" as a separate type. Just go with "Pair", which binds a name+value pair. The => operator then produces a Pair with the rhs coreced into a scalar container.
hoelzro if I were to submit a test for Perl6, I take it I should submit it to perl6/roast? 19:33
pmichaud if we need a separate Enum type, I'd want it to be for more than a "immutable/mutable" distinction.
hoelzro: yes.
hoelzro: or even better, just commit it directly to roast :)
hoelzro awesome =) 19:34
masak pmichaud: +1
pmichaud in other words, Pair is more analogous to what List does than what Array does.
(List doesn't force mutability of its elements, it just binds to them) 19:35
jnthn pmichaud: To be clear: you're suggesting it's => that delivers the mutability, not Pair itself? 19:36
TimToady
.oO(cons cell)
pmichaud jnthn: yes.
jnthn pmichaud: Seems workable.
pmichaud and if you want a Pair that is bound, use :=>
or something like that.
jnthn More smileys in Perl 6 syntax!
moritz +1 to a separate syntax 19:37
pairs are very often used for named arguments
and those are bound RO by default
pmichaud alternatively, let => indicate the binding operation, and introduce a new one that handles mutability. =$> 19:38
moritz =$> is somwhat neat
quietfanatic Now we're pushing the mutability system onto our operators 19:39
moritz note that hashes could still automatically do mutability
pmichaud quietfanatic: actually, it's more "coerce" or "containerize" than "make mutable"
quietfanatic which, arguably, is what we've been doing all along with ::= and such
pmichaud quietfanatic: correct.
moritz so my %h = a => 1, b => 2; %h<a> = 5; # still works even if => constructs immutable pairs by default 19:40
quietfanatic pmichaud: right
pmichaud moritz: yes, I still believe that Hash is analogous to Array
quietfanatic Well, a Hash isn't stored as a list of Pairs, is it?
moritz it's not
it's only notionally one
quietfanatic I mean, it probably doesn't have actual Pair objects inside of it.
though I suppose it could 19:41
pmichaud and I think it's useful to have a non-container-forcing version of Hash ("Mapping"), but it shouldn't be EnumMap which seems to want Positional characteristics.
(and ordering)
mhasch (and distinct values) 19:42
moritz that seems to be what EnumMap is meant to be, according to S12
quietfanatic Just pun the Associative role?
moritz but then we shouldn't have Hash ~~ EnumMap
quietfanatic or, that role doesn't provide a default implementation so never mind
moritz which is what pmichaud++ said all along, iiuc 19:43
pmichaud moritz: yes. :)
I think EnumMap should be a Mapping that is also ordered.
masak sounds about right.
pmichaud and its pairs are bound directly to the enumerated values
hoelzro if I add a new (Rakudo failing) test to roast, should I mark it as "fudged" for Rakudo
pmichaud (currently called "Enumerations")
moritz hoelzro: yes 19:44
hoelzro is still getting his head around all of the projects and how they fit together
pmichaud hoelzro: yes, that would be helpful. But getting the test into roast is more important to us than having it properly fudged.
hoelzro ok
pmichaud i.e., we can fudge the test if we discover it doesn't work. we're mostly interested in having the test.
quietfanatic still wants mutability as a graspable aspect of the type system 19:45
more specifically, I'd like an easy mechanism to point to a mutable object and say "I want this except immutable"
and vice-versa
masak there's something to be said for that. 19:46
pmichaud quietfanatic: there's been a lot of discussion about "readonly" objects.
masak but I believe it's not as easy as it sounds.
based on exactly those discussions.
TimToady graspable, but not shoved in the face of newbies
quietfanatic yes
actually, the distinction between 'value' and 'container' types might be what I'm looking for 19:47
at least, for making immutable things mutable
like we have with putting Scalar objects in a List 19:48
pmichaud I need lunch; I'll bbiaw
quietfanatic but the other way around is a little more tricky
maybe we don't need the other way around as much
19:49 kaare_ left
moritz well, I'd like a simple way to mark user-defined classes so that their objects are immutable once they are created 19:49
for example Date objects are immutable, but the compiler doesn't really know that
there just aren't any methods that mutate the values
TimToady so the compiler knows it, it just doesn't know it knows it 19:50
19:50 bonsaikitten joined
TimToady if this was Haskell we'd s/class/data/ for immutables, and confuse everyone 19:51
it always bothers me that "data" doesn't mean "data" in Haskell 19:52
quietfanatic it's more like 'struct'
19:52 am0c left
TimToady data means metadata, and type means metatype :) 19:53
quietfanatic I...don't think either of those is correct :)
type is equivalent to typedef, by the way, so it really is type. 19:54
TimToady well, but my impression is that they're really type classes, somewhat generic
quietfanatic 'class' makes a type class.
TimToady well maybe it's class is metaclass I was thinking of 19:55
quietfanatic Actually, it's more like if you could templatize a typedef in C++
that's more like it, yeah
(re: metaclass) 19:56
('type' is more like a templatized typedef)
19:56 birdwindupbird joined
TimToady FP languages tend to unify the "meta" with the "generic" 19:56
quietfanatic People show off object-oriented features and inheritance and such in Haskell's type classes, but they're hard to use since they're two levels of abstraction above normal object. 19:57
normal classes are only one level of abstraction up.
TimToady we put generics and metas 2 up, and classes 1 up in P6 19:58
quietfanatic generics?
TimToady roles
quietfanatic metas?
masak generic roles.
quietfanatic ah
TimToady .HOW
masak well, the MOP in general. :)
TimToady actually, .HOW is 3 up, since roles also have meta 19:59
quietfanatic I think of a role as one level up though, just in kind of a different direction :)
hoelzro yay! first pull request to perl6/roast!
jnthn Roles have more than one manifestation in meta-space anyway :)
quietfanatic Also, there's a unique .HOW for every class, right? That suggests to me that it's on the same level
maybe instead of levels of abstraction I mean levels of generalization.
jnthn The default MOP makes it that way, yes.
TimToady a unique metaobject, not necessarily a unique metaclass 20:00
quietfanatic eh? huh? >.<
quietfanatic is now getting confused
what is a metaclass?
TimToady where metaclass means the class of the metaobject
jnthn All of the meta-objects for classes are instances of the class Meatmodel::ClassHOW.
masak there's a unique *instance* of ClassHOW per class.
jnthn Right.
quietfanatic Aah 20:01
TimToady Meatmodel, yum
masak there's only one ClassHOW type, by default.
quietfanatic: sorry, the terminology is confusing.
I tend to avoid the word "metaclass" because it's overloaded :)
quietfanatic so ClassHOW is on the third leve.
masak I'd say it's second level still.
quietfanatic *level
jnthn uh, Metamodel :P
TimToady metacircularity at that level
masak .oO( I never metacircularity I didn't yo dawg ) 20:02
quietfanatic yeah, at the point where classes generalize over classes, you don't really have levels any more.
jnthn Though if you chase 6model down far enough you get to "classes not invented yet". :)
masak \o/
that's *awesome*.
you could build anything!
jnthn Being a class is really just a function of having a meta-object that makes you behave like one.
quietfanatic in fact, I'd say that a metaobject is already breaking the levels, being a first-level thing (an object) describing a second-level thing (a class)
TimToady hence the need to cheat on .HOW's args 20:03
masak .oO( in Perl 6, classes aren't built in, they're just a question of attitude )
quietfanatic stay classy, #perl6
jnthn I don't really thing of it in "levels", tbh. A meta-object is just an object that describes how another object works.
moritz so it's all(1, 3 ... *)
:-)
masak moritz: at the very least! :P 20:04
TimToady but it takes 2 to tango!
masak guys, you're making poor Bertrand Russel faint.
Russell*
quietfanatic In Haskell you have to be acutely aware of what level you're working on.
of course.
moritz you have to in any language
TimToady s/ <?after aware>.* //
moritz it's just that in Haskell the type checker will smite you if you forgot 20:05
quietfanatic What I mean is, in Haskell the three levels are very distinct and completely incompatible.
whereas Perl 6 contains all these back doors that make them the same level in the end. 20:06
masak in Smalltalk, too.
but looking at the Smalltalk MOP nowadays makes it clear that it was married to inheritence in a way that no longer feels healthy.
(Smalltalk 80, that is) 20:07
TimToady that's so we can teach people using "Oh, by the way, an X is just a special kind of Y."
quietfanatic Classical prototype-based object systems don't feel right to me
for the same reason as making every method virtual by default.
oh wait :)
TimToady well Smalltalk -> traits -> P6 roles
which is why we have non-virtual methods too; they're just not the default like in C++ 20:08
and they're really just disguised subs 20:09
quietfanatic inheritance in general is kind of a broken concept IMO
what do you mean?
submethods?
TimToady no, private methods
quietfanatic but private methods are private.
TimToady yup :)
but you can trust others with the care and feeding of them
we just don't make it easy 20:10
masak we have subs *and* submethods *and* private methods -- and the latter two can be combined.
TimToady we want people to do the right thing by accident most of the time
masak and I find I'm using all of those hiding mechanisms.
quietfanatic I don't think inheritance is the right thing though
TimToady uh, no, submethods have different visibility from private methods
masak aye.
moritz aye
TimToady submethods live in normal method namespace 20:11
jnthn masak: Comining submethods and private methods is odd given private methods are non-virtual anyway. :)
quietfanatic The problem with inheritance and virtual methods is that they make each class actually mean two things:
moritz can submethods be multi?
masak jnthn: hm, I might not have done private submethods, come to think of it :)
jnthn (I presume you mean submethod !foo() { })
masak moritz: yes.
jnthn: yes.
TimToady 'course, even virtual methods are just functions if you scratch away the dispatcher
jnthn moritz: Mumble.
quietfanatic the interface the class exposes, and the default implementation of that interface, which have the same name and can't be separated.
masak moritz: private methods can be multi too, just not in Rakudo.
TimToady: yes, but late binding makes them operationally very different even if they are "just functions". 20:12
quietfanatic default-virtual functions make keeping invariants pretty hard too. 20:13
s/functions/methods
well, at this point I'm just ranting. I doubt I'll actually be able to change anything.
moritz quietfanatic: we do encourage role usage in Perl 6 for those reasons
TimToady masak: sure, but that's all in the dispatch that I just scratched away
moritz quietfanatic: and I guess there are places in core where we could more stuff out of classes and into roles 20:14
TimToady quietfanatic: invariants are supposed to be calculatable at CHECK time, assuming there have been no declarations to keep a class open or non-final
we just default finalization opposite to Java 20:15
moritz quietfanatic: so, concrete feedback on how to factor things in more reusable ways are very welcome
quietfanatic moritz: Roles are better, in that they encourage separation of interface from implementation, but they still allow you to do weird wrong things with them.
TimToady alternately, you can look at roles as final classes done right
quietfanatic (if you aren't disciplined)
moritz quietfanatic: you won't get Haskell-level type safety in p6, no matter what you'll try
TimToady well, I suppose one could define the Haskell subset(superset) of Perl 6 20:16
moritz it's still a dynamic language, late bound and thus less safe
quietfanatic TimToady: Explicit invariants made with 'where' clauses, sure, but other properties of classes you expect to hold may not hold if a subclass can override bits of it.
yeah 20:17
moritz: I get that, yeah
TimToady that's what "final" means, there are no subclasses
quietfanatic I'm more concerned with, as TimToady said, getting people to do the right thing by accident most of the time 20:18
and the majority of object-oriented tutorials explain things with inheritance and virtual methods.
masak r: role R[::T] { multi foo(T $x) { say "{T} $x!" } }; class C does R[Str] does R[Int] {}; given C.new { .foo("OH HAI"); .foo(42) } 20:19
p6eval rakudo 843534: OUTPUT«No such method 'foo' for invocant of type 'C'␤ in block at /tmp/s7ey2sVBWu:1␤␤»
masak huh.
moritz hoelzro: you now have commit access to roast, and a bunchof other repos in the perl6 organization
ingy moin
masak r: role R[::T] { multi method foo(T $x) { say "{T} $x!" } }; class C does R[Str] does R[Int] {}; given C.new { .foo("OH HAI"); .foo(42) }
p6eval rakudo 843534: OUTPUT«use of uninitialized value of type Str in string context in method foo at /tmp/7S5gG1LU8V:1␤␤ OH HAI!␤use of uninitialized value of type Int in string context in method foo at /tmp/7S5gG1LU8V:1␤␤ 42!␤»
masak r: role R[::T] { multi method foo(T $x) { say "{T.^name} $x!" } }; class C does R[Str] does R[Int] {}; given C.new { .foo("OH HAI"); .foo(42) }
p6eval rakudo 843534: OUTPUT«Str OH HAI!␤Int 42!␤»
masak \o/
that's pretty nice.
20:20 SamuraiJack left
TimToady S12:188 and S12:2130 talk about closing and finalizing, but we don't implement it yet 20:20
quietfanatic (now I'm just complaining. Come one, TimToady, say something sharp and put me in my place :)
TimToady oh, is that what you're gunning for? :)
quietfanatic not exactly, but whatever
ingy Scissors! 20:21
hoelzro moritz: thanks!
masak pictures ingy running with scissors
ingy: be careful! o.O
hoelzro should I just push my change to roast?
TimToady when I was eating lunch and watching quietfanatic++ enter battle on #perl6, I was desperately trying not to think about Denethor :)
moritz hoelzro: you should adjust your system clock :-) 20:22
ingy masak: I've made it this far running with scissors :)
moritz hoelzro: temp $value = $value + 1; # doesn't do what you think, I guess
hoelzro moritz: what *should* it do?
TimToady we always raised our kids by treating them like kids when they behaved like kids, and treating them like adults when they behaved like adults
moritz hoelzro: temp acts immediately 20:23
hoelzro eesh, my clock is f'ed up
TimToady thinks quietfanatic++ has been behaving like an adult
moritz hoelzro: ie it's like temp $value; $value = $value + 1
quietfanatic So behaving like an adult is going online and fussing about things you don't like?
moritz hoelzro: which means that $value is Any, and you get a warning from the +
quietfanatic :)
hoelzro moritz: those two are equivalent?
moritz quietfanatic: there's fussing and there's fussing
hoelzro: yes
hoelzro ok, that's what I expected
masak r: say "fussing" ne "fussing" 20:24
p6eval rakudo 843534: OUTPUT«False␤»
hoelzro in Perl 5 speak, it's local $value = $value; $value += 1, right?
masak moritz: Rakudo claims they're the same, though :P
jnthn r: my $a = 1; { temp $a = $a + 1; say $a; }; say $a
p6eval rakudo 843534: OUTPUT«2␤1␤»
moritz hoelzro: no
20:24 cognominal joined
jnthn temp doesn't throw away the old value 20:24
moritz erm
wait
now I'm confused
masak why?
jnthn That's what "my" would do :)
hoelzro me too =/
masak nope. 20:25
moritz in p5, 'local $x = $x' the $x on the RHS is the old $x, right?
masak r: my $a = 1; { my $a = $a + 1; say $a; }; say $a
hoelzro temp is like local, isn't it?
p6eval rakudo 843534: OUTPUT«use of uninitialized variable $a of type Any in numeric context in block at /tmp/sOzU03jliG:1␤␤1␤1␤»
moritz and in p6 it's the new $x, right?
masak r: my $a = 1; { my $a = $OUTER::a + 1; say $a; }; say $a
p6eval rakudo 843534: OUTPUT«2␤1␤»
masak right.
moritz hoelzro: temp is like local, but I think the time it takes place is different
jnthn temp takes effect where it's written.
moritz right 20:26
jnthn It doesn't declare a new variable.
ingy masak: is there a p6 implementation that supports pcre style regexes yet?
jnthn It just makes sure that we restore the original value.
hoelzro ok, hang on
jnthn (at LEAVE time)
hoelzro temp $x; is like local $x = $x;
right?
TimToady yes 20:27
yet another default we changed
masak ingy: no, I don't think so.
pmichaud would know, though.
didn't PGE have one? 20:28
mhasch can you temp a lexical variable?
TimToady didn't pugs have one?
mhasch: yes
hoelzro so temp $x = $x +1; saves the value of $x and increments it in the "new" version, right?
so before: 0 after: 1?
TimToady you can just say temp $x += 1
masak TimToady: Pugs had some kind of bridge to Perl 5 there, methinks. 20:29
TimToady that won't work with local
hoelzro ok, that's what I thought.
so...doesn't my test make sense?
moritz hoelzro: I think it does, sorry for all the noise 20:32
hoelzro ok, cool =)
should I just push it up to roast? 20:33
moritz yes
hoelzro ok =
dalek ast: e0f4d0f | (Rob Hoelz)++ | S04-blocks-and-statements/temp.t:
Add test for temp in recursive calls
20:36 pmurias joined
pmurias hi 20:36
is it normal for nqp code to segfault when it's unable to match a multi?
moritz I hope not :-) 20:38
nqp: multi f(int $x) { say('Int') }; f('str')
p6eval nqp: OUTPUT«Sorry, no proto sub in scope, and auto-generation of protos is not yet implemented. at line 2, near "; f('str')"␤current instr.: 'nqp;HLL;Grammar;panic' pc 22407 (src/stage2/gen/NQPHLL.pir:8553) (src/stage2/gen/NQPHLL.pm:326)␤»
moritz nqp: proto f(|$) {*}; multi f(int $x) { say('Int') }; f('str')
p6eval nqp: OUTPUT«Routine declaration requires a signature at line 2, near "(|$) {*}; "␤current instr.: 'nqp;HLL;Grammar;panic' pc 22407 (src/stage2/gen/NQPHLL.pir:8553) (src/stage2/gen/NQPHLL.pm:326)␤»
moritz nqp: proto f(|$) {*}; multi f(int $x) { say('Int') }; f('str')
p6eval nqp: OUTPUT«Routine declaration requires a signature at line 2, near "(|$) {*}; "␤current instr.: 'nqp;HLL;Grammar;panic' pc 22407 (src/stage2/gen/NQPHLL.pir:8553) (src/stage2/gen/NQPHLL.pm:326)␤»
moritz nqp: proto f(*@) {*}; multi f(int $x) { say('Int') }; f('str')
p6eval nqp: OUTPUT«Routine declaration requires a signature at line 2, near "(*@) {*}; "␤current instr.: 'nqp;HLL;Grammar;panic' pc 22407 (src/stage2/gen/NQPHLL.pir:8553) (src/stage2/gen/NQPHLL.pm:326)␤»
masak pmurias! \o/
moritz nqp: class A { }; proto f(*@) {*}; multi f(A $x) { say('A') }; f('str') 20:39
p6eval nqp: OUTPUT«Routine declaration requires a signature at line 2, near "(*@) {*}; "␤current instr.: 'nqp;HLL;Grammar;panic' pc 22407 (src/stage2/gen/NQPHLL.pir:8553) (src/stage2/gen/NQPHLL.pm:326)␤»
moritz nqp: class A { }; proto f(*@args) {*}; multi f(A $x) { say('A') }; f('str')
p6eval nqp: OUTPUT«(signal SEGV)»
masak o.O
masak submits nqpbug
moritz pmurias: I'm sure that's not intended behavior :-)
masak github.com/perl6/nqp/issues/49 20:40
moritz pmurias++ 20:41
masak++
jnthn Maybe that you're passing it something that ain't a 6model object.
moritz nqp: class A { }; class B { }; proto f(*@args) {*}; multi f(A $x) { say('A') }; f(B);
p6eval nqp: OUTPUT«(signal SEGV)»
moritz jnthn: I think B is a 6model object too 20:42
jnthn ah
Then I've no idea what's going on.
Investigation welcome.
pmurias jnthn: QAST::Block 20:43
jnthn That's one.
Well, all of QAST::Compiler hangs off multi-dispatch on QAST nodes, so clearly it can work, if you align the stars correctly. ;-) 20:44
pmurias jnthn: I copied the catch all unknown error reporting thing 20:45
jnthn pmurias: I'm...not entirely sure which thing you're meaning? 20:46
moritz r: macro funny_nil { quasi { {;}() } }; say funny_nil() === Nil 20:48
p6eval rakudo 843534: OUTPUT«True␤»
moritz masak: that test passes in t/spec/S06-macros/macros-d1.t since the toqast merge
20:48 zby_home_ left
moritz (erm, is still skipped, can be unfudged) 20:48
masak ooh 20:49
moritz++
moritz autounfudge it an endless source of karma, it seems :-)
I should really get a patent on it! 20:50
pmurias jnthn: multi method as_post($unknown) {...}
jnthn pmurias: Ah, OK... Did that help?
moritz I mean, on a machine doing an autounfudge :-)
masak that might even work. 20:51
pmurias jnthn: it replaces a segfault with an error message
if I add a new file to rakudo do I have to register it in the build system? 20:53
moritz pmurias: yes
pmurias: you need to add it to tools/build/Makefile.in 20:54
dalek ecs: 5dfd996 | larry++ | S04-control.pod:
resumable failures for coercions

fixes #17
ast: 8a9dbc0 | moritz++ | S (4 files):
various rakudo unfudges
 20:57
pmurias is there vim syntax highlighting for nqp? 20:59
or should I just use the regular p6 one 21:00
jnthn Even Perl 5 highlighters make a decent job of NQP, apart from regexes, where things mess up...
Well, grammars especially.
moritz the p6 one is quite OK for nqp
jnthn Ah, I missed the "I ahve a p6 one to hand" bit of that. 21:01
pmurias jnthn: I only thought about the p6 one after asking the question ;) 21:02
21:03 Coleoid left
pmurias is QAST documented somewhere? 21:04
dalek ast: 45454e3 | moritz++ | S04-blocks-and-statements/temp.t:
clean up temp + recursion tests

  * remove redundant tests
  * add test labels
  * fudge only the failing test for rakudo		 21:06
masak lol, I blogged! \o/ strangelyconsistent.org/blog/july-2...t-synonyms 21:08
pmurias: no, but the source code is in nqp, and very approachable. 21:09
as in, both "in the nqp repo" and "written in nqp".
21:10 bruges_ left
jnthn It really wants some decent docs, but I didn't get a round tuit yet. 21:10
Now that we know that it actually works well enough to build Rakudo though... :)
dalek ecs: c939721 | larry++ | S (3 files):
clean up some ucfirst fossils

fixes #15		 21:11
21:11 snearch left, bruges joined
hoelzro hey guys, I wrote up a test illustrating a strange issue I saw when working on my chat bot: gist.github.com/3194797 21:12
it currently doesn't pass; is this a bug, or am I being stupid?
masak hoelzro: rt.perl.org/rt3/Ticket/Display.html?id=113958 21:13
hoelzro ah, thanks
masak bows
doesn't solve your problem, of course :/
but now you know it's not just you.
hoelzro oh, haha
I forgot about that =/
masak hah, it was you last time too :P 21:14
hoelzro I know!
masak this is why we need new people: they find new bugs for us
again and again :P
hoelzro I spent too much time getting a working Rakudo =P
well, that's part of why I decided to write my bot in Perl6
can anyone suggest a workaround? 21:15
masak hoelzro: do you know about github.com/TiMBuS/Net--IRC/ , by the way?
hoelzro masak: I did not! 21:16
my chat bot is for XMPP, though =)
masak ah.
still, might be good inspiration.
hoelzro so I just need to write an XML parser
oh, good point!
masak oh noes not another XML parser... :P
hoelzro is there one for Perl6? 21:17
masak hoelzro: why in the world would you want to parse XML in Perl 6, aside for the pure exercise value of writing a moderately big grammar?
"write an XML parser" seems to be a phase that about 15% of the neophytes go through...
I haven't seen a complete one, no.
hoelzro masak: oh, I don't care if it's written in Perl6 or not; I would just need an XML parser for an XMPP client
masak oh, good.
hoelzro right now I'm using a very dodgy hack
I'd rather not waste my time writing an XML parser =) 21:18
I have cooler stuff to do!
masak indeed. 21:19
(and writing an XML parser may look trivial, but... it isn't.)
timotimo huh? i'll just throw a regex at it and voila! 21:20
arnsholt hoelzro: If you feel up for the challenge, you could try wrapping libxml with Zavolaj 21:21
Iæm not quite sure if quite all the pieces required are there, but most of them should be there
hoelzro after I get my bot working, I might tackle it =)
arnsholt If you find out what's missing, I can probably implement them =)
hoelzro it's a pretty important library to have!
masak aye. 21:22
masak would love to have it 21:23
arnsholt Quite. But now, sleep
zzzzz &
21:23 birdwindupbird left
masak ...and to have nice repr thingies for access from Perl 6. 21:23
arnsholt: 'night. dream of people wrapping cool things with Zavolaj.
sorear good * #perl6 21:28
pmurias sorear: hi
is there a way to dump the QAST for a given bit of code?
--target=past gives me strange errors 21:29
masak sorear! \o/
pmichaud jnthn: ping 21:32
jnthn pmichaud: pong
pmichaud I have a new .msi created; could you test it and see if it installs to your c: drive this time? 21:33
jnthn sure; url?
pmichaud just a sec, uploading
jnthn
.oO( mmm...url...sounds like öl... )
pmichaud pmichaud.com/sandbox/rakudo-star-2012.07-3.msi 21:34
jnthn pmichaud: Trying. 21:35
pmichaud: yes, installed on C: and works \o/ 21:39
pmichaud okay, I'll replace the one on github with this one then.
jnthn pmichaud++ 21:40
pmichaud thanks for the quick check 21:41
jnthn np 21:42
Just waiting for builds anyway.
jnthn is debugging building Rakudo using PIRT
dalek ar: d961bd7 | pmichaud++ | skel/tools/build/star-product.wxs:
Update Wix file to force installation to C:\ .
 21:44
pmichaud okay, I'm going to go play with robots for a while. :-) bbl 21:45
jnthn Have fun! 21:49
21:56 larks left 21:58 larks joined 21:59 seldon left
hoelzro is there a way to compile a regular expression from a string without using eval? 21:59
ex. Regex.new("^ $foo")
japhb What base class do Perl6::Metamodel::* classes have? What metaclass do they have? Is there a way to introspect them? 22:01
jnthn /^ <$foo>/ 22:02
hoelzro jnthn: thanks! 22:03
22:03 kaxing left
jnthn japhb: They're weird; initial NQPMu but then post-bootstrap they are twiddled to have Any. 22:03
22:04 kaxing joined
japhb jnthn, Ah, interesting. 22:04
22:05 larks left
japhb Are they all supposed to have metaclass Perl6::MetaModel::ClassHOW, just like "normal" classes? 22:05
22:07 larks joined
jnthn No 22:07
japhb r: say Perl6::Metamodel::ClassHOW
p6eval rakudo 843534: OUTPUT«Could not find symbol '&ClassHOW'␤ in method <anon> at src/gen/CORE.setting:9629␤ in <anon> at src/gen/Metamodel.pm:2304␤ in any find_method_fallback at src/gen/Metamodel.pm:2302␤ in any find_method at src/gen/Metamodel.pm:843␤ in <anon> at src/gen/BOOTSTRAP.p…
jnthn r: say Metamodel::ClassHOW
p6eval rakudo 843534: OUTPUT«Perl6::Metamodel::ClassHOW()␤»
jnthn Hmm. It lies :)
japhb Oh interesting
jnthn Anyway, no, it's own .HOW is actually an instance of NQPClassHOW. 22:08
*its
japhb Is that true of all Metamodel classes, or just ClassHOW?
jnthn All.
japhb Ah, OK. 22:09
jnthn It's factored primarily with roles.
The only bit of inheritnace in there is that GrammarHOW inherits from ClassHOW.
japhb Yeah, I'm getting back to work on my "visualizing the type graph" project, and discovered that the handling of the Metamodel classes/roles appears ... very fishy. 22:10
22:11 larks left
japhb jnthn, Hmmm, I have MethodDispatcher, MultiDispatcher, and WrapDispatcher all inheriting from BaseDispatcher. Is that wrong? 22:11
22:12 lue joined
lue hello world o/ 22:12
masak lue! \o/
lue: you have your inventory now.
jnthn japhb: Ah...that sounds sane.
masak lue: and shortcuts for directions, though not yet for verbs. that comes tomorrow. 22:13
22:13 larks joined
lue reads blog posts 22:13
"So you panic for three moves and then you die." :) 22:16
masak yup.
japhb jnthn, why do the Metamodel classes have names that start with Perl6:: in their definition, but appear in the namespaces without that prefix?
jnthn japhb: They're defined with the prefix in NQP, but exported differently.
japhb: But really I think .WHAT.gist ain't meant to include the fully qualified name. 22:17
masak r: module A { class B::C {}; say B::C.new.^name };
p6eval rakudo 843534: OUTPUT«B::C␤»
masak r: module A { class B::C {} }; say A::B::C.new.^name
p6eval rakudo 843534: OUTPUT«B::C␤»
masak +1
hoelzro is there a way to retrieve a role from a name? 22:18
ex. I have a role 'Hexe::Plugin::Echo'
masak r: role R {}; class C does R {}; say C.new ~~ R
p6eval rakudo 843534: OUTPUT«True␤»
hoelzro and I want to dynamically apply the role of that name to an object
ex. $object does %roles($role);
japhb jnthn, sure, I get the mechanism ... I meant "why" as in "what's the reasoning for that difference?" Is it because Metamodel:: within NQP would refer to NQP's metamodel, so this is how you create something in NQP's world that needs to appear in Perl6's world?
masak r: role R {}; class C {}; my $c = C.new but R; say $c ~~ R 22:19
p6eval rakudo 843534: OUTPUT«True␤»
japhb s/appear/appear under the same name/
22:19 lue left
hoelzro masak: right, but I only have the string 'R' 22:20
tadzik ::('R') may work
jnthn japhb: I'm not sure there's a reason beyond me defining it one way in the first place, and then feeling it was a bit unweildy. 22:21
japhb Ah, OK
22:21 lue joined
dalek p/pirt: e27a5eb | jnthn++ | src/ (2 files):
We still need to keep the Regex nodes emitting POST for the sake of NQP (which doesn't use the rest of QAST yet); this should make that possible.
 22:22
p/pirt: 69bc286 | jnthn++ | src/QAST/PIRT.nqp:
Labels need to automatically be uniquified.
p/pirt: 8c14cea | jnthn++ | src/Q (2 files):
Various fixes to regex compilation; this also means we'll be able to eliminate the PAST::Node usages that remain in Rakudo's Actions.
p/pirt: 1f5406a | jnthn++ | src/QAST/Compiler.nqp:
One more regex tweak to get regex compilation under PIRT OK.
p/pirt: dc95f4a | jnthn++ | src/QAST/PIRT.nqp:
For hysterial raisons, PIRT nodes should stringify to their results (for now).
p/pirt: 17a5ef3 | jnthn++ | src/QAST/Compiler.nqp:
Final fix needed to get PIRT able to build Rakudo.
masak r: role R {}; class C {}; my $c = C.new but ::('R'); say $c ~~ R
p6eval rakudo 843534: OUTPUT«True␤»
lue
.oO(stupid randomly panicking kernel...)
tadzik jnthn: what's pirt?
masak tadzik: PIR in tree representation.
tadzik nice
dalek kudo/pirt: 99d850b | jnthn++ | src/Perl6/Actions.pm:
Initial update to get Rakudo to build with nqp/pirt. Eliminates the final PAST::Node usages also. Some test failures; seems CORE.setting memory is a bit lower with this, which was the aim.
 22:25
jnthn tadzik: It's a replacement for POST that I hope reduces build memory a bit further when doing CORE.setting compilation.
hoelzro tadzik: no dice =/
jnthn tadzik: Today it's gone from not existing to being able to replace POST for the Rakudo build. 22:26
Well, modulo a relatively small number of new test failures.
tadzik awesome
tadzik builds 22:27
dalek kudo/pirt: 2b57323 | jnthn++ | src/Perl6/Grammar.pm:
Fix user defined ops, which probably deals with many of the test regressions.
 22:29
jnthn tadzik: Grab ^^ first.
It probably deals with most of the new failures. 22:30
lue considers attempting pirt 22:32
jnthn lue: It's got some rough edges and I won't get to them today.
Now too many though, looking at this test run. 22:34
Yeah, few enough test files that you can count 'em on your fingers. Not bad to say I replaced an entire compilation stage in a day's hacking. :) 22:35
japhb jnthn++ 22:37
dalek c: fe96631 | (Geoffrey Broadwell)++ | type-graph.txt:
Fix missing superclass: Perl6::Metamodel::GrammarHOW is Perl6::Metamodel::ClassHOW
c: e0db752 | (Geoffrey Broadwell)++ | lib/Perl6/TypeGraph.pm:
Defaulting superclass to Any should happen after roles get a chance to supply superclasses
jnthn Seems my first cut of the code is no slower (despite being in NQP - a HLL - rather than PIR) and sees us use less memory. 22:38
japhb That is rather awesome. 22:39
22:39 MayDaniel left
diakopter 2 22:43
dalek ecs: 76c977a | larry++ | S05-regex.pod:
:p is incompatible with scanners

  :p may not be used with :ex, :ov, :g, :nth, or :x modifiers.
fixes #16		 22:45
jnthn OK, time for some rest. 22:49
'night o/
tadzik 'night! 22:53
22:56 hoelzro is now known as hoelzro|away 23:08 sudokode left 23:10 spider-mario left, sudokode joined
dalek ecs: 003891a | larry++ | S0 (3 files):
Explication of lazy warnings

Lazy warning are to be used for +, ~ and ? conversions, as well as for the ??? stub.		 23:23
ecs: 838a4b9 | larry++ | S04-control.pod:
relax prev to numeric/string conversion

This is all Stringy/Numeric conversion, not just the ~/+ prefix operators.
  (And Bool conversion always succeeds, so ? need not apply).
23:32 larks left, larks joined 23:42 takadonet left 23:47 pmurias left

Powered by Cro and the Raku® Programming Language.

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

Thank you!