perl6-projects.org/ | nopaste: sial.org/pbot/perl6 | evalbot: 'perl6: say 3;' | irclog: irc.pugscode.org/ | UTF-8 is your friend!
Set by moritz_ on 4 May 2009.
00:01 rdice left
jnthn Well, Rakudo builds under it fine here. Haven't benchmarked it, but it'd probably shave some off our spectest run if it improves every startup... 00:21
Anyways, sleep...night all
00:25 buubot left 00:26 buubot joined 00:31 bacek__ joined 00:36 Kisu left 00:40 [particle] left, Kisu joined 00:42 wollmers left, eternaleye joined 00:52 nbrown joined 01:04 avar joined 01:10 avar left, avar joined 01:15 agentzh left
s1n so was it changed to have self refer to whatever is constructed at the time of use? 01:19
dalek kudo: r4d7fe56 | pmichaud++ | src/ops/perl6.ops:
Revise d86c014 (NotFound++) a bit so that we're not mixing our types as much.
 01:21
01:21 ZuLuuuuuu joined 01:26 Whiteknight left 01:29 c9s_ joined
meppl good night 01:29
01:31 c9s left, meppl left 01:32 c9s_ is now known as c9s 01:38 agentzh joined 01:40 ZuLuuuuuu left 02:01 cdarroch left, Kisu left 02:10 s1n left, hercynium left 02:11 s1n joined 02:19 alester_ joined 02:32 nihiliad joined
s1n moritz_: ping 02:44
02:50 avar left, avar joined 02:51 alester_ left 03:02 nbrown left 03:14 alester_ joined
mikehh I am still failing t/spec/S09-subscript_slice/slice.rakudo - rakudo (4d7fe56) on parrot r38475 03:15
./perl6 t/spec/S09-subscript_slice/slice.rakudo - runs through all the tests says # FUDGED! then Segmentation fault 03:16
03:18 skids left
mikehh that's om Kubuntu Jaunry (9.04) Amd64 03:18
s/om/on/
all the other spectests pass 03:19
in the test - ok 1..11, ok 12..19 (SKIP), ok 20..25, nok ok 26..27 (TODO), # FUDGED!, Segmentation fault 03:25
s/nok/not/ and Jaunty in prev 03:27
avar mikehh: (I'm not a parrot or rakudo dev): If you compile parrot with "-ggdb3 -O0" cflags and run that under gdb you could produce a more useful bug report 03:50
e.g. $ gdb ./perl6
run t/spec/S09-subscript_slice/slice.rakudo
bt full
but your setup is common enough that they'll probably find out about this soon enough if they haven't already
04:03 SamB left 04:06 SamB joined
pmichaud (segfault on slice.t) Yes, I get that segfault also. It's not a rakudobug, really -- it's something segfaulting on Parrot exit. 04:17
04:38 mikehh is now known as mikehh_, mikehh_ is now known as mikehh__ 04:41 mikehh__ is now known as mikehh 04:42 SamB left 04:59 alc joined 05:06 jhuni joined 05:39 nihiliad left 05:43 ejs joined 05:44 orafu joined 05:48 justatheory left 05:55 DemoFreak joined 06:06 wollmers joined 06:12 orafu left 06:13 orafu joined 06:20 masak joined
moritz_ s1n: pong 06:20
masak good morrow, socketcamels. 06:21
mberends good morrow, mask 06:22
moritz_ h h
06:22 ejs left
masak does a Jim Carrey impersonation 06:23
mberends masak: thanks for the netstat info. could you please check out a test for me? 06:24
masak mberends: certainly. 06:25
mberends masakbot: cd ~/rakudo; svn up t/spec; ./perl6 t/spec/S32-io/IO-Socket-INET.t 06:27
masak output: PASS 06:28
mberends :D 06:29
masak mberends++
moritz_ now we need to find out what happens if somebody has no netcat installed 06:30
06:30 FurnaceBoy left
mberends will try that immediately 06:31
moritz_: ok, it works without either netcat or socat. Would you advise inflicting this socket test on spectest users? 06:36
moritz_ mberends: in principle, yes. But I'd like to have jnthn++ test it on windows first 06:37
06:38 eMaX left
mberends O NOES! it cannot work on windows because of the shelly things. It should skip everything silently though. 06:38
06:39 zamolxes joined
mberends but good idea. let's wait until it's rakudo-am in Bratislava. 06:39
masak if ($*OS ~~ Suck) { ... }
mberends in Dutch slang pipe === blowjob 06:40
masak no, I meant the other way around. them that don't hav'em suck. 06:41
mberends true
06:43 meppl joined
mberends with tests in place I have the courage to tinker with IO::Socket::INET a little more. 06:44
masak tests++ 06:48
mberends hands the proto pumpkin to masak
moritz_ let's add it to the spectests.t and have $people complain if it doesn't work somewhere ;-)
masak :) 06:49
mberends ;-) works here! what do you mean, tests fail? must be your setup! 06:50
masak mberends: I'm holding another pumpkin at present. says "$work" on it. have to put the proto pumpkin here on the ground along with the other spheroid objects.
rakudo: say { a => 1, b => 1 }.invert.perl 06:52
p6eval rakudo 4d7fe5: OUTPUTĀ«[1 => "a", 1 => "b"]ā¤Ā»
masak submits rakudobug
06:52 alester_ left, avar left, avar joined, iblechbot joined
moritz_ why? 06:53
mberends duplicate key
moritz_ .invert returns a list.
mberends ah, not a hash 06:54
masak reading spec now.
I don't seem to find that this is an error.
maybe my brain isn't synched with the latest spec...
moritz_ rakudo: my %h; %h.push: { a => 1, b => 1 }.invert; say %h.perl
p6eval rakudo 4d7fe5: OUTPUTĀ«{"1" => ["a", "b"]}ā¤Ā»
masak moritz_: ah. 06:55
masak stands down red rakudo-alert
mberends next yak: pure Perl 6 Net::SMTP for command line rakudobugs 06:56
masak yay!
06:58 ElectricHeavyLan left
moritz_ let's make Perl 6 attractive for spammers ;-) 06:58
mberends imagine the questions in #perl6 in 6 months time 06:59
"hello, I want send email w/attachment. can u help with this?" 07:00
moritz_ hopes that in 6 months the questions will be like this: "this last release was *so* cool - how can I help?" 07:01
masak moritz_: you should blog-post an FAQ with only questions like that. prime people in the right direction. :) 07:03
ok, #perl6. I'm a bit stumped with .invert at present. maybe you can help me.
rakudo: say $*IN.slurp.join.comb.map(-> $k {$k=>1}).perl 07:04
p6eval rakudo 4d7fe5: OUTPUTĀ«["Land" => 1, "der" => 1, "Berge," => 1, "Land" => 1, "am" => 1, "Strome," => 1, "Land" => 1, "der" => 1, "Ƅcker," => 1, "Land" => 1, "der" => 1, "Dome," => 1, "Land" => 1, "der" => 1, "HƤmmer," => 1, "zukunftsreich!" => 1, "Heimat" => 1, "bist" => 1, "du" => 1, "groƟer" => 1,
.."S...
masak now, I'd like to invert this list of pairs.
how do I do that?
should be easy, right?
moritz_ what do you mean by "invert"?
invert each pair?
rakudo: say $*IN.slurp.join.comb.map(-> $k {$k=>1}).map(*.invert).perl 07:05
p6eval rakudo 4d7fe5: OUTPUTĀ«Method 'invert' not found for invocant of class 'Perl6Pair'ā¤current instr.: 'parrot;P6metaclass;!whatever_dispatch_helper' pc 243809 (src/gen_metaop.pir:8)ā¤Ā»
moritz_ rakudo: say $*IN.slurp.join.comb.map(-> $k {$k=>1}).map(*.hash.invert).perl
p6eval rakudo 4d7fe5: OUTPUTĀ«Method 'invert' not found for invocant of class 'Block'ā¤current instr.: 'parrot;P6metaclass;dispatch' pc 243614 (src/gen_actions.pir:0)ā¤Ā»
moritz_ rakudo: say $*IN.slurp.join.comb.map(-> $k {$k=>1}).map({.hash.invert}).perl 07:06
p6eval rakudo 4d7fe5: OUTPUTĀ«[1 => "Land", 1 => "der", 1 => "Berge,", 1 => "Land", 1 => "am", 1 => "Strome,", 1 => "Land", 1 => "der", 1 => "Ƅcker,", 1 => "Land", 1 => "der", 1 => "Dome,", 1 => "Land", 1 => "der", 1 => "HƤmmer,", 1 => "zukunftsreich!", 1 => "Heimat", 1 => "bist", 1 => "du", 1 => "groƟer",
..1 ...
masak thank you.
moritz_ (chained method calls don't work with *)
masak in the past few seconds, I've realized I don't need to invert. :) 07:07
but that's ok, I have a new problem.
rakudo: %*h.push: $*IN.slurp.join.comb.map(-> $k {$k=>1}); say %*h.perl
p6eval rakudo 4d7fe5: OUTPUTĀ«{}ā¤Ā»
masak I would expect something non-empty.
moritz_ rakudo: our %*h; %*h.push: $*IN.slurp.join.comb.map(-> $k {$k=>1}); say %*h.perl 07:08
p6eval rakudo 4d7fe5: OUTPUTĀ«get_pmc_keyed_str() not implemented in class 'String'ā¤current instr.: '_block14' pc 68 (EVAL_20:51)ā¤Ā»
masak huh?
moritz_ IMHO both are "ouch"
masak double rakudobug coming up!
rakudo: %*h.push: 1 => 2; %*h.perl.say 07:10
p6eval rakudo 4d7fe5: OUTPUTĀ«{}ā¤Ā»
masak rakudo: our %*h 07:13
p6eval rakudo 4d7fe5: OUTPUTĀ«get_pmc_keyed_str() not implemented in class 'String'ā¤current instr.: '_block14' pc 63 (EVAL_11:49)ā¤Ā»
masak rakudo: our $*h 07:14
p6eval rakudo 4d7fe5: ( no output )
dalek kudo: r4d3a38f | (Moritz Lenz)++ | t/spectest.data:
run socket tests, mberendes++
Matt-W Morning 07:20
masak Matt-W: o/
rakudo: my %h; %h.push: $*IN.slurp.join.comb.map(-> $k {$k=>1}); say +.value, "\t", .key for %h.pairs.sort({ -.value })
p6eval rakudo 4d3a38: OUTPUTĀ«5 Landā¤4 derā¤3 Ɩsterreich,ā¤3 Ɩsterreich!ā¤2 duā¤2 vielgerĆ¼hmtesā¤2 vielgeprĆ¼ftesā¤2 inā¤2 undā¤2 vielgeliebtesā¤1 Berge,ā¤1 amā¤1 Strome,ā¤1 Ƅcker,ā¤1 Dome,ā¤1 HƤmmer,ā¤1 zukunftsreich!ā¤1 Heimatā¤1 bistā¤1 groƟerā¤1 Sƶhne,ā¤1
..Volk,ā¤1 begnadetā¤1 fĆ¼rā¤1 dasā¤1 Schƶne,ā¤1 HeiƟā¤1 umfehdet,ā¤1 wildā¤1 umst...
masak should have taken the 10 most frequent. 07:21
rakudo: my %h; %h.push: $*IN.slurp.join.comb.map(-> $k {$k=>1}); say +.value, "\t", .key for %h.pairs.sort({ -.value })[^10]
p6eval rakudo 4d3a38: OUTPUTĀ«5 Landā¤4 derā¤3 Ɩsterreich,ā¤3 Ɩsterreich!ā¤2 duā¤2 vielgerĆ¼hmtesā¤2 vielgeprĆ¼ftesā¤2 inā¤2 undā¤2 vielgeliebtesā¤Ā»
mberends nice
masak I'm a bit bummed at the '-> $k { $k => 1 }' thing, though. 07:22
07:22 ejs joined
masak it's needed because just '{ $_ => 1 }' is interpreted as a hash, and map gets sad. 07:24
maybe overloading map to take a Hash would save a lot of future FAQ grief...
OTOH, maybe it would cause some grief also. 07:25
mberends implementing map on a hash might delay Christmas :/ 07:26
masak mberends: no, that's not what I meant.
mberends: I meant that when the Code is accidentally a Hash, it should DWIM. 07:27
for example [email@hidden.address] $_ => 1 })'
it looks like it should work...
Matt-W There's some rule about when {} is a Hash and when it's a Callable
masak Matt-W: aye. 07:28
mberends whitespace as syntax ?
masak as far as I'm concerned, it's a good rule.
07:28 eMaX joined
masak it's still a bit surprising when it bites you, though. 07:28
mberends: no, whitespace is not a part of it.
meeting & 07:29
07:29 jhuni left
moritz_ if it's a list of pairs inside, it's a hash. otherwise it's a block. 07:31
Matt-W and of course $_ => 1 looks like a pair
moritz_ {; $_ => 1 } is the common Perl 5 disambiguation
07:34 ejs1 joined, H1N1[A] left
Matt-W Isn't this somewhere that the * would be useful? 07:35
moritz_ like * => 1? 07:37
rakudo: say <a b c>.map(* => 1).perl
p6eval rakudo 4d3a38: OUTPUTĀ«Parameter type check failed; expected something matching but got something of type Pair() for expr in call to mapā¤current instr.: 'die' pc 16538 (src/builtins/control.pir:225)ā¤Ā»
moritz_ no. 07:38
rakudo: say (* => 1).perl
p6eval rakudo 4d3a38: OUTPUTĀ«!whatever_dispatch_helper => 1ā¤Ā»
07:38 ejs left
moritz_ rakudo: say *.perl 07:38
p6eval rakudo 4d3a38: OUTPUTĀ«!whatever_dispatch_helperā¤Ā»
moritz_ that looks so wrong. 07:39
07:39 H1N1[A] joined
moritz_ summons masak 07:39
Matt-W rakudo: say (<a b c>.map: * => 1).perl
p6eval rakudo 4d3a38: OUTPUTĀ«Parameter type check failed; expected something matching but got something of type Pair() for expr in call to mapā¤current instr.: 'die' pc 16538 (src/builtins/control.pir:225)ā¤Ā»
Matt-W bah 07:40
okay so you can't fool it into doing that
std: say <a b c>.map(* => 1).perl
p6eval std 26674: OUTPUTĀ«ok 00:02 35mā¤Ā»
moritz_ it just constructs a Pair
07:40 H1N1[A] is now known as H1N1
Matt-W You can do things like map(*.methodname) though 07:41
interesting
Is that a bug, or intentional
moritz_ I don't know 07:42
Matt-W Neither do I
Maybe masak knows
When his meeting is done
moritz_ hopes that TimToady knows :-)
Matt-W Yes I hope so too 07:44
Because if he doesn't...
07:47 ejs2 joined 07:58 ejs1 left
moritz_ just created his first binary from a Perl 6 program 08:17
08:18 donaldh joined 08:21 elmex joined, bacek__ left
masak is back from the meeting 08:28
I think it's a bug.
masak submits rakudobug
hm, wait. 08:33
where are expressions like (* - 1) documented in the spec? S09?
moritz_ S02 08:34
Most of the built-in numeric operators treat an argument of * as indicating the desire to create a function of a single unknown, so: 08:35
* - 1
produces a function of a single argument:
{ $_ - 1 }
masak 'numeric'.
so, not a bug.
moritz_ no, it's the .perl output that looks weird to me
masak oh. 08:36
masak resumes submitting
moritz_ rakudo: say (* => 1).perl
p6eval rakudo 4d3a38: OUTPUTĀ«!whatever_dispatch_helper => 1ā¤Ā»
masak well, that's obbviously wrong.
pugs_svn r26675 | wollmers++ | precised reason for rakudo skip 08:37
08:39 elmex left 08:44 elmex joined
Matt-W so is that perhaps a bug in infix:<=>> 08:45
moritz_ rakudo: say *.perl
p6eval rakudo 4d3a38: OUTPUTĀ«!whatever_dispatch_helperā¤Ā»
moritz_ no. 08:46
it's much simpler :)
masak it's just piping showing through.
moritz_ rakudo: say *.isa(Whatever)
Matt-W Oooh your piping is showing! Scandalous!
p6eval rakudo 4d3a38: OUTPUTĀ«!whatever_dispatch_helperā¤Ā»
moritz_ there is a slight problem 08:47
masak rakudo: say ?(*)
p6eval rakudo 4d3a38: OUTPUTĀ«1ā¤Ā»
moritz_ who to decide which methods should generate a closure? 08:48
Matt-W Something consistent
moritz_ since Whatever ~~ Any, most interesting methods (defined in Any) are also existing methods on Whatever
masak rakudo: say Whatever.flip
p6eval rakudo 4d3a38: OUTPUTĀ«)(revetahWā¤Ā»
masak :) 08:49
moritz_ :-)
masak we should have a .flip that also flips parens.
and turns 'b' into 'd'.
moritz_ that would be .mirror :-)
masak right.
moritz_
.oO( Text::UpsideDown )
 08:50
mberends :) in the days of single sided floppy disks, once installed a splash screen on the second side that displayed a reversed message saying "Error: diskette upside down" 08:54
masak moritz_: ĀæʃuıɄŹ‡ĒÉÆos Ēq Ź‡ÉÉ„Ź‡ Ź‡,upŹƒnoM 08:55
mberends wow! 08:56
masak www.fileformat.info/convert/text/upside-down.htm
09:01 ejs2 left 09:12 kidd` joined 09:16 donaldh_ joined, donaldh_ left 09:21 alc left
pugs_svn r26676 | wollmers++ | rakudo skip 09:30
09:31 ejs2 joined
moritz_ wollmers: please prefix commit messages to the test suite with [t/spec' 09:31
[t/spec] actually
afk
10:05 ejs2 left
wollmers moritz_: ACK 10:08
10:15 Muixirt joined
jnthn H H 10:35
moritz_ LL HL TH VWLLSS LLSPK 10:36
10:38 icwiener joined
arnsholt E A S OE? 10:42
10:45 zamolxes left
masak arnsholt: well, you wouldn't be able to distinguish between "OH HAI" and "NO WAI", for example... 10:46
moritz_ masak, literal: in last year's GSOC we had weekly IRC meetings with status reports, much like #parrotsketch... what do you think about doing it like that this year? (once the coding time has started, of course) 10:48
arnsholt masak: Yeah. It's just that it's a natural extension of the no-vowels thing. I'm a sucker for language games 10:49
Which is part of the reason I like Perl as much as I do, I suspect =)
moritz_ Muixirt: my complains are not that there are no possiblities for optimization, but that the default thing (whiche 90% of all people use) sucks 10:50
masak moritz_: +1 10:51
(for the meetin idea)
s/in/ing/
pugs_svn r26677 | wollmers++ | [t/spec] fudge and unfudge for current rakudo
masak arnsholt: Perl is definitely the language for people who like languages. 10:52
Muixirt moritz_, but if parrot isn't ready (the fast cores) rakudo users have to wait 10:53
Matt-W arnsholt, masak: Definitely!
wollmers masak: true for me, mostly because regex and \p{} 10:54
Muixirt moritz_, did you watch the Lars Bak video? 10:55
moritz_ Muixirt: I know. no reason not to rant.
and "no" to the video
wollmers masak: I do NLP and NER in ~300 languages with Perl 5 10:56
masak wollmers: nice. 10:57
jnthn has got a backtrace of the slice.t crash, but it isn't making it especially obvious what's wrong... 10:59
11:02 payload joined
jnthn Ugh. And it fails in a different way under -G. 11:03
masak I'm still thinking about putting together a grant application for summer Rakudo work. I have another idea which might be considered to be on the critical path as well. 11:05
all those collection classes, Bag, HashSet, and the like.
Muixirt moritz_, what a pity :-) the was a passage in the talk that reminded me instantly about parrot/rakudo 11:06
moritz_, of course i don't know whether Lars Bak looked at parrot/rakudo, and, of course, i might got it wrong 11:07
jnthn masak: They need to get done at some point. :-)
masak yes. and they need a bit of spec work and tests as well.
jnthn Indeed.
masak I think it could be a nice summer project.
jnthn Same.
masak the advantage of doing it in one concentrated time period is that there might actually be some consistence to the design. :) 11:08
s/ce/cy/
ugh.
wollmers moritz_: Should all Unicode related tests go from some-file.t to somefile.unicode.t?
moritz_ wollmers: no 11:09
11:09 payload left
moritz_ wollmers: Unicode is integral part of Perl 6, no need to separate them 11:09
wollmers moritz_: UTF-8;-) 11:10
masak Set, Bag, KayHash, KeySet, and KeyBag. 11:12
11:12 hcchien left, hcchien joined
moritz_ masak: "container types" 11:12
masak Mr Bloch's tutorial on Java's collections really opened my eyes to the richness of collections in Java, and the need to do this well. java.sun.com/docs/books/tutorial/co...index.html 11:13
I should perhaps start by tentatively translating all the code in that tutorial.
...and blogging about it. :)
jnthn :-)
moritz_ mberends: the socket test passes here when run separately, but fails one (the third) when run during 'make spectest' 11:15
at least on one machine
11:16 spx2 left
jnthn Ouch. 11:17
rakudo: my @array = (3,7,9,11); say @array[0..1]
p6eval rakudo 4d3a38: OUTPUTĀ«37ā¤Ā»
jnthn Can somebody try running that but with Parrot's -G on? 11:18
e.g. ../../parrot perl6.pbc -e ' my @array = (3,7,9,11); say @array[0..1]'
oops
-G after parrot
Try it with and without and let me know if you see a difference in the output too.
moritz_ maximum recursion depth exceeded
current instr.: 'parrot;Int;succ' pc 5513 (src/classes/Int.pir:81)
jnthn Same here. :-(
moritz_ (with -G)
and 37 without -G 11:19
jnthn OK, reproducable at least. Hmm.
(This is from the slice.t issue)
11:20 donaldh left, donaldh joined
jnthn Not quite sure, what to make of this. 11:20
It's almost like a GC bug that makes things work. 11:21
(But in slice.t our luck runs out and we segfault during exit)
If I fudge out all the tests in slice.t that do ranges inside postcircumfix:<[ ]> then the segv goes away. 11:23
s1n masak: "In general, it is good API design practice not to make users pay for a feature they don't use." (from the introduction, always been a good quote) 11:26
masak s1n: nod. 11:32
jnthn files a ticket with this discoveries, not wanting to spend all day debugging a GC issue. 11:42
11:47 mmayr joined, mmayr is now known as desertm4x 11:53 fridim_ joined 12:03 kidd` left 12:05 ruoso joined
ruoso Hello! 12:07
masak ruoso: o/ 12:08
ruoso masak, how's everything in web.pm? 12:10
masak ruoso: there's progress. 12:11
ruoso I kinda lost track of it... do you have a dispatcher already? 12:12
masak ruoso: I'm not sure about the big picture yet. there's a lot of details to get lost in. but I like working on it.
ruoso: ihrd has been working on something called Routes. don't know how close to an actual dispatcher that is.
ruoso is it in the github repo?
masak yes. 12:13
ruoso do you have a link to that specific part of the code? 12:14
masak aye, hold on.
github.com/masak/web/blob/master/lib/Routes.pm 12:15
I don't know too much about it -- only that I recognize parts of it from the Dispatcher.pm code in November, that one also written by ihrd++, 12:16
and that Routes is said to have ties to the model in some way.
ruoso hmm... I miss the use of context variable 12:22
s
masak ruoso: I haven't seen this code in actual use yet. it's what I feel is missing right now, an actual use. 12:23
now that we have sockets and the Web.pm request/response groundwork, we should really focus on pushing things out onto the browser, so that we can see them. 12:24
ruoso masak, what happen to the idea of using a generic role for Request and Response? 12:30
s/happen/happened/
12:31 donaldh left
masak ruoso: what was that idea again? I vaguely recall us talking about it, but not why that was preferrable... :/ 12:31
ruoso because it would make the framework useable in more situations 12:32
masak that's a generic answer.
ruoso: put it like this: the current model works for Rack, why won't it for us?
ruoso it's the other way around 12:33
I'd like to use it for non-web things
or at leasy
non http
masak ruoso: ah, now I remember.
ruoso github.com/masak/web/blob/2b4a056b0...doc/stages
masak ruoso: you know, non-http is really outside of the scope of our grant. 12:34
ruoso as I said
at that time
I don't expect you to implement non-http
masak right.
ruoso I just asks for it to be possible for someone else to do it
masak yes.
ruoso i.e.: me
(as I already did with catalyst...)
masak at that time, I didn't know as much about Rack's Request and Response as I do now. 12:35
12:36 abra joined 12:37 zev left
ruoso masak, github.com/masak/web/blob/2b4a056b0...work-roles 12:38
12:38 ZuLuuuuuu joined
ruoso that was my sketch at that time, in case you don't recall 12:38
masak ruoso: I recall. 12:39
12:40 zev joined
masak ruoso: is there something in particular that I can relay to ihrd, the one tasked with implementing this part of Web.pm? 12:40
ruoso two things specially
one is the use of contextual variables
masak (I know he has health problems in the family, so he's not very available right now.) 12:41
pugs_svn r26678 | wollmers++ | [t/spec] fudge test for rakudo; softlink
ruoso the other is the different stages of the action processing
masak I will tell him when he's around. 12:43
12:44 Jedai left 12:45 [particle] joined
Matt-W masak: oh yes, I read your blog about Web.pm and I thought that using parse actions to generate a tree, and then doing top-down processing on the resulting tree is probably going to be very common. 12:56
masak Matt-W: I believe so. 12:57
dalek kudo: rd500f20 | jnthn++ | src/pmc/p6invocation.pmc:
More work on P6Invocation. There'll be room for optimization, but this serves well as at least a POC.
masak Matt-W: I also believe a good project at this point, for whoever has the time to do it, would be to pair up XPath and Match objects.
wolverian would it make sense to compile xpath to rules? probably not. 12:59
masak wolverian: it might make sense to compile them to code. 13:00
wolverian easier to wrap an xpath engine inside a Match impl, or so.
yeah, but that is a pain. 13:01
moritz_ since PGE can't do matches on anything but strings... no way
masak moritz_: you could convert the Match tree to a string :)
moritz_ sends the waves of agonizing pain that this statement caused back to masak 13:02
masak aie. :)
jnthn woo. p6invocation has come together more easily than expected. 13:05
13:05 payload joined
wolverian convert Match to XML and then run xpath on *that* 13:06
13:09 exodist joined
Matt-W replace PGE with an xpath engine 13:09
make everything be XML internally
<assign><my identifier="a" /><string value="foo" /></assign> 13:10
Matt-W runs and hides
jnthn I once saw a library that let you declare regexes in .Net using the query expression syntax. 13:11
Matt-W that sounds like a nice thing for xml people
mberends moritz_: t/spec/S32-io/IO-Socket-INET.t randomly fails tests 2 and 3 when run repeatedly. Using ' netstat --tcp --all --numeric' at the same time shows the tested socket in TIME_WAIT state. Let me try to change the socket options before keeping this in spectest.data. 13:12
jnthn A lot of folks in the .Net world don't really seem to grok regexes, and write all kinds of string manip code that a regex woulda handled for them much more easily. 13:14
Matt-W you have to give them some slack 13:16
a lot of them probably came from C++
and nobody ever introduced them to boost::regex or PCRE
jnthn Oh, quite possibly, yes.
13:17 FurnaceBoy joined
Matt-W Or on the web side, they game from VBScript 13:17
jnthn Aye.
Matt-W And .NET doesn't really encourage use of Regex, it just provides it
jnthn C# 3.0 is a good language, and the .Net stuff ain't so bad either, but the code I see by a lot of people might as well have been written in VBScript or Java. 13:18
Matt-W People don't use it like they should
jnthn It goes for any language though.
Matt-W Absolutely 13:19
jnthn It's not like everyone uses Perl like they should. :-)
Matt-W I've met a fair few people who work with .NET but have no idea how they could benefit from lambda functions
jnthn Aye.
Matt-W Whereas when I've worked in .NET I use them all the time
But that's because I'm a functional programmer whenever I get the chance
dalek kudo: ree10f59 | jnthn++ | (2 files):
Method on P6Invocation to check if we have more candidates, and tests for it.
 13:20
jnthn There's also a camp of people in the .Net world who are along the lines of, "object orientation is the one true paradigm and these extra things pollute the purity of it".
Matt-W Urgh
No no no
OO is much easier if you mix in some functional stuff
jnthn I know, but when they announced anonymous methods in C# 2.0, I was pretty horrified by the number of comments along those lines that I saw. 13:21
Matt-W Particularly if you're playing with threads or event-driven programming
ruoso jnthn, taking an old question... do we have a plan for @list.reduce being called with a multi with varying arity?
Matt-W It's just ignorance
jnthn ruoso: Not that I'm aware of...
Matt-W And I don't think anonymous delegates sit outside the OO paradigm at all
All you have to do is say 'code is an object'
masak jnthn, ruoso: what's the issue? is there a small piece of code that highlights it? 13:22
jnthn Oh, for sure.
ruoso: Part of me still wonders if that's a task for a module.
ruoso masak, the muti-based action dispatcher would require it in order to be sane
Matt-W jnthn: they'd probably also consider Smalltalk's object system to be wrong though :)
13:22 ejs2 joined
masak ruoso: I assumed that's why you asked, yes. 13:23
Matt-W: c2.com/cgi/wiki?ClosuresAndObjectsAreEquivalent
13:25 nihiliad joined 13:28 dalek left
Matt-W masak: I love those pages, where you find your opinion being modified as you read all the arguments on the way down 13:28
13:28 dalek joined
masak Matt-W: me too. 13:28
Matt-W frustrating meeting & 13:29
13:29 sri_kraih_ joined
jnthn oh noes I forgot lunch 13:29
13:30 sri_kraih left
ruoso .oO( good thing to be to the west, so I get a reminder for lunch by people who forgot lunch ) 13:30
masak jnthn: Google Imeages search for "oh noes I forgot lunch" gave this: www.ramblestrip.com/pictures/lolcat...sletgo.jpg 13:31
jnthn :-D 13:32
13:34 dalek left, dalek joined 13:38 amoc left 13:48 desertm4x left
pugs_svn r26679 | lwall++ | [STD] start contextualizing language braids for user-overridability 13:50
TimToady ruoso> one is the use of contextual variables 13:53
13:53 amoc joined
TimToady I'm curious whether this is ++ or -- 13:53
ruoso this is that he was supposed to use contextual variables, and he's not
TimToady okay, good :) 13:54
13:54 PacoLinux joined
TimToady (I hope) 13:54
since STD uses 'em all over the place, probably places where it shouldn't...
ruoso TimToady, take a look at github.com/masak/web/blob/2b4a056b0...work-roles
13:55 pmurias joined
TimToady extra semi on line 50 13:57
PerlJam TimToady: can you make perl give error messages like that? :-) 13:59
TimToady sometimes
the fact that the next statement is "if $.foo;" is indicative 14:00
sort on line 90 missing its :
wolverian can line 20 be self.?begin-closure(self)? though that looks rather weird.
TimToady the bindings in regex should be = rather than := these days 14:01
that's just an optional call, okay if it fails
wolverian and $_.priority can be .priority
14:02 donaldh joined
TimToady line 110 has --> for -> 14:02
masak fixes those issues
TimToady++
wolverian lines 155-156 have extra ;s 14:03
and are they missing ...s
?
PerlJam the extra semicolons there are harmless.
masak probably.
aye. 14:04
removed them anyway.
wolverian++ PerlJam++
TimToady and for some reason STD doesn't like the CONTROL block, but probably a STD bug 14:05
also, it didn't like the use of URI without declaraiton
wolverian can self.*foo be written .*foo? :)
TimToady not unless $_ === self
wolverian yeah, well, it is there. I mean the syntax .? syntax without the preceding noun. 14:06
masak at this point I'd like to point out that this is only a draft. :)
wolverian s:1st/syntax//
TimToady std: say .*foo
wolverian masak: but it is so much fun to bikeshed...
p6eval std 26679: OUTPUTĀ«ok 00:02 35mā¤Ā»
wolverian ah, yeah. thanks. :)
PerlJam masak: but does rakudo reasonably grok most of it?
TimToady well, it mighta found a STD bug
masak wolverian: absolutely. don't mind me. :)
PerlJam: I have no idea. I think so.
TimToady ah, no, it's confused because ControlExceptionDetach isn't declared 14:07
so it assumes it's a listop and eats the block
masak ah.
not much I can do about that, 'sides creating the real thing.
TimToady well, could perhaps give a better error message
after noting that we assumed ControlExceptionDetach was a listop 14:08
masak well, if you insist. :)
TimToady I insist, therefore I exist.
Caesar's version of it :) 14:09
masak :)
.oO( ControlExceptionDetach iacta est... )
14:10 azawawi joined
azawawi hi * 14:10
14:11 skids joined
wolverian
.oO( *.hi($azawawi) )
 14:11
14:14 PhatEddy joined
PhatEddy rakudo: package A::B {}; say ~A::B; say ~A 14:16
p6eval rakudo ee10f5: OUTPUTĀ«Bā¤invoke() not implemented in class 'NameSpace'ā¤current instr.: '_block14' pc 84 (EVAL_20:51)ā¤Ā»
PhatEddy looking at RT #62900 and will probably use above as test ... 14:17
rakudo: package A::B {}; say ~A::B
p6eval rakudo ee10f5: OUTPUTĀ«Bā¤Ā»
PhatEddy is that right ?
PerlJam It doesn't look right to me.
jnthn eated a cheezburger!
PerlJam Your first snippet looked like 2 bugs to me. 14:18
masak ...both of which are in RT already... 14:19
PhatEddy just looking at adding (or documenting tests). If I find the RT for ~A::B I will note the # ... 14:20
masak looks 14:22
PerlJam The tests aren't annotated with RT#s are they?
PhatEddy I did a search on both rt #s in t/spec 14:23
masak nope. the RT tickets are Rakudo-specific.
PerlJam I would be neat if there were a way to do that though
masak PhatEddy: alternatively, it might be that A::B actually should stringify to 'B'... 14:24
PerlJam then when jonathan or pmichaud or whoever implements something, it could say "because these tests pass, these RT#s have been fixed" or something
masak PerlJam: I do that in almost all of my Perl 6 projects. 14:25
PerlJam: I mark things up with a RAKUDO comment and an RT number.
PerlJam rakudo: package A::B {}; say A::B.perl;
p6eval rakudo ee10f5: OUTPUTĀ«Null PMC access in getprop()ā¤current instr.: 'parrot;P6object;HOW' pc 60 (runtime/parrot/library/P6object.pir:100)ā¤Ā»
PerlJam I wonder if that should work 14:26
14:26 SamB joined
masak it shouldn't throw that error, at least. 14:26
masak submits rakudobug
PerlJam right
jnthn aye
I'm not entirely sure what it should do.
But not that. Never that. :-)
PerlJam rakudo: class A::B {}; say A::B.perl 14:27
p6eval rakudo ee10f5: OUTPUTĀ«A::Bā¤Ā»
PerlJam (just checking :)
PhatEddy PerlJam: afaik we CAN add these tests to spec with fudges reading "#?rakudo todo 'RT #NNNNN'"
masak rakudo: package A {}; say A.perl
p6eval rakudo ee10f5: OUTPUTĀ«Null PMC access in getprop()ā¤current instr.: 'parrot;P6object;HOW' pc 60 (runtime/parrot/library/P6object.pir:100)ā¤Ā»
PerlJam rakudo: class A::B {}; say A::B.WHAT 14:28
p6eval rakudo ee10f5: OUTPUTĀ«A::B()ā¤Ā»
PerlJam I guess it doesn't matter if the class names are just Foo or if they are Foo::Bar::Baz
14:29 iblechbot left
ruoso jnthn, is there a way to implement a control exception in rakudo already? 14:31
jnthn ruoso: Not sure, Tene would know. I really haven't been following the exceptions work very closely. 14:32
ruoso Tene?
jnthn I haven't seen it done yet, though.
ruoso are you there? 14:33
jnthn I think he said he was quite busy this week...try leaving a message.
ruoso I'll try to get that sketch code I posted above to work in rakudo
14:35 goksie joined 14:36 PhatEddy left 14:37 [particle]1 joined 14:41 payload left
ruoso rakudo: role A::B { } 14:45
p6eval rakudo ee10f5: ( no output )
Tene ruoso: what do you want to know about exceptions?
ruoso Tene, how can I declare a custom control exception in rakudo? 14:46
14:46 mizioumt joined
Tene You can't. 14:47
pugs_svn r26680 | lwall++ | [STD] improve error message on accidental listops
14:47 abra left
Tene if you can show me what that should look like in rakudo, I can implement it for you, but it might take me a few days. 14:48
ruoso Tene, well... I guess it's just a subclass of ControlException 14:49
that you can call .throw on
Tene and CONTROL blocks catch all subclasses of ControlException? 14:50
ruoso the way you say it seems to imply that they would also be seen by CATCH blocks 14:51
but it's more like
.throw on a control exception will look for a CONTROL block
just as
14:51 frodwith joined
ruoso .throw in a failure will look for a CATCH block 14:52
Tene can you put this stuff in a spec test?
14:52 azawawi left
ruoso right... I just won't be able to do it now... I notice you when I have it there... 14:53
Tene thank you
as jnthn said, I'm really busy with work this week, but I should have some time available.
TimToady rakudo: my %h; %h.push: $*IN.slurp.join.comb.[^3] 14:54
p6eval rakudo ee10f5: OUTPUTĀ«Trailing item in Hash.pushā¤Ā»
Tene also, has "Is ControlException a role or a class" been debated already?
TimToady rakudo: say $*IN.slurp.join.comb.[^3] 14:55
ruoso Tene, every built-in type is a role
p6eval rakudo ee10f5: OUTPUTĀ«LandderBerge,ā¤Ā»
TimToady rakudo: say $*IN.slurp.comb.[^3]
p6eval rakudo ee10f5: OUTPUTĀ«LandderBerge,ā¤Ā»
TimToady rakudo: say $*IN.comb.[^3]
Tene OK
p6eval rakudo ee10f5: OUTPUTĀ«IO()<0xb5fe8808>Use of uninitialized valueā¤Use of uninitialized valueā¤ā¤Ā»
TimToady aw
jnthn Anybody know where in the specs Selector is defined? 14:56
jnthn can't find it...
ruoso pmichaud, github.com/masak/web/blob/2b4a056b0...work-roles in line 102 is failing to parse... any clue
?
jnthn, if you happen to have any clue of the above, I'd apreciate
jnthn ruoso: what error? 14:57
TimToady jnthn: perhaps it's really Matcher
ruoso Statement not terminated properly at line 27, near "$<_parent_"
jnthn ruoso: oh, because Perl 6closures in regexes aren't implemented yet.
TimToady: I'll look for that. It's mentioned in the spec for WALK. 14:58
:omit(Selector) # only classes that don't match selector
ruoso jnthn, how do I workaround it?
TimToady S29 sez: subset Matcher of Object where { $_.can('ACCEPTS') };
ruoso jnthn, considering the specific case, which is what I need to do 14:59
TimToady masak: anyway, you don't need .slurp.join
and shouldn't even need the .slurp since .comb is supposed to work on filehandles
ruoso lunch... brb & 15:00
jnthn ruoso: Actually looking at the error, it may be that we don't ahve := either...
TimToady ruoso: I already said above that regex binding doesn't use := 15:02
15:03 payload joined
TimToady std: / $x := (.) / 15:04
p6eval std 26680: OUTPUTĀ«##### PARSE FAILED #####ā¤Unrecognized regex metacharacter at /tmp/YtuKJvUOTD line 1:ā¤------> / $x := (.) /ā¤ expecting any of:ā¤ quantifierā¤ regex atomā¤ rxinfixā¤ wsā¤Other potential difficulties:ā¤ Variable $x is not predeclared at
../tmp/YtuKJvUOTD line 1:ā¤------> [...
TimToady shower & 15:05
masak TimToady: .slurp.join ... I should take up coffee, it seems. :) 15:07
TimToady: re .comb and filehandles: cool! didn't know that.
TimToady: I don't believe .comb is currently specced to work on filehandles, though. 15:08
dalek kudo: 8009acf | jnthn++ | src/parser/actions.pm:
Add lexical slot for dispatcher to store candidate list holder in methods.
 15:10
15:12 riffraff joined 15:14 aaronpk joined 15:15 goksie left
[particle]1 candidate list holder? 15:16
15:16 [particle]1 is now known as [particle]-
aaronpk moritz_: is it possible to modify your irclog bot to also log usernames? 15:16
[particle]- aaronpk: what's not logged? irclog.perlgeek.de/perl6/today 15:17
jnthn [particle]-: Think callnext, etc.
skids rakudo: @a = $*IN.lines; @a[0..2].say;
[particle]- jnthn: so, list of lists?
p6eval rakudo ee10f5: OUTPUTĀ«Scope not found for PAST::Var '@a' in ā¤current instr.: 'parrot;PCT;HLLCompiler;panic' pc 146 (src/PCT/HLLCompiler.pir:105)ā¤Ā»
skids rakudo: my @a = $*IN.lines; @a[0..2].say;
p6eval rakudo ee10f5: OUTPUTĀ«Land der Berge, Land am Strome,Land der Ƅcker, Land der Dome,Land der HƤmmer, zukunftsreich!ā¤Ā» 15:18
aaronpk [particle]-: the username associated with the nick. on my channel, usernames correspond to the person's real name, while their nick can change.
jnthn [particle]-: No, a P6Invocation...I think. :-)
[particle]- i need to review a few of jnthn++'s commits today 15:19
jnthn [particle]-: I'm not really expecting to get this lot right on the first attempt. Done some planning of course, but taking more the "baby steps" approach. :-)
15:20 donaldh left 15:21 donaldh joined
[particle]- so, is this P6Invocation a cache that gets invalidated if a method is injected into the class/object after the initial invocation? 15:22
jnthn [particle]-: No, it's more iterator-ish.
It'll need to end up doing Iterator-ish stuff. 15:24
[particle]- so if a multi is added/deleted during dispatch, it'll DTRT?
jnthn It's not just about multi.
[particle]- yes, that's an example.
jnthn And it's generated per invocation (for now...)
No, it's not about caching at all.
It's also missing most of what it'll eventually need. 15:25
Baiscally at the moment we do something like...
$P0 = find_method obj, 'name'
[particle]- will it work like for @a { @a.pop } ?
jnthn $P0(...)
[particle]- ...i'm thinking about edge cases again... 15:26
jnthn Apart from callmethodcc encapsulates it a bit...but it boils down to a find_method and then an invoke vtable call.
[particle]- ok
that makes sense.
jnthn Anyway, find_method instead of handing back just the method will wrap it up in one of these, and give it enough stuff that it can (lazily usually) produce the rest of the candidates if needed.
15:27 abra joined
jnthn It will also on invoke fudge itself into the lexpad of the method it invokes. 15:27
That way, callnext anf friends can find this thing and use it to obtain the Next Thing to invoke.
At the moment it's not that far along.
My plan for now is to just handle the $foo.*@candidates() case. 15:28
Where we already have a list of candidates.
[particle]- yeah, i get the idea. seems it's pretty clear in your head, at least at this level
jnthn Well, still various bits to figure out, and the current way isn't as performant as I'd like to end up with. 15:29
But so far, things have gone pretty smoothly. 15:30
Fudging the thing into the lexpad Just Worked at least...got a short PIR test script for P6Invocation checked in.
Anyway, now I'm diverting attention to the meta-class stuff. 15:31
TimToady masak: ./S32-setting-library/Str.pod:You may also comb lists and filehandles. C<+$*IN.comb> counts the words on
jnthn Want to get a bunch of things added into there, or a first cut.
Then try and write .WALK entirely in terms of the meta-class methods.
masak TimToady: oh, ok. 15:32
TimToady rakudo: $*IN.lines(3).say
p6eval rakudo 8009ac: OUTPUTĀ«Land der Berge, Land am Strome,Land der Ƅcker, Land der Dome,Land der HƤmmer, zukunftsreich!ā¤Ā»
masak TimToady: mind if I mention it in IO.pod too?
TimToady skids: ^
not at all
masak mentions
jnthn Plus the tedious work of going through Rakudo and finding everywhere we cheated and didn't use the metaclass and really should be...
jnthn digs back in
skids rakudo: regex ldb12l { ^^ Land .* $$ <.ws> ^^ Land .* $$ } ; join("\n", $*IN.lines(3)).match(/<ldb12l>/).say; 15:33
p6eval rakudo 8009ac: OUTPUTĀ«Land der Berge, Land am Strome,ā¤Land der Ƅcker, Land der Dome,ā¤Land der HƤmmer, zukunftsreich!ā¤Ā»
[particle]- convert to metaclass access first, then rip out old code?
jnthn Well, more replace the code that doesn't use the metaclass to use the metaclass. 15:34
skids rakudo: rule ldb2 { ^^ Land .* $$ }; regex ldb12 { ^^ Land .* $$ <ldb2> }; join("\n", $*IN.lines(3)).match(/<ldb12>/).say;
p6eval rakudo 8009ac: OUTPUTĀ«Land der Berge, Land am Strome,ā¤Land der Ƅcker, Land der Dome,ā¤Land der HƤmmer, zukunftsreich!ā¤Ā»
jnthn At some point I think we'll need to give Rakudo its own metaclass distinct from the one used by the toolchain stuff. 15:35
[particle]- but not yet.
TimToady skids: . match everything, did you mean \N? 15:36
or are you testing backtracking?
jnthn Not yet, but probably not too far into the future.
skids I'm trying to figure out why embedding a rule in a regex is doing non-dwimmy things with $$ and ^^. 15:37
rakudo: rule ldb2 { ^^ Land .* $$ }; regex ldb12 { ^^ Land .* $$ <ldb2> }; join("\n", $*IN.lines(2)).match(/<ldb12>/).say;
jnthn Anyway, this should get us towards having our representation vs metaclass story somewhat more sorted out.
p6eval rakudo 8009ac: OUTPUTĀ«Land der Berge, Land am Strome,ā¤Land der Ƅcker, Land der Dome,ā¤Ā»
TimToady what do you "mean" by .?
skids Bizarre, I must be doing something different on my script here.
TimToady .* is going to match multiple lines 15:38
PerlJam skids: perhaps you want \N instead of . ?
skids Erm, . == end of English sentence
:-)
PerlJam skids: or maybe you want to make * less greedy ?
TimToady are you trying to match one line? 15:39
skids No, it's actually doing what I want, just not on my own install.
15:57 ilogger2 joined 15:58 SamB joined
jnthn rakudo: sub foo() { }; say &foo.name 15:58
p6eval rakudo 8009ac: OUTPUTĀ«Method 'name' not found for invocant of class 'Perl6Sub'ā¤current instr.: 'parrot;P6metaclass;dispatch' pc 243661 (src/gen_actions.pir:0)ā¤Ā»
moritz_ aaronpk: what do you mean? user names appear in the log
aaronpk moritz_: i only see a "nick" column, no "username" column 16:00
moritz_ aaronpk: what would you want to appear in the username?
16:00 cognominal joined
moritz_ the ircname? 16:00
aaronpk when i join an irc server i enter a "nick" and a "username". in my channel, usernames correspond to people's real names, while the nick can change. i want to log their real names too 16:01
moritz_ aaronpk: that's not available in the channels I log, so I won't implent it. Of course you're free to do it yourself 16:02
afk
aaronpk moritz_: in this channel your username is set to "i=moritz". that's the field i'm trying to get to 16:03
16:03 hercynium joined
skids rakudo: rule ldb2 { ^^ Land \N* $$ }; regex ldb12 { ^^ Land \N* $$ <ldb2> }; join("\n", $*IN.lines(3)).match(/<ldb12>/).say; 16:03
p6eval rakudo 8009ac: OUTPUTĀ«Land der Ƅcker, Land der Dome,ā¤Land der HƤmmer, zukunftsreich!ā¤Ā»
Matt-W Woohoo, another working day over 16:04
skids OK, why does that match the second two lines and not the first two, tuens out to be my problem.
16:04 eMaX joined
pmichaud ruoso: (line 102) PGE might not like the leading underscore in the rule name. I'm also pretty sure it doesn't know what to do with <$act.regex> 16:08
good morning, #perl6 16:10
Tene good morning, pmichaud
16:10 justatheory joined
masak good time of day, pmichaud. 16:11
jnthn morning, pmichaud
skids rakudo: rule l { ^^ L $$ }; "\n\nL".match(/<l>/).say; 16:12
p6eval rakudo 8009ac: OUTPUTĀ«ā¤ā¤Lā¤Ā»
skids rakudo: rule l { ^^ L $$ }; "\nL\n".match(/<l>/).say;
p6eval rakudo 8009ac: OUTPUTĀ«ā¤Ā»
skids rakudo: rule l { ^^ L $$ }; "L\n\n".match(/<l>/).say;
p6eval rakudo 8009ac: OUTPUTĀ«ā¤Ā»
jnthn pmichaud: It turns out that I can use multis to make the old and current forms of the metaclass methods co-exist. 16:14
pmichaud jnthn: excellent.
jnthn pmichaud: Added the other variants locally just throwing exceptions, then ran through all of the p6object, nqp, pge and rakudo tests to make sure we never call the wrong one. All passing.
ruoso pmichaud, but it was supposed to accept the leading underscore, right?
pmichaud ruoso: yes, it should.
ruoso righ 16:15
pmichaud rakudo: regex _abc { foo }; "xyz foo bar".match(/<_abc>/).say;
p6eval rakudo 8009ac: OUTPUTĀ«fooā¤Ā»
pmichaud hmm, looks like the leading underscore isn't the problem.
ruoso yeah... it still fails here... 16:16
pmichaud Okay, I think it's the <$act.regex> that it's having difficulty.
ruoso let me get it into a variable first
pmichaud probably won't help -- PGE doesn't know what to do with (Perl 6) variables yet.
ruoso hmm... 16:17
how do I work-around it?
pmichaud maybe put it into a sub variable
my &regex = $act.regex;
.... / <regex> /
that might work. 16:18
ruoso cool
gone to another problem now
how do I work around the { make $act } 16:19
?
16:19 goksie joined, icwiener_ joined
pmichaud that one's tougher. 16:19
I'm not sure how to grab $act reliably. 16:20
ruoso maybe using a closure...
and invoking it with <foo>
how should a closure be written to replace { make $act }? 16:22
pmichaud I'm not sure.
ruoso does it receive the match object as argument? 16:23
pmichaud {{ ... }} will get a match object in the 'match' register... but the internals are PIR
one could do
{{ match.'!make'( ... ) }}
but I'm not sure how to get $act
ruoso right... 16:24
in the previous suggestion
I created a perl 6 level closure that is invoked as a rule
does this rule receive the outer match?
pmichaud yes, the current match object is the invocant.
(I'm not sure that's correct Perl 6, but that's what PGE does.)
ruoso so, in theory, I could call the same method in Perl 6 level, couldn't i? 16:25
like...
moritz_ + my $m = $Ā¢.explain_mystery();
TimToady++
ruoso my &cl = -> $match { $match."!make"( $act ) };
pmichaud sub make_act( $/ ) { make $act } # might work
ruoso just calling make ?
pmichaud make always works on its caller's lexical $/ 16:26
which we set to the invocant parameter here.
ruoso cool
it parses, at least ;)
pmichaud, hmm... it doesn't seem to support <@subregexes>
pmichaud anything dealing with Perl 6 level variables we don't support yet. 16:27
PerlJam rakudo: %*h.push: 1 => 2; %*h.perl.say
p6eval rakudo 8009ac: OUTPUTĀ«{}ā¤Ā»
ruoso right... right...
pmichaud Keep in mind that Parrot has only had working lexicals in the last 6 months. PGE hasn't had a chance to catch up on that yet.
PerlJam rakudo: %*h = (); %*h.push: 1 => 2; %*h.perl.say
p6eval rakudo 8009ac: OUTPUTĀ«{"1" => 2}ā¤Ā»
ruoso pmichaud, right... I assume I could as well write a closure to emulate <@subregexes>... 16:28
pmichaud ruoso: likely.
ruoso should it return a new match? 16:29
pmichaud yes.
ruoso or false, if none match
pmichaud it returns a false match if no match
ruoso is it?
pmichaud it always returns a match
ruoso how do i get a false match?
in Perl 6
pmichaud in PGE, set the match's .to value to be less than its .from
that indicates a boolean false 16:30
ruoso less or equal? or just less
pmichaud less
equal means a zero-length match.
(but successful)
ruoso ok
PerlJam rakudo: %*h = 1; %*h.push: 1 => 2; %*h.perl.say 16:31
p6eval rakudo 8009ac: OUTPUTĀ«Odd number of elements found where hash expectedā¤current instr.: 'parrot;Perl6Hash;!STORE' pc 9254 (src/classes/Hash.pir:237)ā¤Ā»
PerlJam rakudo: %*h.push: 1 => 2; %*h.perl.say
p6eval rakudo 8009ac: OUTPUTĀ«{}ā¤Ā»
pmichaud shouldn't %*h be giving errors as undeclared on all of these?
Infinoid pmichaud: (PGE catching up with parrot lexicals) is that a pmichaud-only task, or is it something you can use some help with? 16:32
pmichaud s/shouldn't/should/
PerlJam dunno
Infinoid (I've been meaning to learn some more about that side of parrot.)
PerlJam (I would guess yes, but I haven't read the spec on globals recently)
pmichaud Infinoid: since PGE's parser is about to be significantly rewritten to match STD's version of regexes, I'm not sure I'd inflict it on anyone else.
PerlJam rakudo: eval "%*h = 1"; %*h.perl.say; %*h.push: 1=>2; %*h.perl.say; 16:33
Infinoid Oh, that does sound daunting. Oh well, I'm still a willing victim for lesser tasks
p6eval rakudo 8009ac: OUTPUTĀ«{}ā¤{"1" => 2}ā¤Ā»
TimToady in STD a false match is a return of (), since the model is using lazy lists to represent possible paths forward
but we're not returning match objects, rather cursors
lists of cursors
ruoso rakudo: "a" ~~ /a/; my $clone = $/.clone; $clone.to = $clone.from - 1; say $clone.perl;
p6eval rakudo 8009ac: OUTPUTĀ«Match.new(ā¤ # WARNING: this is not working perl codeā¤ # and for debugging purposes onlyā¤ ast => "",ā¤ Str => "",ā¤ from => 0,ā¤ to => -1,ā¤)ā¤Ā» 16:34
pmichaud ruoso: right now I think that creating a new Match object (via Match.new) returns one that is "false"
TimToady: a regex should generally return Match objects, though, yes ?
ruoso rakudo: my $m = Match.new; say $m.perl;
p6eval rakudo 8009ac: OUTPUTĀ«too few arguments passed (1) - 3 params expectedā¤current instr.: 'parrot;PGE;Match;new' pc 28 (compilers/pge/PGE/Match.pir:57)ā¤Ā»
pmichaud arggh. 16:35
moritz_ rakudo: my $m = Match.new(0, 1); say $m.erl
pmichaud ruoso: okay, false lead I guess. Sorry about that.
p6eval rakudo 8009ac: OUTPUTĀ«positional inside named args at position 2ā¤current instr.: 'parrot;PGE;Match;new' pc 28 (compilers/pge/PGE/Match.pir:57)ā¤Ā»
moritz_ rakudo: my $m = Match.new(0, 1); say $m.perl
p6eval rakudo 8009ac: OUTPUTĀ«positional inside named args at position 2ā¤current instr.: 'parrot;PGE;Match;new' pc 28 (compilers/pge/PGE/Match.pir:57)ā¤Ā»
ruoso pmichaud, is it ok to clone the input match?
moritz_ rakudo: my $m = Match.new(:from(0), :to(1)); say $m.perl
pmichaud ruoso: yes. You might even ben able to do $/.new
p6eval rakudo 8009ac: OUTPUTĀ«invalid arg type in named portion of argsā¤current instr.: 'parrot;PGE;Match;new' pc 28 (compilers/pge/PGE/Match.pir:57)ā¤Ā»
TimToady pmichaud: a regex should return something that *does* Match; in STD it also happens to be a Cursor
and an instance of the current language 16:36
pmichaud TimToady: okay, got it.
16:36 azawawi joined
TimToady all the .MATCHIFY methods make sure the current Cursor looks like a valid Match object 16:37
moritz_ wollmers: r26676 confuses me - why did you fudge that test? was it failing?
TimToady and also happen to call the implicit reduction action, which is why we don't need {*} 16:38
ruoso pmichaud, hmm... how do I invoke another regex from inside the closure? should I just use the sub form?
pmichaud ruoso: yes, that should work.
jnthn pmichaud: If we're going to add a parrotopaque kinda PMC (along the lines of p6opaque without without the Perl 6 specific bits...will add thsoe in a subclass), should it be a dynpmc? Is this something we need to ask about in parrotsketch? 16:39
ruoso okay... it compiles now...
pasteling "ruoso" at 189.97.249.112 pasted "work-around over <@subregexes>" (17 lines, 422B) at sial.org/pbot/36399 16:40
pmurias ruoso: out of curiosity what are you using Perl 6 rules for?
pmichaud jnthn: I still haven't figured out the criteria for what should be a dynpmc and what shouldn't.
ruoso pmurias, I'm working on getting the regex based dispatch for the web framework 16:41
pmichaud, does the above paste look sane?
TimToady ruoso: in the STD model, you always invoke as a method using some Cursor or other
jnthn pmichaud: OK, I'll try and remember to ask in #ps.
ruoso TimToady, I'm not sure what that means...
pmichaud jnthn: allison has been mentioning that things like CodeString ought to become dynpmcs 16:42
TimToady there's no such thing as an implicit "current match state"
it's all explicitly represented with Cursor objects, which you can think of as a kind of continuation
ruoso right...
jnthn pmichaud: I don't mind if it's a dynpmc or core myself, just would rather know and do it right first time.
ruoso TimToady, but how does that look like in code?
TimToady so {{ return () }} means there are no more possibilities to try
pmichaud what's the advantage to having it as a dynpmc?
TimToady ruoso: see STD.pm :) 16:43
it's all over the place
16:43 lambdabot joined
pmichaud TimToady: how closely do you feel that PGE needs to follow that particular model ? 16:43
jnthn pmichaud: Only needs to be in memory if it's being used by something, so doesn't bulk up the Parrot executable I guess.
TimToady methods are called self.meth or $Ā¢.meth
ruoso TimToady, right... but I don't have a grammar here
pmichaud jnthn: let's put it this way. Anything using PGE or PCT will end up using P6object. 16:44
jnthn pmichaud: However, as it's going to be used by everything that uses the compiler tools...otoh, so is CodeString...
TimToady where self is the state upon entering the regex, and $Ā¢ is the current match state passed into the closure
ruoso I just have a bunch of anon regexes
pmichaud jnthn: right, which is why I didn't understand why CodeString ought to become a dynpmc.
jnthn pmichaud: Me either, since PCT is the likely route for most people building compilers.
pmichaud I mean, if it's _always_ used, then having it as a dynpmc would seem to actually slow things down and increase resource usage. 16:45
jnthn Right.
ruoso does rakudo support $Ā¢?
TimToady ruoso: those are called by assuming the current namespace is a class, and token is a method
pmichaud ruoso: not yet.
TimToady even if the current namespace isn't really a class
moritz_ that's easy in Perl 5 ;-)
jnthn pmichaud: OK, let's try get a general answer form allison in #ps since it would seem whatever applies to CodeString is the rule to follow for parrotopaque. 16:46
pmichaud I still haven't gotten my brain completely around the 'what is a Cursor and what does it do?' yet.
jnthn: sounds good.
ruoso so I presume the paste above is the sanest way to do it in current parrot/rakudo
pmichaud yes, keeping in mind that the @subregexes isn't going to handle backtracking into the regex the way it's written there. 16:47
ruoso you mean the closure wouldn't backtrack... 16:48
pmichaud right
so once it finds a match, you're stuck with that one.
ruoso ok... I think I don't need it to backtrack at this point... 16:49
and since it's a work-around... I think it's enough as it is now...
pmichaud (in case my earlier message got overlooked in the flood) 16:50
16:50 lambdabot left
pmichaud TimToady: how closely do you feel that PGE ought to be following STD's model of Cursor/Match interaction? 16:50
16:53 lambdabot joined
TimToady closely enough that the closures in STD run :) 16:58
but the basic distinction is that Cursors are immutable, and backtracking consists only of throwing away Cursors, not modifying anything to its privious state 16:59
17:00 hercynium left
TimToady having mutable match states is just asking for reentrancy problems 17:00
see P5
that being said, one of the optimizations that STD does is re-using cursors that can't backtrack due to ratcheting. but that's cheating 17:02
17:03 barney joined
pmichaud yes, I'm not quite sure how this all maps into PGE's current model, where backtracking states are actually managed by the program counter 17:03
and backtracking instantly jumps back to the point of the backtrack
[particle]- timtoady: does that break with :keepall? 17:04
moritz_ I think that's orthogonal
pmichaud I don't mind chucking PGE's current model in favor of something else... but I need to understand the something else a bit clearer.
moritz_ :keepall is about keeping matches, backtracking is about finding them
[particle]- i'm not clear on the semantics of :keepall, i should reread before asking. wondered if it kept the state of the backtracking, too. how 'all' is 'all'? 17:06
17:06 pmurias_ joined
moritz_ :keepall forces non-capturing subrules to capture 17:07
TimToady I should probably split out the methods in Cursor.pmc a little better, but basically all the methods starting at _MATCHIFY are the regex engine
moritz_ and it also augments matched literals with match positions etc
but it doesn't provide backtracking information
pmichaud TimToady: okay, that's a very useful guidepost. 17:08
TimToady everything down to sub bless
after that is longest token extractor
ruoso introduces "Faz" - the action dispatching framework... github.com/ruoso/faz/tree/master
[particle]- simple comments in the code could help others 17:09
TimToady 'couse, none of it makes much sense without seeing how it's called from STD.pmc
ruoso unfortunally it dies with Null PMC at this point ;)
moritz_ STD.pm is version 5.0.0.alpha? ;-)
pugs_svn r26681 | lwall++ | [STD] correct version, moritz++ 17:10
TimToady and that STD.pmc code has two basic forms, depending on ratcheting 17:11
the ratchet code is assiduously avoiding calls to lazymap(), since that's slow to emulate in p5
jnthn I'm trying to get the varous orderings that WALK suppots into my head. 17:12
Can somebody review gist.github.com/107058 ?
TimToady my nefarious plan is to retarget viv at some point to putting out rakudo code
it would be nice if lazy lists were implemented by then :)
moritz_ TimToady: thanks for your p6l mail 17:13
pmichaud TimToady: they're on their way, actually.
pugs_svn r26682 | azawawi++ | [S:H:P6] Added 'tmp_prefix' to control where STD stores its temporary files.
17:13 pmurias_ left
ruoso pmichaud, I'm getting a null pmc and I'm not sure where it is coming from... the code is the one at github.com/ruoso/faz/tree/master 17:13
jnthn In what I just pasted breadth and ascendent end up looking the same. Can somebody explain the difference to me?
TimToady I use roles heavily too, which would be much more difficult in p5 without moose
jnthn (or show me an example where it differs) 17:14
pmichaud ruoso: in general I refer Null PMC accesses in class-related stuff to jnthn++ :-)
jnthn ruoso: in general I refer Null PMC accesses in regex related stuff to pmichaud++
;-)
[particle]- ascendent isn't F E B A C D?
jnthn [particle]-: Maybe but that'd make it the same as pre-order. ;-)
17:15 eMaX left
[particle]- ah, hrmm. 17:15
TimToady moritz_: actually sent that mail earlier, but my mailq was clogged
PerlJam jnthn: you just don't have enough node to elucidate the difference adequately :)
jnthn PerlJam: You got a better example?
moritz_ TimToady: np, from your commits I figured that while the syntax was a bit wrong, it couldn't be all that bad ;-)
17:16 ejs joined
jnthn PerlJam: I can't think of one right off that makes it clearer... 17:16
PerlJam me either.
jnthn PerlJam: Or does it more show up when a class appears multiple times in an inheritance hierarchy?
Ah, OK.
[particle]- what i you hook the A over the C? 17:17
TimToady still working on the tests that try to modify regex syntax though. need to work out syntax for deriving new sublanguages inline
merely declaring a token doesn't make it a token for the language you're using for the rest of the block
pugs_svn r26683 | azawawi++ | [Cursor] Added 'tmp_prefix' to control where STD stores its temporary files. 17:18
TimToady and have to be able to derive new Q or Regex language while not in those languages too
moritz_ hrm
TimToady what we have is five or six different langauges braided together
and normal Perl has to be able to redefine any of the braids for the lexical scope 17:19
ruoso jnthn, btw... the null pmc issue was because the role was indirectly loaded, so the parser recognized the name, but the name wasn't installed in the current namespace
jnthn [particle]-: Not much different from what I can see
TimToady something like language Q { token backslash:<Y> {...} }
17:19 icwiener-_- joined
jnthn ruoso: Ah, are you using nested namespaces? 17:19
ruoso yes... 17:20
Faz::Action::Chained
jnthn Those are quirky at the moment.
ruoso but it solves it if I use explicitly
jnthn OK, but you wrote a role inside one of them. That's probably what it didn't like.
Matt-W yeah I have to do that
moritz_ TimToady: but if it's available in regexes and not double quoted strings... should it still go into language Q?
jnthn OK.
TimToady and the current rule_backslash category is completely bogus
no, then it'd go into Regex
ruoso jnthn, it still don't support yada methods in roles, right?
TimToady what's going on from the user's point of view is that $?Regex or some such is redefined for the rest of the scope 17:21
jnthn ruoso: no
TimToady from the view of the compiler it's a dynamic variable $*Regex
[particle]- jnthn: i'd wander over to #moose and ask stevan or nothingmuch 17:22
TimToady that's what my previous change to STD did, make dynamic variables for the braids
jnthn [particle]-: Yeah, good plan. Will do so after dinner.
TimToady (already had one for $*LANG, but just generalized the braiding notion)
(a lot of these context vars probably want to be tied to the Cursor eventually instead; I'm using dynamic scoping at the moment for speed under P5) 17:23
ruoso rakudo: class A { method a { say 'a' } }; class B { has A $.other is rw = A.new handles a }; say B.new.a; 17:25
p6eval rakudo 8009ac: OUTPUTĀ«Statement not terminated properly at line 1, near "handles a "ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
TimToady basically, as soon as the user starts redefining the current language, all the affected languages in the braid become anonymous grammars
ruoso rakudo: class A { method a { say 'a' } }; class B { has A $.other is rw handles <a> = A.new }; say B.new.a; 17:26
p6eval rakudo 8009ac: OUTPUTĀ«Unable to set lvalue on PAST::Val nodeā¤current instr.: 'parrot;PAST;Val;lvalue' pc 623 (src/PAST/Node.pir:163)ā¤Ā»
TimToady I already do that for infix:<foo> and such, but needs to extend to explicit derivation with a mixed in new grammar (including by import)
ruoso how do I define an attribute that is rw, that handles some methods and has a default value? 17:27
TimToady well, enuff branedump
pmichaud TimToady++ # thanks for branedump, very helpful
moritz_ better braindump than brain dumb
jnthn ruoso: Hmm. That doesn't look wrong...
rakudo: class A { method a { say 'a' } }; class B { has A $.other handles <a> = A.new }; say B.new.a; 17:28
p6eval rakudo 8009ac: OUTPUTĀ«Unable to set lvalue on PAST::Val nodeā¤current instr.: 'parrot;PAST;Val;lvalue' pc 623 (src/PAST/Node.pir:163)ā¤Ā»
[particle]- is it parsing <a> = A.new ?
17:28 ejs left
jnthn That's be surprising... 17:28
ruoso std: class A { method a { say 'a' } }; class B { has A $.other is rw handles <a> = A.new }; say B.new.a; 17:29
jnthn rakudo: class A { method a { say 'a' } }; class B { has A $.other handles <a> }; say B.new(a => A.new).a;
p6eval std 26683: OUTPUTĀ«ok 00:03 38mā¤Ā»
rakudo 8009ac: OUTPUTĀ«aā¤1ā¤Ā»
17:29 estrabd joined
pmichaud jnthn: 17:29
rule trait_verb { [ | $<sym>=[of|returns] <typename> | $<sym>=[handles] <EXPR>
note <EXPR>
jnthn That'd be it... 17:30
pmichaud so that ends up grabbing <a> = A.new
jnthn checks STD
[particle]- there you go.
jnthn rule trait_verb:handles {<sym> <noun> }
That's the one.
jnthn will fix later tonight 17:31
pugs_svn r26684 | azawawi++ | [S:H:P6] version 0.51 includes tmp_prefix and latest STD 17:32
jnthn OK, dinner time, before #parrotsketch
17:35 icwiener_ left
TimToady pmichaud: one other note about Cursor.pmc, the quantifying methods are tagged with r, g, f for ratchet, greedy, frugal 17:38
and _rev is for reverse matching for afterness
17:42 LylePerl joined 17:43 szabgab joined
ruoso rakudo: class A { has $!rx; method do { $!rx = / some regex / }; }; A.new.do; #pmichaud 17:48
p6eval rakudo 8009ac: OUTPUTĀ«Null PMC access in isa_pmc()ā¤current instr.: 'parrot;PGE;Match;new' pc 33 (compilers/pge/PGE/Match.pir:65)ā¤Ā»
ruoso rakudo: class A { has $!rx; method do { $!rx = / some regex /; 1}; }; A.new.do; #pmichaud
p6eval rakudo 8009ac: ( no output )
ruoso if the regex fall out as the implicit return, it causes the null pmc
rakudo: class A { has $!rx; method do { $!rx }; }; A.new.do; #pmichaud 17:49
p6eval rakudo 8009ac: ( no output )
ruoso rakudo: class A { has $!rx; method do { $!rx = 1}; }; A.new.do; #pmichaud
p6eval rakudo 8009ac: ( no output )
ruoso it's regex related 17:50
pmichaud rakudo: class A { has $!rx; method do { $!rx = { say 1; } }; }; A.new.do;
p6eval rakudo 8009ac: ( no output )
ruoso submitted rakudobug
pmichaud rakudo: class A { has $!rx; method do { $!rx = regex { some regex} }; }; A.new.do; 17:51
p6eval rakudo 8009ac: OUTPUTĀ«Null PMC access in isa_pmc()ā¤current instr.: 'parrot;PGE;Match;new' pc 33 (compilers/pge/PGE/Match.pir:65)ā¤Ā»
pmichaud It's trying to do a match.
ruoso why?
pmichaud something is causing the regex to be evaluated in boolean context.
ruoso ah
sure... at some point later
maybe
pmichaud I mean as part of returning the regex.
rakudo: class A { has $!rx; method do { $_ = 'foo'; $!rx = regex { some regex} }; }; A.new.do; 17:52
p6eval rakudo 8009ac: OUTPUTĀ«Null PMC access in isa_pmc()ā¤current instr.: 'parrot;PGE;Match;new' pc 33 (compilers/pge/PGE/Match.pir:65)ā¤Ā»
pmichaud hmm.
[particle]- rakudo: sub a { my $x = / some regex; }; 1; 17:53
p6eval rakudo 8009ac: ( no output )
[particle]- rakudo: sub a { my $x = / some regex; }; a();
p6eval rakudo 8009ac: OUTPUTĀ«Null PMC access in isa_pmc()ā¤current instr.: 'parrot;PGE;Match;new' pc 33 (compilers/pge/PGE/Match.pir:65)ā¤Ā» 17:54
[particle]- rakudo: sub a { my $x = / some regex /; }; a();
p6eval rakudo 8009ac: OUTPUTĀ«Null PMC access in isa_pmc()ā¤current instr.: 'parrot;PGE;Match;new' pc 33 (compilers/pge/PGE/Match.pir:65)ā¤Ā»
[particle]- interesting that the first one parsed.
pmichaud yes, looks like a bug in the regex parser.
ruoso rakudo: my $x = /some regex/; $x
p6eval rakudo 8009ac: OUTPUTĀ«Null PMC access in isa_pmc()ā¤current instr.: 'parrot;PGE;Match;new' pc 33 (compilers/pge/PGE/Match.pir:65)ā¤Ā»
ruoso rakudo: my $x = /some regex/;
p6eval rakudo 8009ac: OUTPUTĀ«Null PMC access in isa_pmc()ā¤current instr.: 'parrot;PGE;Match;new' pc 33 (compilers/pge/PGE/Match.pir:65)ā¤Ā» 17:55
ruoso rakudo: my $x = /some regex/; 1
p6eval rakudo 8009ac: ( no output )
pmichaud oh, wait. It might be a p6eval bug.
p6eval is trying to evaluate the result of A.new.do
I'm building and testing locally, though.
[particle]- ah, of course. 17:56
17:56 goksie left
pmichaud okay, I get it locally also. 17:56
nm.
ruoso it fails in rakudo directly as well
pmichaud Yes, I see it here.
17:56 eMaX joined
pmichaud rakudo: class A { has $!rx; method do { $!rx = regex { some regex } }; }; A.new.do; 1; 17:57
p6eval rakudo 8009ac: ( no output )
pmichaud there.
It's not a regex bug, at least not directly. :-)
[particle]- i found a parse bug, though :)
pmichaud Something is trying to evaluate the return value of the program, and *that's* where the bug occurs.
ruoso yes... it sure looks like something outside rakudo itself
pmichaud anyway, the error occurs when the program itself returns a regex. 17:58
rakudo: class A { has $!rx; method do { $!rx = regex { some regex } }; }; $_ = 'foo'; A.new.do; 17:59
p6eval rakudo 8009ac: OUTPUTĀ«Null PMC access in isa_pmc()ā¤current instr.: 'parrot;PGE;Match;new' pc 33 (compilers/pge/PGE/Match.pir:65)ā¤Ā»
pmichaud okay, I think I know how to resolve this one. 18:00
18:00 hercynium joined
pmichaud jnthn: ping (when back from dinner) 18:05
jnthn: actually, dog needs a walk -- I'll catch you during #ps 18:08
afk
ruoso rakudo: sub foo { say $*foo }; my $foo is context = 'hello'; foo(); 18:10
p6eval rakudo 8009ac: OUTPUTĀ«Use of uninitialized valueā¤ā¤Ā»
ruoso rakudo: sub foo { say $+foo }; my $foo is context = 'hello'; foo();
p6eval rakudo 8009ac: OUTPUTĀ«Scope not found for PAST::Var '$+foo' in fooā¤current instr.: 'parrot;PCT;HLLCompiler;panic' pc 146 (src/PCT/HLLCompiler.pir:105)ā¤Ā»
TimToady std: sub foo { say $+foo }; my $foo is context = 'hello'; foo(); 18:11
p6eval std 26684: OUTPUTĀ«Potential difficulties:ā¤ The + twigil is deprecated, use the * twigil instead at /tmp/mKDTftjOXl line 1:ā¤------> sub foo { say $+foo }; my $foo is context = 'hello'; fooā¤ok 00:02 36mā¤Ā»
ruoso TimToady, I was aware of that... I was just wondering if rakudo was still with the old syntax... since $*foo didn't work 18:13
TimToady what I figgered
I was just curious what std would say
ruoso rakudo: sub foo { say $*foo }; my $foo is context; $foo = 'hello'; foo(); 18:14
p6eval rakudo 8009ac: OUTPUTĀ«Use of uninitialized valueā¤ā¤Ā»
TimToady it's possible rakudo still treats $* as global
moritz_ I think it never did "real" context vars 18:15
ruoso it also seems that PGE doesn't like Perl 6 closures
18:16 Jedai joined
ruoso rakudo: sub foo { my sub bar { }; return / foo <bar> /; }; my $rx = foo; say 'a' ~~ $rx; 18:17
p6eval rakudo 8009ac: OUTPUTĀ«ā¤Ā»
ruoso rakudo: sub foo { my sub bar { }; return / foo <bar> /; }; my $rx = foo; say 'foo ' ~~ $rx;
p6eval rakudo 8009ac: OUTPUTĀ«Unable to find regex 'bar'ā¤Null PMC access in invoke()ā¤current instr.: 'parrot;PGE;Grammar;' pc 614 (EVAL_19:239)ā¤Ā»
pmichaud jnthn: reping 18:24
(PGE doesn't like Perl 6 closures) -- sure, this is a long-standing issue. :-)
rakudo still treats $* as global, yes. 18:25
ruoso pmichaud, is there any support for context vars? 18:27
pmichaud in regexes?
or in general?
short answer is no, rakudo doesn't haven't them yet. We have the pieces in place for them, though. 18:28
ruoso hmm... too bad
PerlJam rakudo: say a => 3; # what's going on here? 18:29
p6eval rakudo 8009ac: OUTPUTĀ«ā¤Ā»
18:29 ElectricHeavyLan joined
pmichaud rakudo: say (a=>3); 18:29
p6eval rakudo 8009ac: OUTPUTĀ«a 3ā¤Ā»
moritz_ p6eval: say() seemms to ignore named arguments
pmichaud ah yes.
moritz_ sorry, meant PerlJam
pmichaud it's a named arg.
PerlJam got it.
does say have any named args to accept?
moritz_ no 18:30
18:30 goksie joined
moritz_ at least I can't think of any 18:30
ruoso pmichaud, it seems I can't even use "my sub &foo" in the regex 18:32
pmichaud ruoso: I don't entirely understand... but basically PGE doesn't have any clue about how to parse things that aren't regex syntax. 18:34
ruoso the problem is not parsing 18:35
pmichaud then I fall back on the "I don't entirely understand" part.
ruoso the problem is that it will look the symbol somewhere else, not in the lexical scope where the rx was defiend 18:36
pmichaud it's possible that the regexes aren't properly lexically scoped, yes. See earlier statement about "PGE hasn't caught up to Parrot lexicals yet."
ruoso are you aware on a way to make it work? 18:37
(I'm open to useing PIR where needed...)
pmichaud not off the top of my head. I'm sorry if it sounds like I'm being obstructionist... I just have more problems on my plate than time to fix them all.
ruoso pmichaud, it didn't sound like that... ;) 18:38
skids IMO say should pass named args to Str.
pmichaud okay, good. I just know I'm providing more issues than answers at the moment, and that's not a mode I'm happy about being in. 18:39
skids: hmm?
skids If you want to override say with named arguments for a class, and also Str, then youhave to write in a custom say that passes named args to Str 18:40
ruoso pmichaud, anyway... I'll leave that code there for now then... the issues I'm blocking are contextual variables and closures in the regex....
skids IMO that should be the default behavior.
pmichaud ruoso: noted. I might be able to get contextuals working relatively quickly.
jnthn pmichaud: I'm back now
pmichaud jnthn: question on P6metaclass refactors. Ideally, what should the first argument to add_method, add_attribute, etc. be? 18:41
jnthn The "prototype" (which is an instance of the representation, and what will be our proto-object.) 18:42
18:42 icwiener-_- left
pmichaud I think I'd like the refactored versions to accept *only* prototypes, then. 18:42
and not be calling 'get_parrotclass' to figure it out.
jnthn Indeed, but we don't have those yet. :-)
pmichaud the deprecated versions can do the mapping to the prototype if needed. 18:43
i.e., I don't want them to accept parrotclasses, names of classes, namespaces, etc.
jnthn yes, understood. I'll try and remember that once I get parrotopaque in place; yell at me if I forget. :-)
pmichaud in looking at traces of things (especially signature binds) we end up making *lots* of calls to get_parrotclass. I'd like to optimize that out for cases where we know the incoming argument is well behaved. 18:44
And I'd like to enforce a bit more strictness on callers to only work through the prototype and not be able to pass "whatever" and have P6object dwim... at least not for those metaclass methods. 18:45
(things like 'register', 'new_class', etc. can still have some dwimminess)
japhb Anyone have a link for a screenshot (or online javascript page) showing Perl 6 syntax highlighting that includes {} interpolating into a string? I'm trying to explain the coolness to a friend, and it's clear I need a visual aid ....
moritz_ japhb: wait a sec... 18:46
japhb moritz_: than queue
jnthn Ah, signature binds. Another area that'll want some opt. :-)
pmichaud well, that would be a first step. Basically, we end up doing lots of get_parrotclass stuff whenever we're doing a typematch, and part of the reason is because the existing methods are more forgiving than they should be about the type object. 18:47
moritz_ japhb: feather.perl6.nl/~azawawi/html/spec...ion.t.html has a few basic examples 18:48
ruoso pmichaud, jnthn, the faz project is in github.com/ruoso/faz/tree ... I'll leave it there while I have those blockers... but just noting the repo so you can play with it around if you like...
japhb moritz_: excellent. I'll go wandering around near there. 18:49
moritz_ japhb: if you like vim, you can get svn.pugscode.org/pugs/util/perl6.vim
it hilights interpolating closures correctly 18:50
japhb I'll recommend that too him, he's a vim guy.
pmichaud I wonder if perl6.vim should be distributed with Rakudo, also.
moritz_ if so, we should get the latest version from github.com/hinrik/perl-vim/
japhb pmichaud: that seems like a win. 18:51
moritz_ (oh, it's github.com/hinrik/vim-perl actually) 18:52
[particle]- but it violates dry
japhb [particle]: git submodule ... no need to copy
But it can be packaged into the releases. And (hopefully) kept in sync with changes to Rakudo. 18:53
18:54 ElectricHeavyLan left
japhb (The power of submodules was the one thing I thought git was really missing ... and then it got them.) 18:55
moritz_ japhb: do you know an introductory document to them?
japhb moritz_: Hmmm, I stumbled over one a couple weeks back. Lemme see if I can find it. 18:56
moritz_: A couple github links show up in my history: github.com/guides/developing-with-submodules and github.com/guides/using-git-submodu...ck-plugins 18:58
18:58 LylePerl left
PerlJam japhb: you didn't think partial checkouts were something git was missing too? 19:00
japhb Hmmm, and a quick scan of the Git User Manual's chapter 8 looks decent: www.kernel.org/pub/software/scm/git...submodules
moritz_ japhb: thanks
japhb PerlJam: Yes and no ... there are a couple repos for which a partial checkout would have been nice for me, but I always found it infuriating in my earlier SCM days when I had to expand my checkout scope ... so I eventually got in the habit of always checking out full repos. This is a lot easier when a clone is a fast operation. Unlike, say, in SVK. 19:02
OR git-svn for that matter.
(And to finish the flip side of the thought ... the Venn diagram describing the interaction of the MPlayer and FFMPEG repos was what convinced me submodules weren't just nice, but basically essential to maintain some semblance of sanity and convenience. Apparently those projects felt the same.) 19:07
19:11 wolv joined 19:13 barney left
japhb rakudo: say "Perl 6 is { Ā« magic "sufficiently advanced technology" Ā».pick.uc }" 19:14
p6eval rakudo 8009ac: OUTPUTĀ«Unable to parse block; couldn't find final '}' at line 1, near "\x{ab} magic \"s"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
japhb std: say "Perl 6 is { Ā« magic "sufficiently advanced technology" Ā».pick.uc }" 19:15
p6eval std 26684: OUTPUTĀ«ok 00:04 36mā¤Ā»
19:15 donaldh joined
japhb std: say "Perl 6 is { Ā« magic 'sufficiently advanced technology" Ā».pick.uc }" 19:15
p6eval std 26684: OUTPUTĀ«ok 00:04 36mā¤Ā»
pmichaud rakudo doesn't parse Ā« yet, I don't think.
japhb pmichaud: Ah!
19:16 minazo joined
pmichaud rakudo: say "Perl 6 is { << magic 'sufficiently advanced technology" >>.pick.uc }" 19:16
p6eval rakudo 8009ac: OUTPUTĀ«Perl 6 is MAGICā¤Ā»
pmichaud rakudo: say "Perl 6 is { << magic 'sufficiently advanced technology" >>.pick.uc }"
p6eval rakudo 8009ac: OUTPUTĀ«Perl 6 is MAGICā¤Ā»
pmichaud rakudo: say "Perl 6 is { << magic 'sufficiently advanced technology" >>.pick.uc }"
p6eval rakudo 8009ac: OUTPUTĀ«Perl 6 is ADVANCEDā¤Ā»
pmichaud rakudo: say "Perl 6 is { << magic "sufficiently advanced technology" >>.pick.uc }"
PerlJam heh
p6eval rakudo 8009ac: OUTPUTĀ«Perl 6 is MAGICā¤Ā»
japhb pmichaud: had a misquote
pmichaud rakudo: say "Perl 6 is { << magic "sufficiently advanced technology" >>.pick.uc }"
p6eval rakudo 8009ac: OUTPUTĀ«Perl 6 is SUFFICIENTLY ADVANCED TECHNOLOGYā¤Ā»
PerlJam the << >> parser could look for quotes and warn 19:17
japhb pmichaud: Is there any reason for Ā« Ā» to be difficult to parse? Or is it Unicode problems raising their collective ugly heads? 19:19
pmichaud Unicode problems.
jnthn The one with Parrot hashes and unicode keys?
pmichaud I think so, yes.
19:20 donaldh left, donaldh joined
japhb Is there a way to weight the choice of a .pick ? 19:21
(Other than copying terms to make appropriate ratios)
moritz_ no
19:21 goksie left
moritz_ you could sort on weight * rand(), and then take the first one 19:22
or so
japhb Hmmm, actually, that's not a bad idea, mixed with some of MJD's discussions on algorithms for finding smallest terms "nearly correct" fractions
s/that's/copying terms/
well, you get the point. 19:23
jnthn rakudo: my @values = 1,2,3; my @weights = 2, 4, 2; say @values >>xx<< @weights;
p6eval rakudo 8009ac: OUTPUTĀ«11222233ā¤Ā»
jnthn ^ easy way to do the copying terms way
19:23 donaldh left
japhb jnthn: nice! 19:23
19:23 donaldh joined
arnsholt But having an optional weights parameter might be better_ 19:23
s/_/?/ 19:24
moritz_ rakudo: my %w = a => 1, b => 2, c => 3; <a b c>.sort( { rand * %w{$_} }).[0].say
arnsholt If the array to pick from is large, starting to copy might be an unreasonable suggestion
jnthn arnsholt: multi pick(@list, @weights) { ... }
p6eval rakudo 8009ac: OUTPUTĀ«aā¤Ā»
moritz_ rakudo: my %w = a => 1, b => 2, c => 3; <a b c>.sort( { rand * %w{$_} }).[0].say
p6eval rakudo 8009ac: OUTPUTĀ«cā¤Ā»
moritz_ rakudo: my %w = a => 1, b => 2, c => 3; for 1..10 {<a b c>.sort( { rand * %w{$_} }).[*-1].print } 19:25
arnsholt moritz_: Right. That works too
p6eval rakudo 8009ac: OUTPUTĀ«bbccccbccaĀ»
moritz_ ... if I do it right, yes ;-)
japhb moritz_: nice.
moritz_ arnsholt: pick with weight is a nice feature for a library, IMHO
japhb moritz_: sure, I've got no problem with that at all. It certainly feels more like library functionality than core to me .... 19:26
19:26 azawawi left
arnsholt Yeah, it sounds like a reasonably common operation 19:26
19:26 ZuLuuuuuu joined 19:34 lisppaste3 joined 19:35 donaldh left
ruoso later & 19:51
19:59 wolv left 20:04 amoc joined 20:06 amoc left, wolv joined
dalek kudo: cfc5e82 | pmichaud++ | docs/spectest-progress.csv:
spectest-progress.csv update: 379 files, 11011 passing, 0 failing
 20:11
20:17 amoc joined
jnthn rakudo: multi foo() { }; my @a; @a.push(&foo) 20:18
p6eval rakudo cfc5e8: ( no output )
20:25 PhatEddy joined
PhatEddy rakudo: enum B <a> ; say B::a 20:26
p6eval rakudo cfc5e8: OUTPUTĀ«0ā¤Ā»
PhatEddy rakudo: enum A <a> ; say A::a
p6eval rakudo cfc5e8: OUTPUTĀ«0ā¤Ā»
jnthn :-)
(Though that much worked just fine before the recent refactor...) 20:27
PhatEddy rakudo: class A { say "This is class $?PACKAGE" } 20:28
p6eval rakudo cfc5e8: OUTPUTĀ«Scope not found for PAST::Var '$?PACKAGE' in ā¤current instr.: 'parrot;PCT;HLLCompiler;panic' pc 146 (src/PCT/HLLCompiler.pir:105)ā¤Ā»
PhatEddy that one was closed once I think
jnthn rakudo: class A { say $?PACKAGE } 20:29
p6eval rakudo cfc5e8: OUTPUTĀ«Scope not found for PAST::Var '$?PACKAGE' in ā¤current instr.: 'parrot;PCT;HLLCompiler;panic' pc 146 (src/PCT/HLLCompiler.pir:105)ā¤Ā»
jnthn rakudo: module A { say $?PACKAGE }
p6eval rakudo cfc5e8: OUTPUTĀ«Scope not found for PAST::Var '$?PACKAGE' in ā¤current instr.: 'parrot;PCT;HLLCompiler;panic' pc 146 (src/PCT/HLLCompiler.pir:105)ā¤Ā»
jnthn Yeah, I thoguht so too.
PhatEddy was RT #59484
working on package tests ... 20:30
20:38 rewt left 20:43 LadyLunacy joined 20:52 Kyosuke_Kiryu joined
PhatEddy rakudo: package Simple { sub pkg { $?PACKAGE } } 21:01
p6eval rakudo cfc5e8: OUTPUTĀ«Scope not found for PAST::Var '$?PACKAGE' in pkgā¤current instr.: 'parrot;PCT;HLLCompiler;panic' pc 146 (src/PCT/HLLCompiler.pir:105)ā¤Ā»
21:04 mizioumt1 joined 21:07 payload joined 21:09 iblechbot joined 21:12 rewt joined, ejs joined 21:15 hercynium left
PhatEddy rakudo: package Simple{class Bar {sub baz {'hi'}}}; say Simple::Bar.new.baz; 21:19
p6eval rakudo cfc5e8: OUTPUTĀ«Could not locate a method 'baz' to invoke on class 'Simple::Bar'.ā¤current instr.: 'die' pc 16538 (src/builtins/control.pir:225)ā¤Ā»
PhatEddy oops
dalek kudo: 9a1cacd | jnthn++ | (2 files):
Add .^methods and .^isa introspection methods. Also bump us up to the Parrot build where I put in the P6object changes that are needed to go with this.
 21:20
kudo: c77bd19 | jnthn++ | :
Merge branch 'master' of [email@hidden.address]
PhatEddy rakudo: package Simple{class Bar {method baz {'hi'}}}; say Simple::Bar.new.baz;
p6eval rakudo cfc5e8: OUTPUTĀ«hiā¤Ā»
21:23 Kisu joined 21:25 ejs left 21:26 nihiliad joined 21:33 iblechbot left, amoc left 21:35 amoc joined 21:45 LadyLunacy left 21:52 wknight8111 joined
dalek kudo: 53f7059 | jnthn++ | src/parser/ (2 files):
STD tracking: handles trait verb followed by noun, not EXPR.
 21:59
22:02 skids joined
PhatEddy rakudo: package A { enum B <a> }; package P { role B {}; class C does P::B {} } 22:02
p6eval rakudo c77bd1: OUTPUTĀ«The type P::B does not exist. at line 1, near "{} }"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā» 22:03
PhatEddy rakudo: package A { enum B_different <a> }; package P { role B {}; class C does P::B {} }
p6eval rakudo c77bd1: ( no output )
hatseflats so how are method collisions handled in perl6 when a role has a method named like a method contained in the class importing it? 22:04
jnthn The method in the class always wins. 22:05
hatseflats so there is no late binding?
hmmm
jnthn Two roles both trying to bring in methods of the same name, without any method in the class to resolve it, is an error though (at composition time)
(e.g. compile time)
[particle]- class construction time 22:06
hatseflats veering a bit off, wouldn't it be better to throw an error regardless?
jnthn No, since in Perl 6 writing a method in the class is *the* mechanism for conflict resolution.
PhatEddy In that example above - any good reason for the role and enum in different packages to conflict? 22:07
hatseflats okay
jnthn PhatEddy: No, it's an error.
rakudo: role A { method m { } }; role B { method m { } }; class C does A does B { }
p6eval rakudo 53f705: OUTPUTĀ«A conflict occurred during role composition due to method 'm'.ā¤current instr.: '!meta_compose' pc -614965 ((unknown file):-1)ā¤Ā»
jnthn rakudo: role A { method m { } }; role B { method m { } }; class C does A does B { method m { } }
p6eval rakudo 53f705: ( no output )
hatseflats so besides the difference in overriding behavoir and conflicting imports, and that roles do not add to the inheritance chain, are there any other differences, technically, between extending and using a role? 22:08
or traits for that matter
jnthn Yes and no. 22:09
[particle]- extending?
jnthn [particle]-: synonym for inheriting
For example, a role can also be used as a mix-in.
Also, roles are immutable. Classes are mutable.
hatseflats my bad, I seem to be mixing terminology a lot, I'm not used to perl, and draw a lot of the vocabulary from other sources
and a mix-in would be a non-abstract version of a role, correct? 22:10
jnthn No, a mix-in is just applying a role to an object on a per-object basis.
hatseflats oh, okay
ah yes, that trick is particularly interesting 22:11
jnthn Well, I guess you could say we have the action of mixing in (a verb) more than a construct that we call a mix-in (a noun).
Thus you mix in a role. :-)
PerlJam sprinkles chocolate and pecans and cinnamon on his vanilla ice cream and starts mixing.
hatseflats I see :)
I'm asking because I'm trying to develop a bit or a nice language as part of my school project, and since the perl userbase seems to have the most exp. points, so to speak, on the subject, I ask here. 22:12
In my case, roles will be classes that are used as roles 22:13
[particle]- jnthn: we need to be careful about language, because so many oo framworks have different meanings for different words, so i asked.
jnthn [particle]-: Yes, you're right.
[particle]-: I move between a bunch of different ones all the time...
[particle]- like, perl 6 roles are smalltalk traits, 'cept our roles have state 22:14
jnthn What we call attributes, C# calls fields, and its attributes are something every different...
[particle]- yep, attributes versus properties, too
hatseflats others call them members
jnthn And what Perl 6 calls a trait (mixing in a role to something at compile time) is not what smalltalk calls a trait. 22:15
well
that's not quite what Perl 6 calls a trait, but it's often the upshot... 22:16
hatseflats that's kinda... confusing
regardless, thanks for the info
PerlJam If only they all used standardized terminology ...
22:17 simcop2387 joined
hatseflats heh 22:17
'they all'
pugs_svn r26685 | lwall++ | [Containers] some work on various structures related to KeyHash 22:20
r26685 | lwall++ | define .pick to work rationally on structures whose values represent weights.
22:22 Jedai left
TimToady I'd be open to renaming 'traits' to something less confusing 22:23
hatseflats how about, non-inheriting, non-overriding, non-instantiatable classes 22:25
jnthn TimToady: I'm not so fussed....SmallTalk ain't *that* mainstream.
If Java used it to mean something else, I'd probably feel differently.
22:26 hercynium joined
hatseflats someone has to set a trend 22:26
antiquated words get revised meanings all the time
who's to say that Java won't pick up on Perl-style traits and roles
or maybe we'll have world peace and Perl and Python will combine into one language
PerlJam hatseflats: what makes you think that world would be peaceful? 22:27
hatseflats everyone will be too busy trolling on the internet on the subject to make war? 22:28
arnsholt PerlJam: But at least we would be able to focus on the really important questions 22:29
Like which editor people should use =)
hatseflats heh
jnthn Notepad!!!
PerlJam arnsholt: you use vi to edit perl 6 (see how vi and 6 are the same? :)
arnsholt PerlJam: Obviously. vi is the one true editor, after all =) 22:30
TimToady well, but Python and E-lisp are both just dialects of Perl 6...
hatseflats *cringe*
jnthn oh hey, that means so is lolcode...
PerlJam arnsholt: yeah ... even if it's skin looks like vim, it's still got vi on the insides somewhere.
What *isn't* a dialect of perl 6 ? 22:31
arnsholt Yeah. vim is just vi with makeup on
TimToady strawberries
black holes
PerlJam TimToady: now you're tempting the Damian.
TimToady probably quarks, but we're not sure
hatseflats I'd say Basic is a pretty safe bet to claim as not-perl based
TimToady you don't understand; Perl 6 can easily mutate into *any* other computer language 22:32
even Basic
jnthn { use Basic; 10: PRINT "HELLO" }
hatseflats if that's true
TimToady you forgot the 'goto 10'
hatseflats than isn't perl just scheme with a fancy community? 22:33
TimToady well, lisp with a syntax
the eventual goal is to convert Paul Graham :) 22:34
PerlJam That's more Haskell
hatseflats yeah... how would one classify haskell
as if the gods created the world with it, but then ditched it and wrote the bulk of the things on it in perl 22:35
TimToady well, you can end a Perl program with a right paren
in any case, the universe would consider indentation as syntax to be magical action at a distance, unless there are particles travelling back and forth between the text and the left margin 22:37
.oO(spaces, the final frontier)
hatseflats heh... hehehe :) 22:38
PerlJam use Acme::Bleach
22:38 Kyosuke_Kiryu left
TimToady well, when Haskell gets around to inventing the QM monad, the universe will implode as self-referential, I expect 22:40
pmichaud and here I figured that a self-reference was what kicked the whole thing off in the first place.
hatseflats sorry, ah whatshamacallit monad?
TimToady fortunately monads are still one-dimensional
QM = Quantum Mechanical 22:41
pmichaud Someone overflowed the stack somewhere and this universe is the result.
22:41 Kyosuke_Kiryu joined
jnthn rakudo: sub foo() { my @a = 1; return @a; }; my @x = foo(); say @x[0].perl; 22:41
p6eval rakudo 53f705: OUTPUTĀ«1ā¤Ā»
PerlJam The big thing in haskell is arrows now, not monads. (which works well for entropy and time and such)
japhb PerlJam: OK, last time I looked at Haskell, I had figured out monads, more or less. What are arrows now? 22:42
TimToady better monads :)
japhb oh.
PerlJam japhb: www.haskell.org/arrows/
japhb "But if you're using a structure that's very like a monad, but isn't one, maybe it's an arrow." 22:43
heh
PerlJam or it could be Higgs Boson
TimToady maybe we should rename traits to arrows and confuse everyone 22:44
hatseflats how would that be more confusing?
PerlJam hatseflats: are you saying that we've reached maximal confusion? There is no more? 22:45
TimToady well, I guess an argument can be made that everyone is already confused
hatseflats PerlJam: if the universe dies of maximal entropy, why not a similar fate for us humans
jnthn pmichaud: ping 22:46
pmichaud jnthn: pong
jnthn oh hai
pmichaud also
jnthn: ping :-)
jnthn I'm doing something like
@classes = self.^isa();
lambdabot Unknown command, try @list
jnthn That method is implemented in PIR
And creates an Array and returns it. .return (the_array) 22:47
Which ends up with @classes containing a first element that is that array.
However, that differs from the case I fed to Rakudo above.
Am I doing something obviously stupid?
(the isa method is implemented in PIR, i mean)
pmichaud rakudo: sub foo() { my @a = 1; @a }; my @x = foo(); say @x[0].perl;
p6eval rakudo 53f705: OUTPUTĀ«1ā¤Ā» 22:48
TimToady well, returns should generally be returning Captures, if they don't know what context they'll be used
pmichaud is the_array really an Array?
jnthn result_list = get_hll_global 'Array'
result_list = result_list.'new'()
...
.return (result_list)
hatseflats say $*IN;
err
rakudo: say $*IN; 22:49
?
p6eval rakudo 53f705: OUTPUTĀ«IO()<0xb6b73668>ā¤Ā»
pmichaud jnthn: afaik, that should work out okay. I'd probably want to see more detailed code.
You could try $P0 = result_list.'list'(); return $P0
pugs_svn r26686 | lwall++ | [STD] make lookup of $?FOO overridable in derived languages
pmichaud i.e., to see the definition of whateveris handling '.^isa' 22:50
hatseflats rakudo: say $*IN.lines().pick();
p6eval rakudo 53f705: OUTPUTĀ«Hast seit frĆ¼hen Ahnentagenā¤Ā»
hatseflats ah yes, the austrian anthem 22:51
jnthn pmichaud: If I do .perl of the thingy that I'm returning inside the method it's [A]
So looks right
Let me get you a link though...
pmichaud well, several things will cause it to return [A] though.
indeed, I wonder if it's in fact a reference to an array somehow.
jnthn github.com/rakudo/rakudo/blob/53f70...assHOW.pir 22:52
methods below has the same issue. 22:53
pmichaud oh, fwiw, the line
result_list = get_hll_global 'Array'
is going to be wrong.
japhb PerlJam: Thanks for the Arrows link. Makes sense, but brain didn't expand. Which either means they are common sense, or I'm missing something.
jnthn pmichaud: Because...? 22:54
japhb Perhaps "common sense if monads are already understood"
pmichaud because this code is running in the 'parrot' HLL namespace.
hatseflats haha
jnthn Ah.
pmichaud and Rakudo's 'Array' protoobject will be in a different namespace.
jnthn Yes, yes.
22:55 amoc left
jnthn Well, that said, by the time we get .HLL in place, Rakudo's metaclass and P6object may be been broken out. 22:55
pmichaud could be.
jnthn (As in, Rakudo may have its own one in its own HLL)
pmichaud at any rate, most array-like-stuff will end up being in RPA
i.e., Parrot's RPA will map to Rakudo's Array
is the code in 'isa' "new code", or has it been around a while? 22:56
jnthn New.
Though it'll change once get_parrotclass stops being needed.
pmichaud right.
jnthn Just trying to fill out the metaclass enough so that I can ween other places off using Parrot classes. 22:57
pmichaud some things to try
(1) use new 'Perl6Array' instead of going through the 'Array' protoobject. I'm just curious if it makes a difference.
(2) find out what result_list.'PARROT'() says
oddly, when I run it on my box it appears to do the assignment correctly. 22:59
hatseflats why choose arrays to be pseudo or proto-objects?
jnthn oh, hmm
pmichaud $ cat y
my @isa = 1.^isa();
say @isa.perl;
$ ./perl6 y
[Int, Any]
$
jnthn oh, hmm
What if that's Int.^isa() ? 23:00
Oh, ugh, that is quite bogus too (should just be Any - the Int is coming because we inherit from Parrot's Integer)
pmichaud $ cat z
my @isa = Int.^isa();
say @isa.perl;
$ ./perl6 z
[Int, Any]
$
^H^H^H^H^H^H[B[B[B[A[A
doesn't 'parents' end up returning the class itself in the list? 23:01
istr that the first element of introspecting parents was that it returned the class itself. 23:02
jnthn I think ^isa by default is just the immediate parents
Not the full MRO. 23:03
pmichaud I think that Parrot's "parents" is the full mro.
maybe not, checking.
jnthn No, there's two things
hatseflats perl5: say 42;
jnthn parents and all_parents
First is immediates, second is MRO. 23:04
pmichaud you're correct.
but yes, technically, Int has two parents right now.
jnthn I think something odd is going on though.
Because a shorter snippet just gave me the answer I expected...
23:06 H1N1 joined
pmichaud I think I'm going to add "is vtable(...)" as one of our allowed traits. 23:07
23:07 kane___ joined
pmichaud so we can write vtable functions in the setting. 23:07
jnthn Were we not going to just map 'em all in Object? 23:08
pmichaud I'm not sure they always map.
I might go that approach to begin with.
jnthn OK, no objections from me if you need it. 23:09
Guess it'll be a compiler-handled trait rather than one we dispatch on.
pmichaud I was going to handle it the same as the other sub traits.
i.e., it would call sub_trait_vtable and do the vtable insertion.
at least as a first cut, since that's easy (and consistent with our other trait handling) 23:10
but maybe I'll go with mapping them all in Object for now. 23:11
(these are with respect to refactoring List and Array)
jnthn That's fine.
I'm not going to be doing traits for a bit.
pmichaud Somehow it seems I spend most of my time rewriting/refactoring existing code. :-| 23:12
I guess that's normal, though. 23:13
jnthn Heh, in the next couple of weeks I've probably gotta re-write all of the method dispatcher, .*, .+, .?, wrap, unwrap...
pmichaud I'm glad to let you do that :-)
jnthn Gah. :-( So I take the code out of the setting, pretty much copy-paste, into a separate .p6 file and it works... :-| 23:14
TimToady bootstrap dependency?
jnthn TimToady: Maybe, but then I'm struggling to think where. The code isn't being called until everything else is compiled... 23:15
pmichaud oh, on a related topic
I think we might get to the point soon where we ship with a "bootstrapping" settings.pir 23:16
this is used to compile the real settings, which then replace it.
jnthn I'd been thinking we might end up with that.
pmichaud same here.
but I'm getting to the point where I really prefer to be writing things in p6 (with inline PIR) than pure PIR. 23:17
jnthn Some things are getting quite a nuisance now.
hatseflats stupid question, but why does it make sense that I can access a package created outside the local scope?
jnthn Like, if you re-declare a variable in the setting.
Then warn uses join which isn't existing until we've compiled the setting. 23:18
And so you never get the error.
pmichaud warn could always use the opcode join :-)
jnthn It just whines about not finding join.
pmichaud but yes, our bootstrapping version could provide a join that warn can use.
jnthn Yeah of course. It's just an example of the sort of problem.
Like, most of IO is defined in the setting. 23:19
pmichaud that is then replaced by the 'real' join.
jnthn say() just delegates.
To $*OUT.say
Which doesn't exist.
So you stick a say(...) in actions.pm for debugging and it fails.
TimToady hmm, an exported method has an implicit *%_, which might explain say( a => 42 ) throwing away the pair 23:20
jnthn (which sucks when you're trying to debug...)
pmichaud TimToady: yes, I was thinking that earlier.
But I think it may also related to the discussion on pairs in list(...) and hash(...) 23:21
*relate
TimToady maybe we can refine the automatic *%_ rule somehow
pmichaud for the time being, we could probably have say() warn if it gets any named params.
Is there a strong reason for the automatic *%_ rule? 23:22
jnthn pmichaud: this is odd...
say "before loop: " ~ @classes.perl;
for @classes -> $class {
say $class.perl();
}
Output
before loop: [A]
[A]
pmichaud jnthn: it's possible that .perl is subtly modifying its invocant somehow. 23:23
I know that there was some other oddities with .perl recently.
what is @classes.PARROT; reporting? 23:24
jnthn True, but it's consistent with what goes wrong next...
(Which is that I do $class.^methods and it gives me the methods of Array...
)
I change the .perl to .PARROT and get 23:25
before loop: Perl6Array
Perl6Scalar->Perl6Array
pmichaud aha
yes, something is converting @classes into a Perl6Scalar
meppl good night 23:26
jnthn Note the second line is inside the loop
e.g. say $class.PARROT
pmichaud try @classes[0].PARROT
(before the loop)
wait. 23:27
I might know. What exactly is *in* @classes at this point?
jnthn As far as I can see nothing shoulda touched it other than what came back from @classes = self.^isa();
23:27 PhatEddy left, ZuLuuuuuu left
pmichaud try @classes[0].PARROT 23:27
we know that @classes is a Perl6Array, I need to know its first element. 23:28
jnthn I do say 'before loop @classes[0]: ' ~ @classes[0].PARROT;
and the output is just
before loop @classes[0]:
(e.g. .PARROT gives nothing)
pmichaud because it's an (unnamed) proto.
protoobjects don't get class names by default. 23:29
jnthn ah, yes
pmichaud which proto is it?
jnthn I just defined two classes
class A { method x() { say 1 } }
class B is A { method x() { say 2 } }
And it'd be A's proto
pmichaud testing.
and @classes is.... B.^isa() ? 23:30
jnthn aye
pmichaud you have some local mods in place? 23:31
jnthn The method I'm writing in the setting 23:33
oh
And the return value to list thing you suggested.
pmichaud what's the method you're writing in the setting?
jnthn WALK
(at least partly as an attempt at proof-of-concept of the introspection stuff...) 23:34
pmichaud on my system: nopaste.snit.ch/16478
can I see what you have for WALK ?
jnthn one second, I'm just pulling it out of the setting into an is also to make sure it's really not setting specific... 23:35
pmichaud oh, I know what else it might be.
What *other* methods do you have defined in there...?
jnthn Just git stash'd what i had before.
pmichaud and what class is this being defined in?
jnthn Object. 23:36
pmichaud Yes, that's the problem.
jnthn OH NOES
Yes, I think I can guess where this one is going... :-(
pmichaud Object has its own "list" method, and that's being called as part of the 'list' function.
because
for @classes -> $class { ... }
is really
for list(@classes) -> $class { .... } 23:37
but
because Parrot sticks its methods into the namespace
and because IMCC short-circuits to any named sub in the same namespace
you're not getting the global version of list(), you're getting Object's version of .list()
which returns a list of itself.
i.e., what you're getting is the same as
for Object::list(@classes) -> $class { ... } 23:38
jnthn Yeah. Ugh.
pmichaud I _really_ wish we could get that particular namespace bug fixed. 23:39
jnthn you're right...
nopaste.snit.ch/16479 23:40
That works, but s/Any/Object/ (to add it to Object) and if blows up.
pmichaud right, because Any doesn't have a 'list' method/sub defined. 23:41
jnthn Can't believe we just lost another hour to a stupid Parrot fail attempt at optimization.
pmichaud so IMCC doesn't do the short-circuit optimization.
Perhaps the workaround/solution here is to get PCT to always use :method('name') when generating methods.
And give the sub an empty or nonsense or otherwise mangled name.
jnthn Only works if it's not an our method though... 23:42
The others need to be in the ns.
pmichaud hmmm?
jnthn If it's an our method it has to go in the NS.
I'm not sure what stuff in the setting counts for on that front though.
pmichaud do we have anything that really makes use of 'our methods'?
jnthn I'm not sure. Is it the default? 23:43
The point is more, do other people have code that is making use of them...
pmichaud no, the question is, "what is Spec"?
if people have code that relies on it, and it's not spec, then I'm not too worried about breaking it.
jnthn Sure. 23:44
I don't know which is spec...checking...
pmichaud but I don't remember that methods go into the namespace. This was discussed recently.
I thought they stayed out of the namespace unless 'is export' was present.
(I can be misremembering this.)
jnthn Unlike with most other declarations, C<method> declarations do not 23:45
default to C<our> semantics, or even C<my> semantics, but rather
C<has> semantics.
You're right.
pmichaud right
and the sentence after that.
jnthn yes, that's the upshot of what I pasted.
pmichaud I don't think anyone is using 'our'
if they are, we can certainly arrange for it to go back into the namespace 23:46
jnthn sure
and lo and behold
.namespace ['Perl6Object']
-.sub 'list' :method
+.sub '' :method('list')
Solve the issue.
pmichaud oh, do that for now.
My apologies that it wasn't done sooner.
We should do that for most built-in methods.
jnthn No, no, sorry for dragging you into the debugging.
I shoulda guessed that one sooner... 23:47
pmichaud it often trips me up
I didn't realize you were working inside of Object.
jnthn I did know about Parrot's failtimization.
pmichaud more to the point, however.
because I've also had it happen in settings code
I think I should update PCT to allow :nsentry
and then Rakudo should default to :nsentry('') unless 'our' is present. 23:48
that way the sub can still have a name, but we control where it goes in the namespace.
jnthn *nod*
23:49 c1sung joined
pmichaud this of course, is the other reason we should probably start doing p6-with-inline-PIR instead of pure PIR, to avoid these sorts of things in the first place. 23:50
afk for a while -- need a walk. 23:52
23:53 mizioumt1 left

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!