»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg p6eval perl6: ... | irclog: irc.perl6.org/ | UTF-8 is our friend!
Set by sorear on 4 February 2011.
00:06 araujo left 00:07 noganex joined 00:08 HarryS left
sorear um. 00:10
00:10 noganex_ left
sorear deleting all traces of a Parrot install: makes no difference 00:10
strace says that there was no reference outside the current directory! 00:13
00:17 daniel-s joined 00:18 HarryS joined 00:23 HarryS left 00:28 silug joined 00:35 daniel-s left
dalek ecza: 1a9706b | sorear++ | / (3 files):
Implement autothreading of only subs
 00:42
sorear rakudo: (2 ~~ 2 | 3).perl.say 00:44
p6eval rakudo 048573: OUTPUT«Bool::True␤»
00:46 silug left
dalek ecza: 5b69137 | sorear++ | / (3 files):
Implement junctional ~~, special cases for !op and !=
 00:50
sorear \o/
junctions in niecza are now basically complete 00:51
TimToady yäy 00:57
01:04 awwaiid left
sorear is current Rakudo expected to be buildable? 01:06
01:06 benabik_ joined, benabik left, benabik_ is now known as benabik
sorear tries to build nom instead. 01:08
TimToady yes, I've built it successfully 01:10
Linux edo 2.6.38-8-generic #42-Ubuntu SMP Mon Apr 11 03:31:24 UTC 2011 x86_64 x86_64 x86_64 GNU/Linux
01:14 thou left 01:29 bacek left 01:45 daniel-s joined
daniel-s Hi 01:45
sorear hello, daniel-s. 01:47
daniel-s So, when i find something i don't think is correct in the perl book or documentation, i've been messaging perlpilot 01:48
Should i be writing it in the bug track thingy on github? 01:49
colomon That would probably be a bit better, yes. 01:54
sorear rakudo: say defined Mu 01:56
p6eval rakudo 048573: OUTPUT«Bool::False␤»
sorear rakudo: say (defined (1|2)).perl
p6eval rakudo 048573: OUTPUT«Bool::True␤»
01:57 whiteknight left
sorear o/ colomon 01:58
colomon \o
sorear TimToady: is it OK for spectests to use so(5)? 01:59
TimToady: does that mean that I need a &so in the setting, in addition to &prefix:<so>? 02:00
colomon sorear: do I understand the backlog correctly to mean you are actually handling junction logic in parallel now? or is it just autothreading the way Rakudo has it now? 02:16
daniel-s Perl6: my $a = 6.5; $a.say; 02:19
perl6: my $a = 6.5; $a.say; 02:20
p6eval pugs, rakudo 048573: OUTPUT«6.5␤»
..niecza v6-43-g5b69137: OUTPUT«13/2␤»
daniel-s That's interesting, rakudo and rugs give decimal, the other is fraction 02:21
colomon daniel-s: rakudo is also a fraction internally, but it prints it as a decimal 02:22
02:23 am0c joined
colomon rakudo: my $a = 6.5; $a.WHAT.say; $a.perl.say 02:23
p6eval rakudo 048573: OUTPUT«Rat()␤13/2␤»
colomon rakudo: my $a = 6.5; say ($a + 1/17).perl.say 02:26
p6eval rakudo 048573: OUTPUT«223/34␤Bool::True␤»
02:30 araujo joined
TimToady yes, I think we should provide functional forms of prefixes where appropriate 02:32
TimToady wonders if there's a way to do that systematically 02:34
(other than just forcing everyone to define them) 02:35
bbl &
02:40 ymasory joined 02:53 HarryS joined 03:03 starcoder left 03:04 Chillance left 03:05 starcoder joined 03:09 DarthGandalf left 03:10 sivoais left, sivoais joined 03:11 Util left, Util joined, DarthGandalf joined, jasonmay left 03:13 jtpalmer left 03:15 jtpalmer joined 03:16 hugme left 03:17 daniel-s_ joined, jasonmay joined 03:20 daniel-s left 03:21 daniel-s joined 03:22 mberends joined 03:23 Su-Shee_ joined 03:26 Su-Shee left 03:37 daniel-s left 03:42 _jaldhar joined, starcoder left, starcoder joined 04:01 thou joined 04:03 thou left 04:09 silug joined 04:11 satyavvd joined 04:15 envi joined 04:44 ymasory left
pmichaud suggestion for new method on List: 04:47
method xxlast(Int $n = 1) { ... } # makes any finite List into an infinite list by repeating the last $n elements 04:48
04:48 silug left
pmichaud thus @array.xxlast will have its last element repeated infinitely 04:48
04:48 silug joined
pmichaud @array.xxlist(3) will repeat the last three elements infinitely 04:48
*xxlast 04:49
(feel free to suggest better method names)
in general, xxlast($n) on a list acts like @list[*], (@list[*-$n .. *-1] xx *) 04:50
(but can do so lazily)
04:51 Alax joined
Alax 能用中文? 04:52
pmichaud the infinitely extensible list in S03:4314 then becomes @array.xxlast
05:02 silug left
TimToady Alax: 时大时小 05:06
今小... 05:07
JimmyZ 用中文 05:08
and masak
我用日本語。 05:09
05:11 hudnix left 05:14 sftp left, birdwindupbird joined 05:15 Khisanth left
TimToady pmichaud: .repfinal maybe 05:17
05:18 _jaldhar left 05:21 Eevee left
sorear daniel-s_: that's a bug in niecza. 05:22
05:25 daniel-s_ left, daniel-s_ joined 05:26 tadam joined
sorear daniel-s_: that's a bug in niecza. 05:26
daniel-s_ what is?
sorear rationals printing with a slash
daniel-s_ oh
sorear colomon: "autothreading" is the official name. A quite bad one, if you ask me. 05:27
daniel-s_ sorear: want me to list it on here: 05:28
github.com/sorear/niecza/issues
05:28 tadam left
sorear No 05:28
05:29 Khisanth joined
sorear there are already tests for it 05:30
I don't need a bug tracker to tell me things that comm -2 -3 <(find . | cut -c3- | sort) <(sort ~/rel/niecza/t/spectest.data) or git grep '#?niecza' can tell me
05:43 molaf joined
mberends sorear: smartass ;) 05:43
05:47 tadam joined 05:48 tadam_ joined, tadam left, tadam_ is now known as tadam
tylercurtis niecza: say(3/2); 05:49
p6eval niecza v6-43-g5b69137: OUTPUT«3/2␤»
tylercurtis rakudo: say(3/2) 05:50
p6eval rakudo 048573: OUTPUT«1.5␤»
tylercurtis rakudo: say .perl given 3/2
p6eval rakudo 048573: OUTPUT«3/2␤»
sorear \o/ nom builds for me 05:53
master however does NOT
05:54 woosley joined
sorear (let me try this again...) 05:54
dalek ast: af86cfd | sorear++ | S03-junctions/autothreading.t:
Fudge and simplify S03-junctions/autothreading for Niecza
 05:55
ast: b0f1ece | sorear++ | S03-junctions/boolean-context.t:
Defossilize S03-junctions/boolean-context.t a bit. == does not accept Any arguments
05:58 tadam left 06:04 am0c left 06:09 nymacro joined 06:14 Mowah joined 06:15 koban` joined
dalek ast: 97213bc | sorear++ | S03-junctions/ (2 files):
Fudge and defossilize S03-junctions/misc
 06:16
ecza: 5168d25 | sorear++ | / (4 files):
~~Junction, non-autothready not, ne as !eq
 06:17
ecza: d22c6cb | sorear++ | lib/ (2 files):
Tweaks to Test to handle junctions better, fake &WHAT and &HOW
ecza: b037fe8 | sorear++ | t/spectest.data:
Add S03-junctions to spectest.data
06:26 yinyin joined 06:27 fhelmberger joined, Alax left 06:30 jfried joined 06:32 amkrankruleuen left 06:37 amkrankruleuen joined 06:43 amkrankruleuen left 06:44 amkrankruleuen joined 06:53 starcoder2 left 06:55 starcoder2 joined 06:58 daniel-s_ left 07:00 amkrankruleuen left 07:04 amkrankruleuen joined 07:09 envi left 07:11 envi joined 07:12 mj41 joined 07:14 kaare__ joined, Eevee joined 07:25 amkrankruleuen left 07:26 amkrankruleuen joined 07:34 hanekomu joined 07:41 amkrankruleuen left 07:42 flussence left, amkrankruleuen joined 07:44 flussence joined 07:48 kaare__ left 07:49 amkrankruleuen left, amkrankruleuen joined
dalek ecza: 8f41f1f | sorear++ | / (2 files):
Micro-optimization to builtin type checking
 07:54
ecza: b6269f8 | sorear++ | / (6 files):
A new home for the random small scripts I used for this
ecza: 0e96ea2 | sorear++ | / (3 files):
Fix Mu ~~ $smartmatch
07:59 silug joined
moritz sorear: return.t failed 6 tests 08:02
t/spec/S04-statements/return.t
t/spec/S32-hash/pairs.t dies 08:04
t/spec/S03-operators/repeat.t Failed test: 28 08:05
moritz does discontinuous integration
sorear ...TEST_JOBS?
why out of order?
08:05 Su-Shee_ is now known as Su-Sh, Su-Sh is now known as Su-Shee
moritz TEST_JOBS=2 08:06
no, I just missed repeat.t while it was scrolling by, but saw it in the summary
sorear return is doing something odd.
moritz niecza: say +"3a" 08:07
p6eval niecza v6-46-gb037fe8: OUTPUT«Unhandled exception: System.FormatException: Unknown char: a␤ at System.Double.Parse (System.String s, NumberStyles style, IFormatProvider provider) [0x00000] in <filename unknown>:0 ␤ at System.Double.Parse (System.String s, IFormatProvider provider) [0x00000] in
..<filename unkno…
moritz I'm pretty sure that should return 3, warn
sorear I'm dreading having to write a number parser, but it seems inevitable
moritz we'll need one for rakudo too
sorear c.f. the recent debacle where parsing just the right string makes Java's number parser hang 08:08
moritz just so that +$string returns the right type
sorear which went unnoticed for 18 years because number parsers are so hard to 1. visually scan for bugs 2. test for bugs
I mean floating point base conversion
integer base conversion isn't that bad at all 08:09
Oh. 08:10
I fixed eval to pass control exceptions
eval_dies_ok 'return' causes a return from &eval 08:11
if &eval was a bare block, it would still cause a return from &eval_dies_ok
08:11 silug left
sorear wonders whether to back out this change, or what 08:12
or, hehehehehe, I could add a CONTROL { } to eval_dies_ok :) 08:13
or, I could change the test such that it isn't checking eval behavior
08:15 amkrankruleuen left 08:18 MayDaniel joined, MayDaniel left
felher l/go laber 08:19
08:21 amkrankruleuen joined 08:24 aindilis left 08:40 frettled_ joined 08:43 frettled left
dalek ecza: 2742d7d | sorear++ | / (3 files):
Block parameters default to Mu, method invocants to ::?CLASS
 08:44
sorear moritz: opinions on the best return.t action?
08:45 awoodland left
moritz sorear: currently I think eval should imply CONTROL, just like it implies CATCH 08:45
sorear: though in the long run I'd prefer it if it implied none of the two
sorear right now niecza's eval implies neither 08:46
so I guess that's +1 for
"eval_dies_ok does CONTROL"
moritz sounds sane-ish 08:47
08:54 frettled joined
sorear rakudo: my ($x) = (^*); say $x; 08:56
p6eval rakudo 048573: OUTPUT«Cannot take numeric value for object of type Whatever␤ in 'Any::Numeric' at line 1496:CORE.setting␤ in 'prefix:<^>' at line 7674:CORE.setting␤ in main program body at line 22:/tmp/dlq0LlewTF␤»
09:01 frettled_ left 09:04 uniejo joined
dalek ecza: d124518 | sorear++ | t/spectest.data:
Remove S32-hash/pairs; it relies on Pair cmp Pair, which was never really supported
 09:07
ecza: e9691cc | sorear++ | lib/ (2 files):
xx and Whatever/Array, fudging eval_*_ok to choke on control exceptions
ecza: 1131cc6 | sorear++ | / (2 files):
Honor parameter types in inline blocks
 09:24
ast: 625cd92 | sorear++ | S03-junctions/ (2 files):
Unfudge for block Mu parameters
09:31 woosley left 09:35 pernatiy joined
dalek ecza: 0da903b | sorear++ | lib/ (2 files):
Implement junction arguments to subscripters
 09:37
ast: 8315f60 | sorear++ | S03-junctions/ (2 files):
Unfudge and fix tests for junction indexes
sorear niecza> my %hash = :foo, :quux; ?(%hash{any <foo bar>}:exists) 09:39
Bool::True
niecza> my %hash = :foo, :quux; ?(%hash{all <foo bar>}:exists)
Bool::False
\o/
sorear -> sleep (maybe?)
moritz sorear++ 09:47
09:53 daniel-s joined 10:01 MayDaniel joined, math1 joined
math1 Bonjour 10:01
10:02 JimmyZ joined 10:04 daniel-s left 10:06 mtk left 10:11 _jaldhar joined 10:15 mtk joined
math1 Il y a t'il quelqu'un qui pourrait me renseigner sur un problème de compilation de module PERL via CPAN ? 10:15
PHP::Interpreter
tadzik I thought this is the line from the Call of Cthulu 10:16
flussence rough translation: "this won't install, halp?" 10:17
math1 Hi i have a problem to compile PHP::Interpreter with cpan. Does anyone can help me ? 10:18
flussence math1: you might get better answers on a perl5 channel, but is there an error message? 10:21
10:42 MayDaniel left 10:47 daxim joined
math1 yep 10:47
pastebin.com/uey5i3dk 10:49
flussence the comments in that test seem to suggest they know it'll fail, so it should be safe to force-install it 10:53
10:54 f00li5h joined, f00li5h left, f00li5h joined
math1 I dit it. I "make installed" it 10:54
It works partly but not completly 10:55
I mean I compiled it
I instanciate a php class in my perl script 10:56
It works… In fact It's about another problem… lol 10:57
I have a segmentation fault when I use PHP::Interpreter in my perl script
flussence hm, that module hasn't been updated since 2008. Maybe it just doesn't work with newer PHP versions. 10:59
math1 I pastebin.com/xv9PFa8k 11:00
I found many problem about this...
www.perlmonks.org/?node_id=612502
I followed this topic too 11:01
The solution is on the last post
11:08 wamba joined 11:09 math1 left 11:10 pernatiy left 11:11 silug joined 11:12 wamba left 11:14 donri joined, pernatiy joined 11:15 wamba joined 11:21 Chillance joined 11:22 silug left 11:24 koban` left 11:43 Vlavv` joined 11:44 Vlavv left 11:49 pernatiy left 11:50 molaf_ joined 11:53 molaf left 12:00 yinyin left 12:01 Vlavv` left 12:03 kaare__ joined 12:13 Vlavv` joined 12:15 _jaldhar left 12:18 satyavvd left 12:32 wknight8111 joined 12:33 bluescreen10 joined 12:37 bluescreen10 left 12:42 Chillance left 12:46 thou joined 12:53 bluescreen10 joined 12:54 Chillance joined 13:02 Holy_Cow joined 13:18 isBEKaml joined 13:22 koban` joined 13:23 ymasory joined 13:39 birdwindupbird left 14:01 wamba left 14:02 spq joined, hudnix joined 14:11 pernatiy joined 14:21 hercynium joined, silug joined 14:28 DarthGandalf left, MayDaniel joined 14:32 MayDaniel left 14:34 daxim left 14:38 jfried left 14:39 jfried joined 14:40 silug left 14:43 wamba joined 14:46 kaare__ left 14:55 uniejo_ joined 14:57 uniejo_ left 14:58 koban` left 15:02 slavik joined 15:09 isBEKaml left
sorear good * #perl6 15:22
15:22 bluescreen10 left 15:23 bluescreen10 joined
TimToady o/ 15:28
sjn o7 15:33
perplexa o>
15:35 woosley joined, woosley left 15:37 Holy_Cow left 15:40 MayDaniel joined 15:48 MayDaniel left 15:50 bluescreen10 left
sorear bleh 15:54
after removing my old Parrot/Rakudo install, a clean build of Rakudo master also fails
strace confirms that except for system libraries it's not picking up anything outside the Rakudo root 15:55
what else could it be?
15:56 nymacro left
sorear gdb ./perl6 / run -> got SIGSEGV in Parrot_Object_getprop from sort_candidates 15:56
jnthn :/ 15:58
15:58 DarthGandalf joined
jnthn Hard to guess. 15:58
sorear: Low memory environment?
Wonder if it triggers GC more than most people get it and tickles a missing write barrier or something. :s 15:59
16:00 nymacro joined
sorear jnthn: yeah 98 mwords 16:01
96
jnthn: otoh, nom builds successfully 16:02
the master segfault occurs in startup, before it even gets to the repl
however, the stage1 starts successfully, as evidenced by being able to compile core.pm
jnthn sorear: Did nom build in a sane-ish amount of time? 16:04
sorear stefan@stefans:~/dl/rakudo-ng$ parrot_install/bin/parrot -g ms2 ./perl6.pbc > say 2 + 2
4
jnthn: yes
>
looks like def. a GC issue
shall I clean again and try with --gen-parrot --gen-parrot-option=(what do I put here to enable line numbers)? 16:05
jnthn: even master built in a sane amount of time
core.pm took 320 out of 384 mb and didn't hit swap
took just under 10 mins iirc 16:06
not like the 10 hour builds that drove me away from Rakudo in the first place... I'm impressed by everyone's work
jnthn sorear: nom should use a good bit less memory for Cusrsor/Match nodes. We need PAST switched to 6model to win a bunch more on build memory. 16:07
sorear: Sorry, not sure what flag you're looking for...
Though Parrot's Configure.pl --help should tell you.
sorear the question was more "shall I" than "how do I" 16:08
jnthn You could clean and try again, or I could hunt for a missing write barrier...
sorear I can look for how on my own, but won't bother if it won't help
jnthn If it's in sort_candidates that gives me a good guess.
16:08 am0c joined
sorear eh, probably second-level damage 16:09
jnthn Maybe, but it's also an easy area of the code base to believe would have a missing WB.
sorear "Invalid runcore gcdebug requested" 16:10
WTF #parrot WHY
that was our only prayer
jnthn sorear: Does -G help? 16:11
(that disables GC)
oh, wait...did you do that already...
no, you did ms2
Ah, that makes it even more likely to be missing WB.
sorear -G also fixes it
parrot ought to have an option to print the size in bytes of the nursery 16:13
and another option to set it explicitly
so you can reproduce this
16:15 jedai left, jedai joined
sorear help, I'm getting back my old cynicism and mistrust about #parrot 16:15
16:18 nymacro left
sorear rakudo: each(1,2,3).say 16:20
p6eval rakudo 048573: OUTPUT«Could not find sub &each␤ in main program body at line 22:/tmp/hvtZU5YRd0␤»
sorear jnthn: what do you think about each junctions?
jnthn I...didn't think too much about them. :) 16:21
sorear I don't like em
jnthn They're newer than when I last did any work on Junctions in Rakudo.
16:22 kaare__ joined
sorear each junctions are supposed to wrap a lazy list instead of an RPA 16:22
which means that most of niecza's junction logic is duplicated
jnthn They promise ordering also? 16:23
16:23 bluescreen10 joined
sorear yes 16:23
16:23 mj41 left
jnthn meh, I barely consider that a junction 16:23
grrr...whenre's it spec'd... 16:24
Junctions section in S03 seems not to mention them.
16:24 JimmyZ left
jnthn S02 mentions an Each type 16:25
sorear S09:994
jnthn (e.g. it's disctinct from Junction in a typeful way)
sorear S09:1066
16:25 DarthGandalf left
sorear oh yay, "conjectural" 16:26
that means I can kill it with fire if I don't want it
jnthn [This section is considered conjectural.]
:)
I can see the point of it.
But yeah, it looks somewhat distinct from junctions.
sorear meh 16:27
there's little practical difference between if (kind == EACH) and if (type == Each), either way it involves massive code duplication in autothreading
16:31 DarthGandalf joined
jnthn bbiab 16:37
16:43 flussence left 16:44 flussence joined
dalek p: 473c265 | tadzik++ | src/HLL/Compiler.pm:
Quit REPL and print a newline on ^D

This also works in Rakudo/nom, which makes RT #70297 fixed in nom.		 16:45
16:46 silug joined
sorear rakudo: say (5 cmp "5") 16:48
p6eval rakudo 048573: OUTPUT«0␤»
sorear rakudo: say (5 eqv "5")
p6eval rakudo 048573: OUTPUT«Bool::False␤»
pmichaud rakudo : say (5 cmp "5") 16:49
rakudo: say (5 cmp 5)
p6eval rakudo 048573: OUTPUT«0␤»
pmichaud hmmmm
sorear rakudo: say (5 cmp 5.0) 16:50
p6eval rakudo 048573: OUTPUT«0␤»
pmichaud ah, we're not converting to bool
sorear rakudo: say (5 cmp 6.0)
p6eval rakudo 048573: OUTPUT«-1␤»
sorear rakudo: say (5 cmp 4.0)
p6eval rakudo 048573: OUTPUT«1␤»
pmichaud we're just returning the result of cmp_i_i_i directly
easy fix 16:51
sorear I think this behavior is in violation of S03:3308 and S03:3296
discuss
16:51 geekosaur left
pmichaud which behavior are you referring to? 16:51
16:52 allbery_b joined
sorear two values never compare as Order::Same / 0 unless they are exactly the same type 16:52
16:52 cdarroch joined, cdarroch left, cdarroch joined
pmichaud I've had longish discussions about problems with cmp in the past 16:53
what should they return if not same type? 16:54
(i.e., this is an unresolved section of the spec)
(or the spec is showing fossils) 16:55
sorear last I heard/dreamt, it was $a.WHAT.perl cmp $b.WHAT.perl 16:57
pmichaud that doesn't seem plausible
16:57 wanradt_ joined
pmichaud more specifically, what result are you expecting for 5 cmp "5" ? 16:58
(last I heard/dreamt, &infix:<cmp> "fell back" to string semantics or something like that.) 16:59
(but I could've missed some later discussion on the topic)
sorear pmichaud: Less 17:04
pmichaud sorear: you mean Order::Increase? (less than zero?) 17:05
sorear uh
rakudo: say (1 cmp 2).perl
p6eval rakudo 048573: OUTPUT«-1␤»
sorear rakudo: say ("a" leg "b").perl
p6eval rakudo 048573: OUTPUT«-1␤»
jnthn Doesn't cmp return 0 for same, -1 for less, 1 for greater?
sorear some .perl
jnthn: no 17:06
jnthn Ah
Well
It seems it does in Rakudo ;)
sorear yes, Order::Increase, pmichaud
jnthn Oh...unless it's an enum that's numifying in an entertaining way :)
pmichaud why would 5 cmp "5" be Order::Increase, then ?
17:06 awoodland joined
sorear because "Int" lt "Str" 17:07
pmichaud huh?
sorear cmp isn't supposed to make sense, it's supposed to produce a total order. c.f. cmp(Complex,Complex)
Complex isn't an ordered field and there's no sensible <=> for it 17:08
pmichaud you mean that my @a = (5, "4"); say sort @a should result in a list with 5 and "4" ?
(in that order)
sorear yes.
pmichaud I'd love to see that discussion. :-)
sorear if the user has a list of mixed values and wants them all to be sorted as numbers, they should :by(&[<=>])
moritz iirc cmp should fall back to infix:<leg> (string comparison) if the types don't match 17:09
but of the values, not of their types
sorear the main problem I have with my interpretation is that Ranges don't work
pmichaud what moritz++ said is what I remember
sorear 1 .. 2.5 # the smallest Num is larger than every Int. Oops! 17:10
ok.
do either of you remember where you got this from?
moritz here
sorear also, what about 1 .. 40.5 ? 17:11
5 gt 40.5
pmichaud ranges use the type of the lhs, iirc
so since 1 is numeric, it uses a numeric comparison always
moritz 1 and 40.5 are both Numeric 17:12
pmichaud I think these were discussed/decided/made tentative on #perl6, but the spec might not have been updated
moritz: they're both numeric but not necessarily same type
moritz it's in the spectests, more or less
pmichaud: yes, but cmp should know how to compare Real numbers, in the very least
pmichaud Real probably makes sense
and indeed, Rakudo has &infix:<cmp>(Real, Real) defined. 17:13
TimToady it's actually an impossible problem to make consistently consistent
sorear \o/ TimToady
TimToady any given type may define an order that is inconsistent with Real v Real
pmichaud I think sorear++ is just looking for a spec update or fossil extraction 17:14
since the spec still talks about cmp in terms of eqv semantics
TimToady the sort on type first is the only consistent approach, but doesn't dwym at all
17:14 fhelmberger left
sorear yeah, I want ===, eqv, and cmp done _right_ before v7 17:15
TimToady the whole thing starts to fuzz over into natural sorting as soon as you start to twim
*d
pmichaud sorear: no problem, if you're willing to accept that "right" might change over time :-P
sorear so I want to understand the rules...
so you say "eqv" and "cmp" are no longer related? 17:16
moritz things that are eqv should also be cmpable
TimToady please don't treat me as an oracle in this case :) 17:17
moritz erm, if $a eqv $b then ($a cmp $b) == Order::Same
pmichaud I tend to think that eqv and cmp occupy different spaces
$a eqv $b implies $a cmp $b == Order::Same, but if $a !eqv $b we can't say anything about $a cmp $b 17:18
not based solely on eqv-ness, at any rate.
moritz right 17:19
pmichaud TimToady: you're not the oracle here -- you're The Decider. :-P
(in this case)
TimToady decoupling cmp makes sense from a usability perspective; you want all lists to be sortable, even if they aren't :)
sorear pmichaud: S03 says that for immutable types, === and eqv coincide. Agree? 17:20
TimToady hence, we impose orderings on Complex and Pair
even though they naturally don't work that way
17:20 lumi joined 17:22 masak joined
masak oh hai 17:22
sorear hello masak
TimToady so I think we say that cmp works like eqv only when the types are the same
tadzik hai masak
pmichaud sorear: I agree, unless it's wrong, in which case I disagree. :-) :-)
moritz sorear: sounds right
sorear masak: you just landed in the middle of an Exciting Spec Discussion.
masak ooh! 17:23
TimToady
.oO(re-exciting re-spec re-discussion)
masak re-backlogs
17:24 wamba left, Mowah left
sorear so, to summarize what I've understood so far: 17:24
=== : Stateful indistinguishability
17:24 wamba joined
sorear eqv : Snapshot indistinguishability 17:24
TimToady and if we impose Numeric//Stringy generic semantics on cmp of different types, we strong suggest that Numeric and Stringy types respect the same ordering within their types
masak sorear: that is my understanding as well.
sorear cmp : Within each type, an order which generates eqv, but compares different types in a DWIM way
TimToady *strongly
however, we can't do that for both Numeric and Stringy simultaneously and stay consistent 17:26
"5" cmp "a" vs "5" cmp 4 etc. 17:27
vs "5" cmp "40"
so either we unify the Numeric/Stringy into natural sorting (which probably is inconsistent with the individual types eqv semantics) or we just suffer out-of-orderness with "5" cmp 5.0 17:28
that's the fork 17:29
17:29 Mowah joined
colomon thought it was always a given that "5" cmp 5.0 would not be a numeric comparison.... 17:29
TimToady we can pretty easily write the non-natural orderings, which tends to argue that cmp should default to natural 17:30
pmichaud 05:22 <quietfanatic> What bothers me isn't the ambiguity in distinction, it's that there's two possible distinctions that aren't the same.
05:23 <TimToady> if things like that bother you, either you should or shouldn't be a language designer. :)
(sorry, had to repost that useful quote. )
TimToady lol, really 17:31
pmichaud colomon: it's not ever been that way in the spec. :-)
colomon "Some do, and some don't." (Reading too much Pooh this year.)
TimToady compounding the difficulty is that some sort algorithms blow up if you give circular orderings 17:32
sorear heh, or worse... 17:33
I've seen qsort segfault before
colomon errr... seems like the spec definition of cmp is just plain broken? it doesn't seem to define an ordering at all.
sorear colomon: that's why I started this discussion
colomon sorear++
pmichaud (re-started :-) 17:34
TimToady the only solution consistent with letting types define their own cmp is to order based on types first 17:35
if we drive cmp in that direction, then we need to make it drop-dead easy to get at the natural ordering differently 17:36
and maybe make sort default to the natural ordering
colomon TimToady: I know that's been said above and here again, but I didn't understand the logic either time. Can you try one more time for the very slow amongst us/
?
TimToady but individual types may not override the natural ordering
PerlJam thinks that if there really were a "natural ordering" we wouldn't be having this conversation ;) 17:37
pmichaud colomon: I think TimToady is saying we need two forms of ordering, and that sort uses "natural ordering" (which isn't defined by &infix:<cmp>)
note that quite a bit uses &infix:<cmp> besides, sort, also :-) 17:38
TimToady if you create fallbacks between types that allow mixing of values in the result, then you automatically get inconsistencies if the two types have different ideas on how to sort values that could be construed either way
sorear does natural ordering need to mix stuff such that 4 before "5" before 6 ? 17:39
TimToady so suppose we have a Real generic fallback that works like you might expect, and we introduce a MyInt type that has cmp sorting in reverse order, to be perverse 17:40
two MyInt types will sort in the opposite order than two equivalent values that go through the Real order
colomon ooooo.... and that can create a situation where A is less than B, and B is less than C, and C is less than A? 17:41
sorear if not, then I propose sub cmp($a,$b) { ($x ~~ Real ?? "Real" !! $x.WHAT) leg ($y ~~ Real ?? "Real" !! $y.WHAT) || ($x ~~ Real ?? $x <=> $y !! $x.cmp($y)) }
TimToady colomon: which is why I talked about circularities blowing up sort algos
sorear along with #`(Mu) method cmp($other) { ~self leg ~$other } 17:42
pmichaud note that Parrot allows circularities already, with its broken 'cmp' semantics.
TimToady sorear: natural sorting will intermix numbers and strings that start with numbers
I believe natural sorting can be defined consistently, but it's not trivial, and must enforce its own ideas on the individual types 17:43
17:43 zby_home_ joined
TimToady s/consistently/consistently for all types that map to builtin types/ 17:43
pmichaud "10", 2, "5" in Parrot end up with "10" cmp 2, 2 cmp "5", and "5" cmp "10" all being > 0 :-) 17:44
er, wait, that's not right 17:45
anyway, it's possible to achieve in Parrot :-)
jnthn it's possible to fail to achieve in Parrot too :P 17:46
sorear so it's like cmping over .Str.comb(/ <STD::number> || . /) ?
pmichaud parrot takes the type of the first operand as the type of comparison
colomon almost seems like what we're arguing is that there should be no cmp at all...
pmichaud so "3" cmp 20 and 20 cmp "3" are both >0 17:47
colomon just like you can no longer as for the length of a string, but have to ask for the number of chars or graphs or bytes or whatever 17:49
PerlJam colomon: only if there's no object system to go with no cmp.
flussence
.oO( maybe there should be a .COMPARES, like .ACCEPTS? )
colomon no, my point is if you had to explicitly ask for what ordering to sort by, this problem goes away. 17:50
PerlJam colomon: useful defaults (DWIM) is way better than requiring the poor, unwashed masses to make a decision. A little more pain for the language designer but much happiness for the language users. (<insert star trek quote here>) 17:51
17:52 wamba left
colomon PerlJam: useful defaults imply that there is a sensible default. Other than just making leg the default, like in Perl 5, I'm no longer sure there is such a thing. 17:53
PerlJam colomon: or, put another way ... if we get rid of cmp, then we'll have to provide some standard comparators that are drop-dead easy to use (so that our users don't go crazy) and one of them will contain the comparison that we're talking about for cmp 17:54
pmichaud comparison*s*, I think.
PerlJam yes.
moritz PerlJam: cmp is *very* convient, because it means you can sort arrays of numbers with the default op
pmichaud ...except when you can't. :-) 17:55
moritz PerlJam: and at the same time you can sort arrays of strings with the default op
pmichaud: usually you can.
pmichaud: which why it's convenient
pmichaud but it is very nice to be able to do @list.sort(+*) to mean numeric sort
without having to also specify a :by
17:55 lumi left
moritz but still much more obscure than @list.osrt 17:55
*sort 17:56
PerlJam unless @list.sort does something unexpected :)
17:56 lumi joined
pmichaud (my point being that +* for numeric sort requiers that sort know how to automagically sort numerically when needed) 17:56
*requires
(as opposed to defaulting to always-leg semantics)
colomon I could maybe see making .sort actually examine the list to see if it is all numbers or number strings.
TimToady the two inconsistent things we expect from default sort: a) sort naturally, and b) sort fast 17:57
pmichaud afk, lunch 17:58
PerlJam The first is DWIM and should be huffmanly short, while the second thing falls under the rubric of "optimization" and so should be huffmanly long. Right? 17:59
17:59 thou left
colomon TimToady: okay, one more thing I'm not understanding in this discussion: why a distinction between cmp and "natural sort"? If you can come up with a consistent "natural sort", shouldn't that be how cmp is defined? 17:59
18:00 wanradt___ joined
moritz colomon: natural sort usually does much more than what we envision for cmp semanits 18:00
TimToady yes, we can, but then we can't also let types define their own cmp inconsistent with that
18:01 wanradt_ left
moritz ie compare the numbers in 'ab100' and 'ab99' by number 18:01
TimToady it's easy to specify leg or <=> semantics with ~* or +*, so I don't think either of those should be sort's default 18:02
18:02 impious joined
TimToady natural sorting tends to be a standalone set of remapping semantics, whereas cmp is a system of cmp operators 18:02
colomon except if cmp compares types first, then you cannot specify leg or <=> with ~* or +*, because there are Stringy and Numeric subtypes 18:03
TimToady we're confusing two different meanings of "cmp" here 18:04
it's not inconsistent with the type's defining their own cmp, since Stringy and Numeric both coerce to some most-general type and use that type's cmp 18:06
18:06 impious left
TimToady there's also the notion of a general cmp that is a collation order outside of types, and that's what we need to give a different name 18:06
18:06 am0c left
TimToady several different names, it seems 18:07
PerlJam So, when two different types meet in a comparator they're both coerced to a more general type? 18:09
TimToady pretty much have to be, or you end up with a combinatorial explosion 18:10
in the general case of collating sequences, they're remapped to a different string that you can sort stupidly
sorear sorry, distracted. 18:11
I'm back
PerlJam The "general cmp that is a collation order outside of types" could be called "col" :) 18:12
A col B (how does B collate with respect to A)
or vice versa, I'm not sure about transitivity here.
moritz when I see "col" I think "column", not "collate" 18:13
PerlJam moritz: that's just a matter of (re)education ;)
yath +1
18:13 aindilis joined
masak "col" could also mean "color". 18:13
I get that ambiguity ("color"/"column") a lot when programming board games.
tadzik or "collegaue", or however you write that 18:14
18:14 MayDaniel joined
moritz tadzik: cow-orker :-) 18:14
sorear TimToady: why do you think types need to be able to override cmp?
TimToady isd "increase, same, decrease"
are you asking why types should be allowed to define their own ordering? 18:15
PerlJam obscure enough that no one has any preconceived notions :)
colomon sorear: because you need to be able to provide a comparison operator for whatever weird type you have.
well, "need"
TimToady sorear: but your question tangles the two different meanings of cmp that we're trying to detangle
PerlJam I want to sort Circles, Triangles, Squares, and any other manner of polygon. 18:16
hence, I probably need to define my own cmp
TimToady (if that's the cmp that we end up calling cmp)
18:17 MayDaniel left
TimToady it's like trying to distinguish the Perl language's + operator that enforces numericity with a Python object's notion of what a + method does 18:17
PerlJam isd might work, though I tend to want to put decrease on the left for some reason, so .... dsi :) 18:18
TimToady here we can't just rely on the difference between single and multiple dispatch, as we can with +
or maybe we should, and comparisons on a type should go through single dispatch 18:19
to force the two arguments to be considered under the same type
or we disallow/discourage creation of cmp multis with differing arg types 18:20
or whatever we call the type's view of ordering, if not cmp
PerlJam: except decrease first would be >=< instead of <=>
the main problem with isd is that it's not as funny as "leg" :) 18:21
moritz isd sounds like some disorder :-)
PerlJam In my head dsi is a nice little graphic that looks like a cup, while yours looks like an upside down cup.
moritz the mixture between IRS and a disorder :-) 18:22
18:22 pyrimidine left
TimToady there's also bea, if we name the type's comparison based on before/eqv/after 18:23
and reserve cmp for the intertype ordering
PerlJam bea is my mother-in-law so ... -1 from me :)
TimToady you don't want to be ordered by your mother-in-law? :P 18:24
I think 'bea' for intra-type and 'cmp' for extra-type makes some sense to me 18:25
where the default cmp is the default sort order 18:26
which is probably not natural sorting
but something simple like type-first, then type's bea (which for compound types collates each element in turn)
and then we give a simple name to natural sorting 18:27
that doesn't use cmp
PerlJam feels rightish to me.
moritz thinks natural sorting belongs into a module, because there are so many ways to do it
TimToady thinks languages that define one by default are at an advantage in the user's mind 18:28
it would be trivial to override the built-in in p6; we designed it that way...
use order MyNaturalSort; order @list; 18:29
or just 'use MyNaturalSort' overrides order lexically 18:30
18:30 jferrero joined, Tedd1^ left
TimToady or just give it a different name and use it, to avoid confusion 18:31
order vs sort is probably good huffman, where order is doing a bit more work
otoh, order seems more like it ties directly to Order:: 18:33
so I wonder if cmp goes away entirely and we end up with $a ord $b for an ordering comparison 18:35
which both constructively and destructively interferes with prefix:<ord>
I don't like $a order $b because that's too heavy for the binary comparison, and visually confusing with an order listop
oh, wait, now I'm confusing the two different things. order would use ord, and sort would presumably use cmp, which is the dumb one 18:37
or maybe sort should use "$a srt $b" :)
sorear srt sounds like minmax to me 18:38
also like the Pentium FDIV algorithm
masak grmmmbl... minmax... :/
TimToady should go take a shower before he thinks too hard(er) 18:39
sorear this discussion is making me even more confused. I think. 18:41
TimToady I think in this case it indicates progress. :) 18:43
masak rakudo: say (1 minmax 2).perl 18:44
p6eval rakudo 048573: OUTPUT«1..2␤»
TimToady because when I see how we get confused here, it's indicative that neurotypicals will get even more confused at just these points, and we need to be carefuller
masak rakudo: say ([minmax] 4, 2, 7, 1, 3).perl 18:45
p6eval rakudo 048573: OUTPUT«1..7␤»
pmichaud back from lunch
masak rakudo: say (1 minmax (4 minmax 7)).perl 18:46
p6eval rakudo 048573: OUTPUT«1..7␤»
masak rakudo: say (5 minmax (2 minmax 18)).perl
p6eval rakudo 048573: OUTPUT«2..18␤»
masak rakudo: say ((5 minmax 18) minmax (2 minmax 11)).perl 18:48
p6eval rakudo 048573: OUTPUT«2..18␤»
masak remarkable.
sjohnson masak: !
masak sjohnson! \o/ 18:49
18:49 allbery_b is now known as geekosaur
masak sjohnson: how's life with Perl? 18:52
sjohnson pretty good
been doing quite a bit of p5 stuff last few weeks at work
one of them being a frontend for vim
which is making my life much easier and less headache prone 18:53
moritz thought vim already was a frontend
masak was just going to say that :)
sjohnson well, let me give you an example..
if you accidently pipe something to vim without doing the - thing
it causes a headache
also if you do: vim something.pl | grep something
(don't ask why i'd ever do it, it's a bad habit of using the up arrow in bash improperly!) 18:54
masak this sounds eerily familiar.
sjohnson well i prevented both those things from happening. also hate the msg that comes up when a swap file is edited, i'd rather it tell me who's editing it and not ask me anything, and if i'm editing it myself on another screen window, have it switch the screen window for me
jdhore1 sjohnson, I can haz link?
sjohnson jdhore1: sure 18:55
oh yeah, also another thing it does
is warn you with a question that you don't have write permission on a file
jdhore1: preparing for a git push 18:56
masak sjohnson: ah. I might've been thinking about pipe.vim: www.vim.org/scripts/script.php?script_id=2769
sjohnson: have you seen it?
sjohnson nope 18:57
not really sure i understasnd what this does *scratches head*
masak it lets you pipe things into vim. 18:58
sjohnson well.. i mean i do that already with using vim - 19:01
echo moose | vim -
you're probalby thinking of something more advanced, though
sjohnson doesn't use the advanced stuff
19:08 benabik left, benabik joined
masak no, I don't know offhand how pipe.vim is different from 'vim -'. 19:10
sorear right. unix sucks. 19:11
sjohnson jdhore1: just writing a bit of documentation, in case you actually decide to try this out
sorear the system is very low on DWIM-factor 19:12
jdhore1 sjohnson, I can read the code as documentation :P
sorear hello sjohnson.
sjohnson sorear: howdy!
sorear bah. 19:14
'my $i = -10_000_000; 1 while $i++'
this is basically a best case for the niecza optimizer... it's still 2.5x slower than 5.12
masak sorear: any idea why? 19:15
sorear this is an embarrasment...
perl5 is classically regarded as one of the slowest currently used interpreters, and I'm losing to it with a JIT
masak :P 19:16
sorear well, it's probably relevant that Perl 5 can run that loop without allocating 19:17
since the $x container is a Giant Union Of Doom that can hold unboxed ints 19:18
19:18 icwiener joined
sorear wonders if adding polymorphically unboxing containers to niecza would make any sense 19:18
sadly, this being a managed environent, I can't overlap pointer and integer fields
masak management is killing performance! :P 19:19
19:20 birdwindupbird joined 19:30 math1 joined 19:37 envi left 19:38 sftp joined
sorear so maybe Variable could be a structure with an integer type_tag, 64 bits of nonpointer storage, and a pointer 19:46
and it could directly hold Int, Num, Str, Bool, Any, and use the pointer to a P6any otherwise
sjohnson jdhore1: github.com/smujohnson/web1-home/bl...bin/vim.pl
sorear would be a real pain to program for C# though 19:47
I'd rather use heavily macroed C for this :)
sjohnson masak: also remembered a few other things i didn't like: vi md5sum some.file # won't edit md5sum
(another up-arrow mistake i commonly made)
also, vi some .file # if you accidently put a space, it will not whine when you try to quit that there's "more files to edit" anymore 19:48
dalek ecza: b10ed5a | sorear++ | src/ (2 files):
Cut an extraneous let out of postfix while. Little effect.
sjohnson (url script a few lines above if you want to peek at it)
masak sjohnson: we seem to have different command shell problems :) 19:49
19:51 lumi left 19:52 lumi joined
sorear probably even more relevant 19:53
19:53 kaare__ left
sorear 1 while $i++ # 5 indirect calls per iteration 19:54
sjohnson masak: my colleagues use the frontend at work, and say it's a godsend :) 19:56
if even 1 person gets some use out of it, it was all worth it. well, it was worth it to prevent headaches for myself, so that's good too.
20:02 NINA-2632 joined, NINA-2632 left
masak sjohnson: of course. I'm glad to hear you're building solutions to problems for you and your colleagues. 20:04
jnthn rakudo: my @a; for 1..5 -> $i { @a.push(sub { $i }) }; .() for @a 20:13
p6eval rakudo 048573: ( no output )
jnthn rakudo: my @a; for 1..5 -> $i { @a.push(sub { say $i }) }; .() for @a
TimToady sorear: Lisp implementations often bank on the pointer never using some of the low bits or some of the high bits.
p6eval rakudo 048573: OUTPUT«1␤2␤3␤4␤5␤»
jnthn rakudo: my @a; for 1..5 -> $i { @a.push(sub foo { say $i }) }; .() for @a
p6eval rakudo 048573: OUTPUT«1␤2␤3␤4␤5␤» 20:14
jnthn rakudo: my @a; for 1..5 -> $i { @a.push(method foo { say $i }) }; .() for @a
p6eval rakudo 048573: OUTPUT«Useless declaration of has-scoped Method in a module; add our or my to install it in the lexpad or namespace␤Subfoo (subid "82_1307132053.89459") has no $!llsig and no $!lazysig␤ in <anon> at line 7674:CORE.setting␤ in main program body at line 1␤»
jnthn wow :)
TimToady rakudo: my @a; for 1..5 -> $i { @a.push(anon method foo { say $i }) }; .() for @a 20:15
p6eval rakudo 048573: OUTPUT«Subfoo (subid "82_1307132137.63473") has no $!llsig and no $!lazysig␤ in <anon> at line 7674:CORE.setting␤ in main program body at line 1␤»
TimToady rakudo: my @a; for 1..5 -> $i { @a.push(anon method foo { say $i }) }; .($_) for @a 20:16
p6eval rakudo 048573: OUTPUT«Subfoo (subid "82_1307132201.72778") has no $!llsig and no $!lazysig␤ in <anon> at line 7674:CORE.setting␤ in main program body at line 1␤»
jnthn Not going to bother investigating; just been re-working that chunk in nom and thought "huh, something looks adrift here..." :) 20:17
sorear jnthn: any progress on the WB? 20:19
20:21 bluescreen10 left
jnthn sorear: Sorry, no - am trying to sort some nom stuff out ATM> 20:22
sorear has been toying with the idea of a c5niecza with OpTARG and CvPAD and SvGMAGICAL and all the other stuff that C# makes so hard 20:24
20:25 mj41 joined
masak jnthn: hm, is it OK with you if I submit that one to RT? 20:27
20:29 wknight8111 left
jnthn masak: Why not? :) 20:29
masak submits rakudobug 20:30
jnthn masak: It'll make sure it gets a test that way :)
And...on step closer to 2000 :P
*one
20:30 lumi left
masak ;) 20:30
20:31 lumi joined 20:32 jdv79 left 20:33 jdv79 joined
jnthn perl6: say 1; { $^a }; say 2; 20:35
p6eval rakudo 048573: OUTPUT«1␤Not enough positional parameters passed; got 0 but expected 1␤ in main program body at line 2:/tmp/p5C2qefKMq␤»
..niecza v6-55-gb10ed5a: OUTPUT«1␤Unhandled exception: No value in mainline available for parameter $a␤ at /tmp/x9QQ3TLHjU line 1 (MAIN mainline @ 4) ␤ at /home/p6eval/niecza/lib/CORE.setting line 1565 (CORE C715_ANON @ 2) ␤ at /home/p6eval/niecza/lib/CORE.setting line 1566 (CORE module-CORE @
..53) ␤ at /home/…
..pugs: OUTPUT«1␤*** Blocks with implicit params cannot occur at statement level␤ at /tmp/j79Hm4u9Pu line 1, column 1␤»
TimToady std: say 1; { $^a }; say 2;
p6eval std 4b1b100: OUTPUT«ok 00:01 114m␤»
TimToady hmm, supposed to complain about that
jnthn TimToady: It's illegal? 20:36
That is, compile-time illegal?
TimToady seems as though it ought to be
jnthn +1
TimToady and STD has code that used to catch that...
20:36 bluescreen10 joined
jnthn std: sub foo { { $^a } } 20:37
p6eval std 4b1b100: OUTPUT«ok 00:01 113m␤»
jnthn How about that one?
20:37 ymasory left
TimToady that too 20:37
can't return a bare block, it's taken as inline 20:38
jnthn OK, that makes my life easier. :)
I'll let you fix the error handling in STD and then just nab that ;) 20:40
sorear what he said.
jnthn: "2000"? 20:41
jnthn: also, what bit of code are you reworking?
20:42 jdv79 left
jnthn sorear: Just about everything to do with blocks 20:42
Decided I'd rather make sure closures are right early. :) 20:43
Rakudo and closures have had some...history.
sorear How much memory does a nom sub use?
I mean a Sub object, cloned
tadzik hello hello 20:45
sorear hello tad zik
jnthn sorear: Dunno, but I'm pretty sure that the Parrot sub PMC itself dominates 20:46
sorear: Should be at least 1 PMC lighter than pre-nom ones
Which means at least 2 memory allocations lighter.
tadzik what sense does a method make if it's not inside a class?
sorear so nom still uses Parrot Sub PMCs for closures?
jnthn sorear: Yes. Don't actually have a choice there, so far as I can see. 20:47
sorear tadzik: lets you use 'self' to refer to the first argument
tadzik funky
jnthn sorear: Probably some chance of getting Parrot subs to take a weight loss program some day though... :)
Anyway, sub = 1 P6opaque + 1 Parrot sub PMC 20:48
tadzik perl6: method foo($a, $b) { say $self, $a, $b }; foo('bar', 'asd')
p6eval niecza v6-55-gb10ed5a: OUTPUT«===SORRY!===␤␤Variable $self is not predeclared at /tmp/eumny8oRbn line 1:␤------> method foo($a, $b) { say ⏏$self, $a, $b }; foo('bar', 'asd')␤␤Undeclared routine:␤ 'foo' used at line 1␤␤Potential difficulties:␤ 'method'
..declaration outside of…
..rakudo 048573: OUTPUT«Useless declaration of has-scoped Method in a module; add our or my to install it in the lexpad or namespace␤===SORRY!===␤Symbol '$self' not predeclared in foo (/tmp/OC8vICwrFA:22)␤»
..pugs: OUTPUT«*** ␤ Unexpected ","␤ expecting "::"␤ Variable "$self" requires predeclaration or explicit package name␤ at /tmp/4yPZLgXZnK line 1, column 31␤»
jnthn self, not $self 20:49
tadzik bam! bam! bam!
oh, right
perl6: method foo($a, $b) { say self, $a, $b }; foo('bar', 'asd')
p6eval pugs: OUTPUT«*** No compatible multi variant found: "&foo"␤ at /tmp/hRuV1rzFEI line 1, column 42 - line 2, column 1␤»
..niecza v6-55-gb10ed5a: OUTPUT«===SORRY!===␤␤Undeclared routine:␤ 'foo' used at line 1␤␤Potential difficulties:␤ 'method' declaration outside of class at /tmp/v_0uMKw5Mn line 1:␤------> method foo($a, $b) ⏏{ say self, $a, $b }; foo('bar',
..'asd')␤␤Unhandled exception: Check f…
..rakudo 048573: OUTPUT«Useless declaration of has-scoped Method in a module; add our or my to install it in the lexpad or namespace␤Could not find sub &foo␤ in main program body at line 22:/tmp/N1pIP_Tc2L␤»
20:49 jdv79 joined
jnthn tadzik: also need my :) 20:49
tadzik jnthn: myself? :P
moritz perl6: my $m = method foo($a) { say self, $a }; $m(3, 5)
tadzik perl6: my method foo($a, $b) { say self, $a, $b }; foo('bar', 'asd')
p6eval niecza v6-55-gb10ed5a: OUTPUT«Potential difficulties:␤ 'method' declaration outside of class at /tmp/awbr8GzSO3 line 1:␤------> my $m = method foo($a) ⏏{ say self, $a }; $m(3, 5)␤␤Unhandled exception: trying to dereference null␤ at /home/p6eval/niecza/boot/lib/CORE.setting line
..469 (CORE di…
..pugs: OUTPUT«*** ␤ Unexpected "$m"␤ expecting "=", "::", context, ":" or "("␤ Variable "$a" requires predeclaration or explicit package name␤ at /tmp/lOPUlY5UZ_ line 1, column 4␤»
..rakudo 048573: OUTPUT«Useless declaration of has-scoped Method in a module; add our or my to install it in the lexpad or namespace␤Subfoo (subid "78_1307134187.44165") has no $!llsig and no $!lazysig␤ in main program body at line 7674:CORE.setting␤»
rakudo 048573: OUTPUT«Not enough positional parameters passed; got 2 but expected 3␤ in 'foo' at line 2:/tmp/ikX7Hs5SYx␤ in main program body at line 22:/tmp/ikX7Hs5SYx␤» 20:50
..pugs: OUTPUT«*** No compatible multi variant found: "&foo"␤ at /tmp/bBl_PL3Tkc line 1, column 45 - line 2, column 1␤»
..niecza v6-55-gb10ed5a: OUTPUT«===SORRY!===␤␤Undeclared routine:␤ 'foo' used at line 1␤␤Unhandled exception: Check failed␤␤ at /home/p6eval/niecza/boot/lib/CORE.setting line 469 (CORE die @ 2) ␤ at /home/p6eval/niecza/src/STD.pm6 line 1146 (STD P6.comp_unit @ 31) ␤ at
../home/p6eval/niecza/src…
tadzik what a mess did I make
moritz "Unhandled exception: trying to dereference null"
looks like a niecza bug :-)
ok, all of them are buggy 20:51
tylercurtis perl6: my method foo($a, $b) { say self, $a, $b }; foo('bar, 'baz', 'quux'); 20:52
20:52 lumi left
p6eval rakudo 048573: OUTPUT«===SORRY!===␤Unable to parse postcircumfix:sym<( )>, couldn't find final ')' at line 22␤» 20:52
..pugs: OUTPUT«*** ␤ Unexpected "baz"␤ expecting term postfix, operator, ":" or ")"␤ at /tmp/YgWt6VxESy line 1, column 56␤»
..niecza v6-55-gb10ed5a: OUTPUT«===SORRY!===␤␤Any()Unable to parse argument list at /tmp/mz6nJHblGH line 1:␤------> d foo($a, $b) { say self, $a, $b }; foo(⏏'bar, 'baz', 'quux');␤Couldn't find final ')'; gave up at /tmp/mz6nJHblGH line 1:␤------> a, $b) {
..say self, $a, $b…
tylercurtis perl6: my method foo($a, $b) { say self, $a, $b }; foo('bar', 'baz', 'quux');
p6eval pugs: OUTPUT«pugs: Missing invocant parameters in '&foo': 0 received, 1 missing␤»
..rakudo 048573: OUTPUT«barbazquux␤»
..niecza v6-55-gb10ed5a: OUTPUT«===SORRY!===␤␤Undeclared routine:␤ 'foo' used at line 1␤␤Unhandled exception: Check failed␤␤ at /home/p6eval/niecza/boot/lib/CORE.setting line 469 (CORE die @ 2) ␤ at /home/p6eval/niecza/src/STD.pm6 line 1146 (STD P6.comp_unit @ 31) ␤ at
../home/p6eval/niecza/src…
20:52 molaf_ left 20:53 lumi joined 20:54 tadam joined
moritz "my mthod" doesn't make any sense to me 20:54
*method even
TimToady it prevents it from being a "has method" 20:55
jnthn moritz: my just means install it in the lexpad.
TimToady it's just a function that happens to treat its first arg as an invocant
20:56 mj41 left
moritz rakudo: my $x = my method () { say self }; $x(3) 20:56
p6eval rakudo 048573: OUTPUT«3␤»
moritz \o/ 20:57
sjohnson masak: yep, perl has saved me from more headaches than advil or tylenhol
TimToady rakudo: my $x = my method () { say self }; 3.$x
p6eval rakudo 048573: OUTPUT«3␤»
masak rakudo: my $x = my method () { say self.WHAT }; $x($x) 20:58
p6eval rakudo 048573: OUTPUT«Method()␤»
21:01 birdwindupbird left
math1 Hi i have a problem to compile PHP::Interpreter with cpan. Does anyone can help me ? 21:01
tadzik math1: this is a Perl 6 channel, you probably need help with Perl 5. Try #perl-help on irc.perl.org 21:02
21:02 cooper|ipad joined
math1 Alright. Ok man 21:02
sjohnson math1: or #perl on freenode 21:03
math1 cheers
tadzik have fun 21:04
21:06 mberends left
TimToady the problem with recognizing statement-level { $^a } as bogus is that { $^a }(1) is not bogus. 21:06
the old error message outlawed the latter, and was worked around by changing <block> to <pblock> 21:07
so if we do recognize it at compile time, it'll have to be recognized at statement reduction time
not in add_placeholder, where the old code is 21:08
you can still trigger the old code though 21:09
std: loop { $^a }
p6eval std 4b1b100: OUTPUT«===SORRY!===␤Placeholder variable $^a may not be used here because the surrounding block takes no signature at /tmp/qfnzs9DWOn line 1:␤------> loop { ⏏$^a }␤Check failed␤FAILED 00:01 112m␤»
TimToady rakudo: loop { $^a }
p6eval rakudo 048573: OUTPUT«(timeout)»
TimToady heh
jnthn :P 21:11
TimToady so that's the mechanism I was thinking of, it just doesn't work out for bare blocks
since we don't know they're bare yet 21:12
unless we outlaw statements of the form {...}()
masak that'd be a shame. 21:13
TimToady which seems like overkill, esp since rakudo picks up the error at run time
>>todo
jnthn May be easier with semantic analysis. 21:14
TimToady Might notice it when trying to inline bare blocks 21:16
21:21 lumi left 21:22 mikehh joined, lumi joined
mikehh I am getting a build failure with latest rakudo master Dahut-12-g048573b on parrot RELEASE_3_4_0-153-gc2e3158 21:24
mikemol colomon: I only just now noticed that you implemented the Pi spigot as a sequence, but you didn't add it to justrakudoit.wordpress.com/2011/03...ces-of-pi/
Mind adding that? 21:25
(Or mind if I do? Though it's easier to preserve attribution if you do)
mikehh see nopaste.snit.ch/49579
looks like a problem with Test.pm 21:27
jnthn TimToady: Is there a good criteria for when a bare block can be inlined, ooc? 21:29
e.g. in if $foo { bar() } else { baz() }
21:29 math1 left
jnthn In that case do things like $! get in our way? 21:29
Or generally, any contextual where the outer block has a contextual of the same name? 21:30
21:30 mj41 joined 21:37 ymasory joined
TimToady I'd imagine such in-lining would emulate the bits of entry/exit that are needed, and throw away the rest 21:37
if we find our semantics continually clobbering our performance, though, we can revisit our semantics 21:38
jnthn Probably do-able. 21:41
21:42 bluescreen10 left 21:45 tadam left
jnthn 1 files changed, 61 insertions(+), 119 deletions(-) 21:46
21:46 Psyche^ joined
jnthn Yeah, I like that kinda refactor... 21:46
masak \o/
21:47 MayDaniel joined, lumi left 21:48 lumi joined
jnthn yay, "my $x := -> Mu $a { say($a) }; $x(42)" works again. 21:48
21:48 wamba joined
jnthn Now to figure out where the missing decontainerization is... 21:49
oh, it's beer time!
21:50 Patterner left, Psyche^ is now known as Patterner
masak rakudo: class A { has @.x }; my $a = A.new(:x(1,2,3)); my @x := $a.x; @x[1] = 42; say $a.x.perl 21:50
p6eval rakudo 048573: OUTPUT«[1, 42, 3]␤»
masak does anyone else consider this to be a problem, from an encapsulation point-of-view?
jnthn If you didn't want to share it, write @!x. 21:51
And your own accessor that clones or something. 21:52
masak right.
TimToady
.oO(beer decontainerization!)
jnthn \o/
dalek kudo/nom: 92f19d1 | jnthn++ | src/Perl6/Grammar.pm:
Toss unused token.
kudo/nom: ccc2fb3 | jnthn++ | src/Perl6/Actions.pm:
Big refactor to try and get some closure handling stuff back in place and working. Also some basic pointy block support.
kudo/nom: 6e94078 | jnthn++ | src/CORE.setting/Code.pm:
A cheaty Code.clone.
jnthn phew, closures actually seem to work
> sub foo(Any $i) { -> { say($i) } }
> my $a := foo(1); my $b := foo(2); $a(); $b();
1
2
masak I guess we can even provide attribute metaobjects that have the right default (that of cloning) from the start. so no real harm done.
jnthn masak: right
masak it's just a bit... unnecessary... that the default default is disencapsulation. 21:53
jnthn my module EXPORTHOW { class MyMetaAttr { ... } }
oh wait, not that
my class MyMetaAttr { ... }; my module EXPORTHOW { $?PACKAGE.WHO<class-attr> := MyMetaAttr } # that... :) 21:54
masak TheDamian was concerned about attribute introspection breaking encapsulation. he should be concerned about the default accessors breaking encapsulation.
jnthn masak: It's nothing to do with the accessors really.
masak the accessors are handing out the array reference from inside the object. 21:55
jnthn masak: They just return the array. Doing an index into the array is a method on the array, thorough the array's public interface.
masak by then it's too late.
jnthn What do you want the accessors to hand out?
Do you want an array accessor to shallow-clone the entire array? 21:56
masak a different array. like you say, a clone.
jnthn I guess with laziness it's not *so* expensive.
masak yes, I guess I see the issues with cloning.
TimToady we really need "capabilities"
where the default object returned is immutable, and you need a special view of the object to let you write it 21:57
jnthn TimToady: It's the making it immutable cheaply bit that'll take a little effort.
TimToady if I were doing this in the bad old days I'd use the 1-bit of the pointer to indicate rw-ness....
I'd make immutable the default
jnthn the 1-bit...eewwww! 21:58
:P
TimToady and add in the capability bit for the rw, at the expense of having to strip the bit
21:58 bluescreen10 joined
TimToady or just trap the SEGV and test for rw in the signal handler ;) 21:59
masak :P
'night, #perl6
stay immutable.
21:59 masak left
TimToady I'm hard to mute. 22:00
22:00 mj41 left
colomon mikemol: huh? 22:00
tylercurtis rakudo: class ReadonlyPositional does Positional { has @!arr; method at_pos($i) { my $x = @!arr.at_pos($i); $x } }; class A { has @!foo; method foo { ReadonlyPositional.new(:array(@!foo)) } }; my A $a .= new(:foo(1,2,3); my @x := $a.foo; @x[1] = 42; say $a.foo.perl 22:01
mikemol A while back, I indicated I thought Perl6 should implement rosettacode.org/wiki/Pi as a sequence
p6eval rakudo 048573: OUTPUT«===SORRY!===␤Confused at line 22, near "my A $a .="␤»
tylercurtis rakudo: class ReadonlyPositional does Positional { has @!arr; method at_pos($i) { my $x = @!arr.at_pos($i); $x } }; class A { has @!foo; method foo { ReadonlyPositional.new(:array(@!foo)) } }; my A $a .= new(:foo(1,2,3)); my @x := $a.foo; @x[1] = 42; say $a.foo.perl
p6eval rakudo 048573: OUTPUT«ReadonlyPositional.new(arr => [])␤»
mikemol From that, you wrote justrakudoit.wordpress.com/2011/03...ces-of-pi/ justrakudoit.wordpress.com/2011/04/26/more-pi/ and justrakudoit.wordpress.com/2011/04...eam-of-pi/ 22:02
tylercurtis rakudo: class ReadonlyPositional does Positional { has @!arr; method at_pos($i) { my $x = @!arr.at_pos($i); $x } }; class A { has @!foo; method foo { ReadonlyPositional.new(:array(@!foo)) } }; my A $a .= new(:foo(1,2,3)); my @x := $a.foo; @x[1] = 42; say $a.foo.perl; say $a.perl
p6eval rakudo 048573: OUTPUT«ReadonlyPositional.new(arr => [])␤A.new(foo => [1, 2, 3])␤»
tylercurtis rakudo: class ReadonlyPositional does Positional { has @!arr handles <elements>; method at_pos($i) { my $x = @!arr.at_pos($i); $x } }; class A { has @!foo; method foo { ReadonlyPositional.new(:array(@!foo)) } }; my A $a .= new(:foo(1,2,3)); my @x := $a.foo; @x[1] = 42; say $a.foo.perl; say $a.perl
p6eval rakudo 048573: OUTPUT«ReadonlyPositional.new(arr => [])␤A.new(foo => [1, 2, 3])␤»
tylercurtis rakudo: class ReadonlyPositional does Positional { has @!arr handles <elements>; method at_pos($i) { my $x = @!arr.at_pos($i); $x } }; class A { has @!foo; method foo { ReadonlyPositional.new(:array(@!foo)) } }; my A $a .= new(:foo(1,2,3)); my @x := $a.foo; @x[1] = 42; say $a.foo[1] 22:03
p6eval rakudo 048573: OUTPUT«Any()␤»
tylercurtis rakudo: class ReadonlyPositional does Positional { has @!array handles <elements>; method at_pos($i) { my $x = @!array.at_pos($i); $x } }; class A { has @!foo; method foo { ReadonlyPositional.new(:array(@!foo)) } }; my A $a .= new(:foo(1,2,3)); my @x := $a.foo; @x[1] = 42; say $a.foo[1]
p6eval rakudo 048573: OUTPUT«2␤»
tylercurtis (I think that just directly returning the result of @!array.at_pos should work, but ISTR some discussion of Rakudo not doing rwness/roness of subs in some ways.) 22:05
Although that approach is wrong for some use cases, in that between getting foo and accessing its elements, the underlying array could be changed. 22:10
TimToady readonly doesn't necessarily imply value semantics, that's what 'is copy' is for 22:12
the readonly default is intended only to protect the caller from side-effects in the callee, not vice versa 22:15
22:15 hercynium left 22:20 bluescreen10 left 22:22 bluescreen10 joined 22:35 bluescreen10 left 22:39 bluescreen10 joined 22:41 lumi left, lumi joined 22:43 bluescreen10 left
dalek p: 7ca852e | jonathan++ | src/ (2 files):
More decontainerization twiddling.
 22:46
22:49 lue left, hanekomu left
dalek kudo/nom: 9df2d17 | jnthn++ | src/Perl6/Actions.pm:
Get corrector semantics and save an op off every method call, thanks to putting decontainerization in a righter place.
 22:49
22:51 whiteknight joined 22:56 wanradt___ left
sorear jnthn: what was that change about? 22:59
are you trying to implement non-rw subs?
23:00 MayDaniel left 23:02 bluescreen10 joined 23:03 wamba left
jnthn sorear: It was about quite a few things. :) 23:07
sorear: Well, depends if you mean nqp one or Rakudo one.
sorear: Rakudo one just means "don't decontainerize invocant here, it's done elsewhere now" 23:08
And only needs doing in some places
Otherwise you can't write a (grrrr...but still...) method foo($self is rw :) { $self = ... }
23:11 _jaldhar joined
sorear jnthn: grr? 23:12
jnthn sorear: It's possible but evil :)
sorear: Anyway, in theory it works in nom now.
sorear meh
jnthn In practice I didn't do "is rw" yet...or did I...
sorear it's what makes hash autovivification work
jnthn oh, yes 23:13
sorear so I put up with it
jnthn :)
23:13 bluescreen10 left
sorear ponders stealing jnthn's SIG_ELEM_* defines 23:14
jnthn Yes but mabye call them PARAM_ like I should have. ;)
(I mostly killed the term "sig elem"... :)) 23:15
sorear How does SLURPY_BLOCK work?
that (and, ok, attributives) is the only one I don't have an _exact equivalent_ to
jnthn I don't think I ever implemented slurpy blocks. 23:16
Nobody's ever missed them so far. :/
iirc, I couldn't think of an easy way to do them, so I figured "eh, I'll see if anybody moans about them being missing" :) 23:17
Occasionally when you do that, the scary bit of spec in question goes away. : 23:18
:P
rakudo: say pir::eq__III(1,1)
p6eval rakudo 048573: OUTPUT«===SORRY!===␤error:imcc:The opcode 'eq_i_i_i' (eq<3>) was not found. Check the type and number of the arguments␤ in file '(file unknown)' line 76407100␤»
jnthn ...line wut? 23:19
23:20 silug left
jnthn rakudo: say pir::iseq__III(1,1) 23:21
p6eval rakudo 048573: OUTPUT«1␤»
sorear jnthn: well it's an imcc error, you know Rakudo generally makes super-bloated PIR 23:23
reading bind.h makes me feel so _behind_ 23:24
jnthn sorear: The one in nom?
sorear aye
23:24 Mowah left
jnthn heh...niecza does rather more than rakudo/nom :) 23:24
I'm not even Turing complete yet. :) 23:25
sorear can't even emulate a tag system or a Minsky machine? 23:26
doesn't nom have recursion?
and integers?
jnthn oh, yes, it does
But...no relational ops :)
sorear so set up a bunch of mutually tail-recursive functions to emulate the states of a Minsky machine
do you have Int.Bool?
jnthn Not yet 23:27
Working on those things at the moment.
sorear rummages through his collection of hats, pulls out the one labeled Unlambda User 23:28
jnthn
.oO( A curious implementation strategy. Meta-classes and multi-dispatch before booleans )
sorear church encoding can give you your booleans 23:29
sub true($x,$y) { $x }
jnthn Hey, stop reminding me how many years it's been sicne my computation theory lectures :P 23:30
.oO( I wonder if you can build a turing machine out of different depths of inheritance hierarchy... )
 23:31
sorear .o( I had best hurry up, nom isn't holding still ) 23:34
23:41 _jaldhar left 23:44 awoodland left 23:50 _jaldhar joined 23:52 silug joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!