»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.
pmurias you need installed clisp with configure quicklisp 00:00
sorear with -Bclisp?
pmurias yes
sorear what's quicklisp?
dalek ecza: cf9e0a4 | sorear++ | src/ (2 files):
Remove some gimme5 cruft from this STD too
 00:04
sorear rakudo: say ( 1 ).WHAT
00:04 shi joined
pmurias sorear: cpanm for lisp 00:04
p6eval rakudo : OUTPUT«Int()␤»
sorear rakudo: say ( 1, ).WHAT
p6eval rakudo : OUTPUT«Parcel()␤»
pmurias www.quicklisp.org/beta/
quicklisp allows loading lisp modules with (ql:load "module") which automatically fetches from the internet/and installs that module 00:05
sorear: quicklisp is required for installing the json library 00:11
sorear pmurias: how helpful was the use of json? 00:14
pmurias: you've talked in the past about moving to a binary format 00:16
00:16 awwaiid joined
sorear I am considering this 00:16
00:19 bluescreen left
pmurias the use of json was a major annoyance 00:21
i had to unpack the arrays buy hand
* by
hmm, sbcl seems to be faster the the mono backend on "while $a < 1000000 {$a = $a + 1} 00:24
i'll investigate it tommorow
s/the/then
sorear: what do you want to move to? 00:25
sleep& 00:30
00:30 pmurias left 00:32 Sarten-X left
TimToady sorear: 'is unary is defequiv' is sort of the opposite of a fossil; it's more like info hung out there for the use of macros someday 00:34
but it might be the wrong place to hang the info anyway
they are fossils in the sense that they were put there a long time ago
back before STD could even be executed 00:35
00:35 bluescreen joined 00:48 felliott left 00:54 felliott joined 00:55 lakshman joined
lakshman Is there an expert in perl 6 datastructures? 00:56
TimToady depends on whether you're mad about them :) 00:58
lakshman mad about them? 00:59
TimToady a number of people understand S09 pretty well, but nobody's implemented it all yet
I guess I don't know what you're actually asking
they could use a champion, if that's what you're asking 01:00
but there are a number of people who can answer questions about the design 01:01
01:01 qiyong joined
TimToady lakshman: do you have a more specific question? 01:03
lakshman no I need to come up with specific questions i'll get back to you
Thanks
TimToady no problem
01:04 coldhead joined 01:14 Sarten-X joined 01:21 lakshman left 01:24 felliott_ joined, felliott left 01:25 felliott_ is now known as felliott 01:27 shi left 01:32 jaldhar_ joined 01:34 icwiener left 01:38 gdey left 01:44 gdey joined
sjohnson hmm, new vim with blowfish encryption for text files is kind of handy 01:50
01:55 gdey left 02:01 noganex_ joined 02:03 Chillance left 02:04 noganex left 02:08 jaldhar_ left 02:12 Sarten-X left 02:17 risou joined, kanishka joined 02:18 Sarten-X joined 02:19 risou left 02:42 stkowski left 02:45 bluescreen left 02:50 whiteknight left 02:52 justatheory left 02:54 minimalist joined 02:57 nadim left 03:16 cdarroch left 03:21 lue joined
sorear phenny: tell pmurias Half the reason I picked JSON was to make life easier for people like you. Let's discuss what you could have used later. 03:23
phenny sorear: I'll pass that on when pmurias is around.
sorear TimToady: ok. 03:24
03:24 lakshman joined
sorear wb lakshman 03:24
03:24 agentzh joined
sorear TimToady: I'll make niecza ignore those traits, then. 03:24
or maybe they could be transformed into something readable at runtime 03:25
lue I'm not dead yet, zebras! 03:26
03:29 jferrero left 03:36 ryan__ joined 03:46 dukeleto left 03:47 dukeleto joined 03:59 Trashlord left 04:03 Trashlord joined 04:04 HarryS left 04:23 Su-Shee_ joined 04:26 Su-Shee left 04:27 BaggioKwok joined 04:29 BaggioKwok left 04:33 stifynsemons joined 04:36 Sarten-X left
stifynsemons how does one invoke a subroutine indirectly, as in &{"test"} from perl5? 04:37
04:37 kanishka left
sorear I guess it would be something like callframe.my.{'&test'}.() 04:38
sub test has to be declared 'is dynamic' for this to work
TimToady well, ::("test")() ought to work
sorear generally speaking Perl 6 doesn't swing that way
TimToady but I doubt rakudo does that yet
sorear subs are pretty hidden 04:39
TimToady: why not ::('&test')() ?
TimToady all symbolic refs are syntactically distinguished from hard refs
sorear wouldn't ::("test") be referencing a typename?
TimToady yeah, probably wants the &
stifynsemons Indirect name lookups not implemented yet :-(
04:39 jaldhar_ joined
stifynsemons thanks though 04:40
sorear the Perl 6 Way doesn't use indirect name lookups very often
it's one of those things we support more for supporting 5->6 than actually being important 04:41
stifynsemons so if I want to map a user supplied string into a function name, how would I do that the Perl 6 way?
sorear is it important that the user be able to run any Perl 6 function?
including stuff like CORE::tan and CORE::exit? 04:42
stifynsemons not any perl6 function in general; normally I add a prefix which restricts this.
sorear it's better form to use an explicit hash for that sort of thing
stifynsemons I could implement it I suppose as a set of anonymous subroutines stored in a hash table, I think.
I can appreciate that. 04:43
sorear they don't have to be anonymous
sub foo { ... }; sub bar { ... }; my %hash = :&bar, :&foo;
stifynsemons oh, that's not much different that what I intended that way ... since I don't have to assign the sub definitions directly to the hash table. 04:44
thanks
04:44 Sarten-X joined
stifynsemons that seems to work well 04:45
TimToady pugs: sub foo { say "yup" }; ::("foo")() 04:46
p6eval pugs: ( no output )
TimToady pugs: sub foo { say "yup" }; MY::{'foo'}() 04:48
p6eval pugs: OUTPUT«*** ␤ Unexpected "MY"␤ expecting ";", Doc block, block declaration, declaration, construct or expression␤ at /tmp/rFEs7rPRrg line 1, column 24␤»
dalek ecza: bec0cd5 | sorear++ | lib/Builtins.cs:
Abstract map logic away from simple batching
 05:00
05:12 minimalist left
sorear rakudo: say ~(1,2 Z+ 10,20,30,40) 05:14
p6eval rakudo : OUTPUT«11 22␤»
sorear rakudo: say ~(1,2 Z 10,20,30,40)
p6eval rakudo : OUTPUT«1 10 2 20␤»
sorear TimToady: correct?
TimToady yes 05:18
05:21 kanishka joined 05:47 ryan__ left 05:55 takadonet joined
takadonet yo 05:55
05:56 takadonet left, takadonet joined
takadonet try this again... 05:57
sorear hi takadonet
takadonet sorear: how are u sir?
sorear I hear you do BioPerl6
ryan__ was looking for gsoc stuff earlier
takadonet for doing work on bioperl6? 05:58
sorear possibly 05:59
takadonet Pyrimidine would be the man to discuss that
aka github.com/cjfields 06:00
he also the pumpkin of perl 5 version of bioperl
what library (on linux) do you need so you can view previous commands enter when in the interactive mode? 06:02
i always forgot
forget* 06:03
arnsholt readline 06:04
06:05 stifynsemons left
takadonet arnsholt: thx 06:05
06:07 felliott left 06:19 takadonet left 06:22 fhelmberger joined 06:29 satyavvd joined
sorear TimToady: What should 1,2 Z> 1,2 Z> 1,2 do? 06:46
06:46 Trashlord left 06:47 woosley joined 06:48 cxreg2 is now known as cxreg 06:49 kanishka left 06:52 Trashlord joined 06:53 kanishka joined
TimToady I'd think it should return (1>1>1), (2>2>2) 06:53
False,False, in other words 06:54
sorear I wonder if Z needs to be sensitive to precedence for that to work
dalek ecza: b86e304 | sorear++ | lib/Builtins.cs:
Streamline flattening process for map
 06:55
ecza: f3b81bb | sorear++ | / (5 files):
Implement Z, Zop, X, Xop
TimToady Z is always a list infix in precedence, but it does need to be sensitive to chaining semantics, I expect 06:56
sorear -> sleep 06:59
07:13 nadim joined 07:15 envi joined
coldhead congratulations on the looming nuptials, TimToady 07:43
07:45 Mowah joined 07:58 hudnix left, hudnix joined
dalek ast: 051af06 | moritz++ | S0 (3 files):
some rakudo unfudges, mostly related to infix orelse
 08:07
moritz_ good morning
tadzik o/ 08:19
08:23 jaldhar_ left, _jaldhar_ joined 08:34 agentzh left 08:42 agentzh joined 09:00 kst left 09:03 protorom joined 09:04 wiseguyxp_ left 09:13 Sarten-X left, kanishka left, satyavvd_ joined 09:15 satyavvd left 09:21 Sarten-X joined 09:28 Su-Shee_ is now known as Su-Shee 09:35 REPLeffect left 09:37 REPLeffect joined 09:44 lue left, lue joined 09:46 woosley1 joined 09:47 woosley left, cjk101011 joined 09:52 Chillance joined 10:01 satyavvd_ left 10:02 woosley1 left 10:22 _jaldhar_ left, _jaldhar_ joined 10:23 peters_mops left 10:29 _jaldhar_ left, _jaldhar_ joined
moritz_ rakudo: say 1 !+ 3 10:35
p6eval rakudo : OUTPUT«===SORRY!===␤Can't negate + because it is not iffy enough at line 22, near " 3"␤»
10:44 coldhead left
dalek kudo: f4ec358 | moritz++ | docs/ChangeLog:
[docs] update ChangeLog
 10:49
10:59 Sarten-X left 11:01 pmurias joined 11:07 Sarten-X joined 11:15 whiteknight joined 11:35 colbseton joined 11:39 colbseton left 11:42 Trashlord left 11:52 masak joined
masak привет, zebras. 11:52
11:54 agentzh left 12:01 MayDaniel joined
masak has been wrapping his head around the SOLID principles lately 12:03
they're mostly what I expected them to be, except the Open/closed principle, which didn't make any sense whatsoever until it suddenly did. 12:04
it's also interesting/valuable to discover places in well-designed libraries where these principles are broken, sometimes deliberately as a tradeoff. 12:08
12:09 shi joined 12:27 MayDaniel left 12:33 colbseton joined 12:51 BinGOs left 12:53 pmurias left, edenc left, perigrin left, perigrin joined 12:54 edenc joined 12:55 BinGOs joined
masak if C is a class fairly far down in an inheritance hierarchy, and it has an (overridden) method .foo, will the PRE blocks in the .foo methods of C's base class(es) also trigger when C.foo is called? 12:57
is the answer different if .foo is a multi method?
&
jnthn Sleep. I got it. \o/ 13:02
13:02 Trashlord joined 13:12 felliott joined 13:17 orafu left 13:19 _jaldhar_ left, _jaldhar_ joined 13:21 protorom left 13:25 drbean left
moritz_ congratulations jnthn. I'm jealous :-) 13:35
jnthn moritz_: Aww. Sorry.
hugme: hug moritz_
hugme hugs moritz_
13:49 MayDaniel joined
colomon remembers rocking a very small baby at 2am, reading a Gene Wolfe paperback in a strange sleepless haze. 14:05
Now-a-days, however... well, last night I was rocking a much larger toddler, and reading Lois McMaster Bujold on my Droid. 14:06
14:08 kanishka joined 14:16 hudnix left
jnthn ooh, bacek++ has added the write barriers to nqp \o/ 14:20
moritz_ and the parrot folks have temporarily unmerged the gen_gc branch 14:22
jnthn oh.
moritz_ because it had troubles to build on some platforms
jnthn Aww.
OK.
colomon write barriers? 14:23
moritz_ and performance regressions in register access 14:25
14:28 vmspb joined 14:30 satyavvd joined
jnthn colomon: Triggers some GC action when an object is written to and before some allocation happens. 14:31
s/allocatin/something that may trigger a GC run/
14:36 satyavvd left
masak a write barrier is a GC trigger? 14:37
gee, sounded more like it would be some sort of barrier against writing... :)
jnthn masak: It's not a trigger as such 14:38
masak: It's more to do with generation management, so the next run doesn't eat stuff it shouldn't.
masak ah, ok.
jnthn masak: If an object in one generation comes to point to things in other generations, that needs some special handling on the next GC run.
iiuc, anyway ;)
masak it has a ring of truthiness to it... 14:39
now, what about my PRE/POST question above?
what's the relation between a method in a derived class, its (top-level) PRE/POST blocks, and methods in more base classes and their (top-level) PRE/POST blocks. 14:40
s/\./?/
jnthn masak: AFAIK none at the moment. (more)
14:41 xinming_ left
jnthn masak: I'm not sure we should go in the direction of enforcing that they're always run up/down the hierarchy. It may be very Liskov-ish, but we aren't so Liskov-ish to always give the methods in a parent class a chance to run either. 14:42
14:42 xinming joined
jnthn Further, if the invariants relate to arguments passed to the method, how does a PRE/POST block in a method in a parent class make sense if it's never invoked? 14:42
masak yes, that's what I'm fishing for.
but S04:1459 says "call all the appropriate per-method C<PRE> phasers", which makes it sound like there's a chain of PRE phasers. 14:43
jnthn Oh. :/
I'm not sure that makes sense - or is practical.
masak there might be, even for one method, of course, but there's no real need to divide things into several PRE phasers.
no, I don't think it ultimately does make sense.
jnthn Right.
My feeling is that we'll have the meta-programming capabilities to let people who want Beta-style OO where you always call the least derived method and it chooses how to delegate to more derived ones. 14:44
masak just consider .foo($a) in Parent and .foo($b, $c) in Child -- how would PRE be able to run in Parent if Child.foo was called?
jnthn s/who want/implement/
masak jnthn: that sounds a little like proto methods.
jnthn Well, not quite. A bit broader: it sounds like dispatchers. 14:45
moritz_ I think only PRE phasers from the candidate list should be considered
ie those that could be reached with nextsame()
masak that makes a bit more sense.
jnthn Apart from instead of the list you're iterating over being most to least derived, it's the other way. 14:46
masak *nod*
jnthn moritz_: Doing nextsame() into a method (note, *not* multi - that's different) can fail.
e.g. if I have a method m($a) { } in a parent class and a method m($a, $b) { nextsame() } in its child, that *will* fail. 14:47
For multis though, you'd not get that situation.
Of course, hopefully people will pay attention to Liskov and never put themselves into such a situation. :)
masak feels he has not completely encompassed the problem space yet 14:48
jnthn I don't particularly want to make method single-dispatch more complex for the sake of encouraging people to do bad OO.
masak it makes a lot of sense that people who dabble with PRE/POST would be Liskov-aware.
moritz_ what about class A { method x($y) { my $k = $y ** 2; PRE { $k <= 10 } }} ; class B is A { method x($y) { ... } }
does it have to run the parent method up to the point where PRE can be evaluated? 14:49
masak moritz_: uhm.
moritz_: PRE blocks are run before the method body.
says so in S04.
jnthn masak: That's only half the truth.
masak that's why they're called PRE :)
jnthn masak: The method callframe surely needs to exist.
But before the "main" body, yes.
masak that's what i meant :)
jnthn OK :)
That PRE would work albeit with a warning, I guess. :) 14:50
rakudo: say Any <= 10
p6eval rakudo : OUTPUT«Use of uninitialized value in numeric context in <anon> at line 22:/tmp/vQPJ_UVjZ_␤Bool::True␤»
jnthn :)
14:50 risou joined
jnthn I guess the rule is "PRE should only care about attributes and arguments" 14:51
masak PRE runs after INIT, but before ENTER. so there's no really good way to assign to $k each time. 14:53
but the assignment might as well be done inside the PRE block (the declaration still being before/outside it). problem solved. 14:54
moritz_ INIT my $x = $y * $y;
jnthn Don't that run, like, one, ever? :) 14:55
masak yes.
jnthn *Won't
*once
...can't type. :)
masak so it doesn't quite cut it.
14:57 risou left 14:58 kanishka1 joined, kanishka1 left 14:59 kanishka1 joined 15:00 kanishka left
masak hm, if I implement ENTER blocks in Yapsi, it'll be easier to do named subs... 15:12
moritz_ why is that? 15:13
masak because I plan (at least initially) to let 'sub foo { ... }' be sugar for 'my &foo;', 'ENTER { &foo := { ... } }' 15:16
15:17 icwiener joined
masak yapsi: say 42 15:17
p6eval yapsi: OUTPUT«===SORRY!===␤Unable to find module 'Yapsi' in the @*INC directories.␤(@*INC contains:␤ lib␤ /home/p6eval/.perl6/lib␤ /home/p6eval/p1/lib/parrot/3.0.0-devel/languages/perl6/lib␤ .)␤»
moritz_ so... a sub doesn't have a body until it's called? 15:25
moritz_ confused
oh wait, ENTER from the outer scope
masak aye.
maybe there's a "Liskov dialect" or Perl 6, accessible through a pragma, wherein DBC is possible, and methods are required to override with the same signature. that would make PRE/POST blocks along the inheritance chain much more feasible. 15:26
jnthn masak: lisklass Foo { ... } # :-) 15:27
masak but they're slightly odd creatures. if there's a parameter $x, and the PRE block of Parent.foo mentions $x, we expect it to bind to the $x of Child.foo when it is called.
jnthn Eww. :)
masak very anti-lexical of them.
moritz_ I think we need to signature-bind to all the parent methods where PRE blocks are run 15:28
masak yes, probably. that would solve that. 15:29
jnthn phenny: tell pmichaud when you get chance, please take a look over github.com/rakudo/rakudo/blob/nom/...P.markdown - it's my work plan for the nom branch. kplzthx
phenny jnthn: I'll pass that on when pmichaud is around.
masak markdown \o/ 15:30
jnthn ETOOLAZYTOPOD :)
masak well, Github highlights both, IIRC. 15:31
jnthn Right. :)
15:31 hudnix joined
jnthn But the difference between what I have to type to hae a text file and a markdown file is...tiny. :) 15:31
So there's kinda no reason not to do it.
15:32 stifynsemons joined 15:34 stifynsemons left 15:36 Sarten-X left 15:44 Sarten-X joined
masak right. it's Structure for Free. 15:46
dalek p: c8ce669 | jnthn++ | src/pmc/nqpmultisig.pmc:
Fix a mark bug that saw some people get segfaults in multi-method dispatch situations.
 15:49
masak I have a heisentest here: a Perl 6 program that Yapsi should run but happily runs on the command line. when I copy+paste it as a test, Yapsi refuses to compile it (as it should). 15:52
15:52 dukeleto left
masak I find myself wanting a test to fail. :) 15:53
15:53 dukeleto joined
moritz_ is there a "not" missing in your previous statement? 15:53
"should run but happily runs"
masak right. "that Yapsi shoudn't compile"
fwiw, the program is 'my $a = 1; { say $a; my $a = 2; say $a }' 15:54
std: my $a = 1; { say $a; my $a = 2; say $a }
p6eval std 625303c: OUTPUT«===SORRY!===␤Lexical symbol '$a' is already bound to an outer symbol (see line 1);␤ the implicit outer binding at line 1 must be rewritten as OUTER::<$a>␤ before you can unambiguously declare a new '$a' in this scope at /tmp/7qHlmTzxbR line 1:␤------> my
..$a…
15:55 Sarten-X left
jnthn I can't remember who reported the nqp multi dispatch test segfault, but the patch I just commited above should fix it. 15:56
walk & 15:58
dalek p: 5ed3fad | jnthn++ | src/metamodel/reprs/P6str.c:
Fix an issue in the P6str repr spotted by sorear++.
 16:00
16:03 Sarten-X joined 16:04 Sarten-X left 16:05 vmspb left
masak ah. the heisentest turned out to be due to state bleedover between runs. 16:11
testers take note: a test showing that something fails isn't much worth on its own, since there are so many orthogonal ways something can fail. :)
arnsholt Indeed. State behaving in funny ways makes for interesting debugging of tests 16:12
masak I have a few globals in Yapsi. one of them is the culprit here. 16:13
arnsholt I've had some fun (fsv. of fun) debugging backtracking shenanigans in Parrotlog =)
16:13 wooden left 16:15 qiyong left, Patterner left 16:16 cognominal left 16:17 Psyche^ joined, Psyche^ is now known as Patterner, ryan_ joined
masak yay! changed the global into a private attribute on the ::Actions class, and that brought out the failing test! \o/ 16:18
if that isn't a victory for narrow scopes, I don't know what is. 16:19
16:19 HarryS joined
arnsholt Absolutely 16:21
16:23 Sarten-X joined
jnthn back :) 16:34
16:35 stifynsemons joined 16:36 whiteknight left 16:37 cognominal joined 16:39 stifynsemons left 16:52 plobsing left, stifynsemons joined 16:56 stifynsemons left 17:14 plobsing joined
dalek psi: 639e391 | masak++ | lib/Yapsi.pm:
[Yapsi] narrowed the scope of %vars

Instead of a lexical inside ::Actions, it is now a private attribute. This makes a lot more sense, because the variable should be reset with each new parse anyway.		 17:15
masak std: unless a {} 17:17
p6eval std 625303c: OUTPUT«===SORRY!===␤Function 'a' needs parens to avoid gobbling block at /tmp/XQXOz5gt1Q line 1:␤------> unless a⏏ {}␤Missing block (apparently gobbled by 'a') at /tmp/XQXOz5gt1Q line 1:␤------> unless a {}⏏<EOL>␤ expecting any
..of…
sorear good * #perl6 17:30
masak sorear! \o/ 17:32
diakopter * * *
jnthn It's not * here
oh, wait..
masak { $_ * $_ }
jnthn :)
diakopter std: * * *
masak oh wait, { $^a * $^b }
p6eval std 625303c: OUTPUT«ok 00:01 119m␤»
masak { $_ * $_ } was the old sugaring...
*desugaring 17:33
17:38 dual left
masak std: my @a = 1, => 2 17:38
p6eval std 625303c: OUTPUT«===SORRY!===␤Preceding context expects a term, but found infix => instead at /tmp/atbVfl3TGD line 1:␤------> my @a = 1, ⏏=> 2␤Parse failed␤FAILED 00:01 122m␤»
masak rakudo: my @a = 1, => 2
p6eval rakudo : OUTPUT«===SORRY!===␤Method 'returns' not found for invocant of class 'Undef'␤»
masak submits LTA rakudobug
dalek psi: b62dada | masak++ | t/compiler.t:
[t] two new tests

One fails.		 17:42
psi: 97dea43 | masak++ | t/compiler.t:
[t/compiler.t] checked error message on failure

Just knowing that something fails isn't very informative. It turns out we were sometimes failing for the wrong reasons. This commit is more discerning about the actual failure.
17:44 dual joined 17:45 MayDaniel left 17:51 justatheory joined 17:54 MayDaniel joined 17:56 kanishka1 left, kanishka joined
TimToady is anyone here currently using rakudo on Mac? What's the best way to install? 17:59
17:59 MayDaniel left
jnthn TimToady: I think layla++ asked that recently...I *think* the answer was that there wasn't any packages around, so best was just to build it from source. 18:00
jnthn is failing ot use the ir clog search... 18:02
TimToady thanks, we'll do that 18:04
jnthn TimToady: irclog.perlgeek.de/perl6/2011-02-01#i_3242639
masak yes, build from source. 18:13
18:14 kaare_ joined
masak Mac OS X is basically BSD with pretty graphics. ;) 18:14
tadzik isn't that Darwin? 18:20
arnsholt Darwin and BSD userspace, IIRC 18:24
And I use Rakudo on OS X. I build from source 18:25
18:25 kanishka left
colomon I use Rakudo on OS X, built from source. 18:28
Someone was working on building package distribution recently, though. 18:29
18:29 pmurias joined
pmurias sorear: hi 18:30
phenny pmurias: 03:23Z <sorear> tell pmurias Half the reason I picked JSON was to make life easier for people like you. Let's discuss what you could have used later.
18:44 kaini joined 18:45 kaini left 18:50 kaini joined 18:56 jvv joined
sorear pmurias: hi 19:00
lue good morning zebras! o/ 19:11
tadzik hey lue! 19:13
masak lue \o
sorear hi lue 19:14
lue is so close to beating TeX and friends with several forms of aquatic life. 19:23
tadzik ...why can we have 61 seconds? 19:24
lue I think I asked before, and I believe it has to do with leap seconds. 19:25
tadzik perlcabal.org/syn/S32/Temporal.html#line_72
19:25 wiseguyxp_ joined
sorear tadzik: leap seconds are added to the end of existing minutes 19:26
tadzik: the Earth rotates on its axis once every 86400.xxxxx seconds
tadzik I see
sorear we divide the sidereal day into 24 hours with 60 minutes each
TimToady they should just put some big weights in some towers so they can adjust the rotation rate to match the atomic clock 19:27
sorear most minutes have 60 seconds, but to keep UT1 synchronized with the Sun we add 1 second to a minute occasionally
erm, the solar day 19:28
the sidereal day only has 1436 minutes...
lue TimToady: or giant rockets.
TimToady or more likely (meaning not likely at all) pump a great amount of water to/from reservoirs on the equator
sorear because the Earth rotates on its axis every 1436 minutes, but the Earth is also revolving around the Sun 19:29
TimToady we should do something about the year length too
sorear every 366 rotations of the Earth, we see the Sun rise 365 times because 1 of them was cancelled out by yearly motion
TimToady you can get it back by traveling around the world in the right direction 19:30
sorear I wonder if we could cancel out global warming by cleverly manipulating the Milankovitch cycles
lue
.oO(Or perhaps open a hole in the atmosphere that's always on the Dark Side™ of the Earth, to let all the heat out)
 19:33
pmurias re what to replace json with something with a structure and existing implementations would be preferable
sorear: why do you want to replace json? (in what ways does the replacement have to be better) 19:34
sorear pmurias: you said json made your work harder
the main reason I picked json was to make implementing backends (readers in genral) *easier*
so obviously something is very wrong
pmurias common lisp doesn't have pattern matching 19:35
19:35 shi left
pmurias so unpacking arrays had to be worked around 19:35
sorear what could you have dealt with easier? 19:36
19:38 envi left
pmurias i'm not sure i don't have much experience with different serialisation formats 19:40
in the end i just converted json into sexpressions and copied out parts of docs/nam.pod 19:41
19:48 molaf joined
masak I should probably attack serialization soon in Yapsi. I'm getting to the point where I need static lexpads and BEGIN to work. 19:50
19:51 cjk101011 left
dalek ecza: 311ba52 | sorear++ | / (2 files):
Reorganize and combine TODO lists
 20:04
20:04 rgrau joined 20:08 felliott left 20:10 Tedd1 left, wiseguyxp_ left 20:19 stifynsemons joined 20:24 felliott joined 20:28 felliott left 20:29 stifynsemons left
masak that's a nice read. EASY, MEDIUM, HARD, NASTY. 20:36
I'm surprised to see &eval and BEGIN both fall under NASTY. 20:37
tadzik I like how this looks like „that's easy. Do it”
20:37 zby_home_ joined
sorear masak: NASTY means basically "I don't understand this well enough to even start" 20:38
the other three are mostly broken down by time required 20:39
EASYs will take me less than an hour if I'm in a proper flow state
moritz_ masak: I guess both BEGIN and eval carry context between run- and compile time
which makes it hard
sorear exactly
jnthn considers BEGIN decidedly hard and scary 20:40
sorear eval is probably the easier of the two, since it doesn't depend on the immediate metamodel and doesn't require general serialization
20:41 stifynsemons joined, colbseton left
jnthn sorear: I don't consider eval at all easy, but I think it's a lot less hassle to do right than BEGIN is. 20:42
moritz_ an approximation of eval is much better than nothing. 20:44
jnthn True. Rakudo had an eval that couldn't see outer lexicals for many months.
moritz_ and it was enough to pass quite some spectests
jnthn Right. 20:45
sorear I could do *that* right now 20:46
sorear adds to agenda 20:47
moritz_ sorear: that would be enough for eval_{lives,dies}_ok
which evalutes in context of Test.pm anyway, not in the context of the test file 20:48
sorear: another TODO: MAIN subs 20:53
niecza: say %*ENV<PATH> 20:54
masak moritz_, sorear: thanks for explanations about why BEGIN and &eval are nasty. what you say matches what I've been thinking about them so far for Yapsi.
p6eval niecza v2-42-g311ba52: OUTPUT«␤»
moritz_ perl6: print %*ENV<PATH> 20:55
p6eval niecza v2-42-g311ba52: ( no output )
..rakudo : OUTPUT«/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games»
..pugs: OUTPUT«*** Can't modify constant item: VUndef␤ at /tmp/q2nhWB6tGn line 1, column 7 - line 2, column 1␤ /tmp/q2nhWB6tGn line 1, column 7 - line 2, column 1␤»
moritz_ another TODO
masak ...already in there. 20:56
moritz_ in the issue tracker?
sorear: I want run() ... and preferably not in SAFE.setting, but in a FULL.setting or so :-) 21:00
sorear right. FULL is spelled CORE for now 21:04
I'm not 100% happy with this factoring
niecza: say ?(slurp "/etc/passwd") 21:05
p6eval niecza v2-42-g311ba52: OUTPUT«===SORRY!===␤␤Undeclared routine:␤ 'slurp' used at line Any()1␤␤Unhandled exception: Check failed␤␤ at /home/p6eval/niecza/boot/lib/SAFE.setting line 377 (SAFE die @ 2)␤ at /home/p6eval/niecza/src/STD.pm6 line 1148 (STD P6.comp_unit @ 73)␤ at 21:06
../home/p6eval/niecz…
sorear goodie, p6eval is using it
stifynsemons Another newbie question: my %g; %g<gamma>.[].push(5) doesn't auto-vivify hash element gamma. So I have to use my %g; %g<gamma>=[]; %g<gamma>.[].push(5). Is there a simpler syntax?
moritz_ sorear: so how do I run neicza with CORE instead SAFE 21:07
sorear moritz_: don't pass --safe to the executable
stifynsemons: I suspect push %g<gamma>, 5 should work
stifynsemons: generally you shouldn't use .[].; that overload exists for the sake of string interpolation 21:08
moritz_ what's wrong with .[] ?
stifynsemons push %g<gamma>, 5 doesn't work in 201101 rakudo star 21:09
sorear moritz_: it's a no-op, except when it's not
(%g<gamma> //= []).push(5) will definitely work 21:10
masak rakudo: my %g; push $g<gamme>, 5; say %g.perl
sorear but is rather on the ugly side
p6eval rakudo : OUTPUT«===SORRY!===␤Symbol '$g' not predeclared in <anonymous> (/tmp/yn4U_mE8fT:22)␤»
masak rakudo: my %g; push %g<gamme>, 5; say %g.perl
p6eval rakudo : OUTPUT«Method '!fill' not found for invocant of class ''␤ in 'List::push' at line 2881:CORE.setting␤ in main program body at line 22:/tmp/l0BsSJw77w␤»
moritz_ I'm pretty sure both the sub and method form should work
masak is surprised the spectests wouldn't pick something like that up.
masak submits rakudobug 21:11
stifynsemons++
moritz_ masak: there are tests for that. They are just fudged :-)
stifynsemons thanks
ok, for now, I'll just %<gamma>.[].push(5) for now :-)
masak moritz_: right, but ISTR this *used* to work. have they been fudged recently?
stifynsemons: leave comments with the RT ticket number in; that way, you'll remember to improve the code when the bug is fixed. 21:12
sorear in my family, "fudge" is considered a low-tier swear word. Makes reading about spectests a little amusing. :)
moritz_ rakudo: my %h; %h<a>[0] = 3; say %h.perl
p6eval rakudo : OUTPUT«{"a" => [3]}␤»
moritz_ masak: I remember *that* working
I didn't fudge any autoviv tests lately 21:13
alpha: my %a; push %a<b>, 3; say %a.perl 21:14
p6eval alpha : OUTPUT«No applicable candidates found to dispatch to for 'push'␤in Main (file <unknown>, line <unknown>)␤»
moritz_ -> sleep 21:15
masak 'night, moritz_. dream of excellent autovivification.
lue I am beginning the beating process on XeLaTeX, albeit only in my head :) . afk. 21:19
sorear TimToady: ping. 21:21
TimToady: suppose I have a given statement with 256 whens, each labelled when 0, when 1, ... when 255; obviously this should be a jump table 21:22
TimToady: suppose further that the 63 case ends with $_ = 195; proceed;
TimToady: what should happen? does it make sense to use the jump table here? how?
TimToady: what if it were $_ = 16; proceed; ?
masak interested question: is there a nice Perl or Unix tool that would allow me to start two processes and have the output of one fed as input to the other? I'm asking for p4. 21:23
sorear masak: man 1 sh
sh -c 'program1 | program2'
masak ok, I was unclear. 21:24
they reciprocally feed input to each other.
they are "talking" to each other.
I might need to do some minimal processing in the middle, too. at least to get things started.
sorear you're asking for deadlocks if you do that because of buffering semantics
masak yes, I gathered as much from reading 'perldoc -f pipe'. 21:25
sorear there is an easyish way though
masak I don't need it to be fast, just reliable.
best idea I have so far is to re-run things lots of time, each time a little further. 21:26
that's risky if there's randomness involved, which there might be.
sorear mkfifo $TMP; <$TMP a | b >$TMP
21:28 dukeleto left, dukeleto joined
masak hm... 21:28
with additional piping, that might fly. 21:30
sorear++
21:42 y3llow_ joined, pothos_ joined, plobsing left 21:43 y3llow left 21:44 y3llow_ is now known as y3llow, pothos left, pothos_ is now known as pothos
stifynsemons I noticed that there is implicit whitespace in a grammar like: grammar E { rule TOP { ^[ <a> <b> ]$} token a { a } token b { b} } 21:59
so that E.parse("a b") matches but E.parse("ab") doesn't.
sorear that's because you used "rule" 22:00
stifynsemons is this the way it is supposed to be? I couldn't tell from the synopsis.
sorear if you had used token TOP, there'd be no issue
stifynsemons so if a rule includes a token, it puts space between them?
sorear it's even simpler than that 22:01
every " " in a rule becomes <.ws>
stifynsemons oh, that is simple
masak and <.ws> *requires* whitespace between \w things.
stifynsemons that's even what I want :-)
masak ...due to the way we most often want identifiers to work in languages. 22:02
stifynsemons I assume that multiple spaces maps to a single <.ws> as well
masak yes. 22:03
22:10 coldhead joined, plobsing joined 22:11 HaraldJoerg joined
sorear well, it does, but that's just an optimization 22:12
<.ws> is defined such that <.ws><.ws> === <.ws>
don't try <.ws>+ though; that will hang
22:20 aindilis left
masak trying to pipe two p4 algorithms together is the most confusing thing I've attempted in a long, long time. 22:25
diakopter <.ws>+ should probably be fixed not to hang.. 22:28
masak +1
sorear -1
22:29 plobsing left
masak *all* quantified empty things should be fixed not to hang. 22:29
that's what Perl 5 does.
diakopter how about a warning then
sorear deliberately introducing inconsistencies into a programming language to make one edge case nicer is never a good idea
masak: ? 22:30
masak <.ws>+ is a special case of [<something-that-matches-empty-string>]+
sorear if you can make this sound consistant, I'll support t
mm
masak it's just good customer service not to hang in that case.
it can be easily detected, and it's never what the user wants.
sorear given "" ~~ /<ws>+/; what is +$<ws> ? 22:32
given "" ~~ /<ws>?/; what is +$<ws> ?
masak in fact, a Thompson engine will *have* to address this case, since it's founded on having to progress one character with each iteration.
and LTM, while not exactly a Thompson engine, expresses the same kinds of sentiment. 22:33
sorear niecza's LTM Thompson engine does handle this case 22:35
22:35 kfo joined
sorear I still like the idea of doing logic programming using regexes that never match anything 22:35
masak it's working! \o/ \o/ \o/ 22:38
sorear++
sorear what did I do?
masak I had a bit of a problem with flushing, but ++$| solved that.
sorear: you gave the clue about a named FIFO.
sorear: man, you should see this. the algorithms are actually playing each other.
22:39 kfo_ left
jnthn Wow! 22:39
masak now I just need to do a harness script that does all-against-all over a range of "board" sizes. and collects statistics. 22:40
jnthn That'll heat the CPU some. Good job it's winter. :-) 22:41
masak :)
sorear clearly, the next step is genetic algorithms. 22:44
with a simulated tournament for fitness ranking
masak look! gist.github.com/824207 22:45
sorear: it would be, if p4 wasn't a solved problem. :)
Tene masak: depends on what metric you're measuring fitness by ;) 22:47
least cpu time used to win, perhaps? 22:48
masak ooh
still, not even that's too hard when one knows the answer.
would be more interesting, I think, with an intractable problem. 22:49
Tene Yes, rather.
22:53 Trashlord left 22:54 Trashlord joined
pmurias masak: we can't reliably determin if a regex matches an empty string 22:56
masak at runtime, we can. 22:57
and that's what matters.
pmurias in some cases we can in others not
masak "am I in a quantifier? will the next iteration take me exactly nowhere? exit loop."
pmurias and what do we do succed/fail? 22:58
22:58 bacek left
masak I don't understand the question. the quantifier succeeds. 22:58
pmurias no it doesn't
masak I see.
Tene it... doesn't? 22:59
22:59 JodaZ left
Tene Why not? 22:59
pmurias because it doesn't terminate
Tene a* should match the empty string successfully.
22:59 bacek joined
pmurias (a*)* would hang 22:59
pmurias thinks 23:00
masak exactly the (a*)* case is used as an example with the Thompson engine. 23:01
pmurias (a*)* works in Perl 5
masak the general situation is 'a quantifier with something that (as a default) matches the empty string).
s/\)./'./
23:03 autark joined, kjeldahl left
pmurias we could define the quantifers that way that they stop when the thing doesn't match anything 23:03
masak I now have the harness playing every algorithm against every algorithm, for game sizes between 6 and 20. :)
pmurias: yes, that's what I'm saying. 23:04
'night, #perl6. 23:06
23:06 masak left
colomon masak++ 23:08
though 20 is nothing....
23:11 HaraldJoerg left
colomon thinks his algorithm should do pretty well until about 60 or so (barring bugs that slipped through testing, of course). 23:14
23:16 autark left 23:19 HaraldJoerg joined 23:24 plobsing joined 23:25 HaraldJoerg left
dalek p: d6338b2 | jnthn++ | src/NQP/Actions.pm:
Toss 'is pirflags' and add 'is vtable'. Calls method on the meta-object, though none implement it yet. Eventually this allows overriding of some Parrot v-table methods again when using 6model, which we'll need for other Parrot bits to interop with languages using 6model.
 23:29
p: a022547 | jnthn++ | t/nqp/52-vtable.t:
Update test file to use is parrot_vtable rather than is pirflags. Still doesn't pass again yet, though.
23:30 drbean joined
sorear jnthn: how long until it will be possible to have other parrot languages using 6model objects? 23:31
jnthn sorear: Well, in a sense, they already could, if they're prepared to use the new nqp already. 23:32
23:32 MayDaniel joined
jnthn sorear: I think the main blocker is probably documentation and examples of doing stuff other than what NQP does with it. 23:32
sorear will it eventually be possible to use new objects without nqp code? 23:33
jnthn Yeah.
The weaving at the moment is one of developer (e.g. my ;)) convenience than need. 23:34
sorear Dedenting heredocs turns out to be less trivial than I thought.
jnthn You can already do everything in PIR with 6model that you can do from NQP, so any other languaes that can generate such PIR should do just fine. And since doing most things with 6model is method calls anyway...it shouldn't be so bad. 23:35
sorear perl6: my $x = 1; say "$x".WHAT
p6eval pugs: OUTPUT«Str␤»
..rakudo : OUTPUT«Str()␤»
..niecza v2-42-g311ba52: OUTPUT«␤»
sorear niecza: my $x = 1; say "$x".typename
p6eval niecza v2-42-g311ba52: OUTPUT«Num␤»
jnthn o.O 23:36
23:36 shi joined
sorear jnthn: niecza treats "$x" as just $x 23:36
jnthn Ah :)
sorear wondering if maybe it... shouldn't 23:37
otoh, "foo" really shouldn't call &infix:<~>
stifynsemons how does one delete hash keys? 23:40
TimToady by spec, %hash<key>:delete, but that's not in rakudo yet 23:43
sorear %hash{"key"}:delete
TimToady %hash.delete('key') in rakudo
sorear TimToady: is the latter spec? does niecza need to support it?
jnthn knows why we have the %hash<key>:delete, but his brane still refuses to find it attractive...
stifynsemons thanks 23:44
TimToady sorear: no, not spec 23:47
jnthn The adverb way will be the only one? :( 23:48
TimToady please tell me how .delete(Str) can possibly be a degenerate case of a multidimensional list of slices 23:51
jnthn Hmm. I guess I'm not arguing for keeping .delete(Str) really...I've just never really managed to like the :delete approach.
Maybe it feels too end-weighty. 23:52
Maybe I'll get used to it with time.
TimToady delete is supposed to return the deleted elements exactly as if it were subscripted anyway
sorear FWIW, niecza has a .delete-key which is used by postcircumfix:<{ }>
jnthn Just one of those rare-ish cases where something in Perl 6 grates a bit. It may just be me.
TimToady I don't mind internal methods
sorear although it's slightly misnamed now that it accepts slices (or at least its sister at-key does) 23:53
TimToady: did you see my question about jumptables in the backlog? 23:59

Powered by Cro and the Raku® Programming Language.

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

Thank you!