»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.
00:05 fgomez joined
colomon timotimo: gist the code? 00:09
00:24 snearch left 00:33 Guest29702 left 00:59 mucker joined
timotimo i'm a bit ashamed of it :| 01:00
and the linear algebra portion is hardly self-explanatory :)
colomon are you making any progress debugging it? 3D lines intersection? 01:01
TimToady sigh, is-prime seems to work correctly up to base 11, but uses up 15GB trying to do base 12, which is only about 800th the size of the base 18 case, which will obviously be hopeless 01:15
(in rakudo here)
17000 big integers should not take 15GB of memory... 01:17
is-prime is not what is chewing up the memory, just to be clear about; moritz++ 01:21
*about that
01:28 whiteknight left
colomon depends on how big, no? 01:29
TimToady is there a way to force a GC run from within rakudo? 01:31
the following appears to grow continually: 01:32
r: my @current; loop { my @next = (1..100000).roll(10000); @current := @next; 0; }
should timeout... 01:33
p6eval rakudo 2e3365: OUTPUT«(timeout)»
au drop to PIR level and run "sweep 1"?
TimToady it doesn't grow if I take out the @current := @next 01:35
MikeFair_ Hey all 01:37
TimToady: Does it grow if you take out the 0; at the end? (that zero might prevent tail recursion detection) 01:38
TimToady: Or is it needed
TimToady: Or do you think it might the "my" inside the loop
raiph r: my @current; my @next; loop { @next = (1..100000).roll(10000); @current := @next; } 01:39
p6eval rakudo 2e3365: OUTPUT«(timeout)»
MikeFair_ hehe - go raiph
raiph r: my @current; my @next; loop { @next = (1..10).roll(10); @current := @next; }
p6eval rakudo 2e3365: OUTPUT«(timeout)» 01:40
TimToady well, I was trying to avoid something collecting the results of the loop, but I shoulda put the 0 on the outside to do that
but even an explicit sink on the loop doesn't prevent the leek
*leak
nor does the sweep 1, au++
MikeFair_ That last example from raiph is shocking to me
01:41 FROGGS_ joined
MikeFair_ one iteration times out?? 01:41
raiph r: my @current; my @next; loop { @next = (1..1).roll(1); @current := @next; }
TimToady no, the loop times out
colomon MikeFair_: loop is infinite loop
p6eval rakudo 2e3365: OUTPUT«(timeout)»
MikeFair_ colomon: oh, duh
TimToady you can't see the memory growth with evalbot, you have to try it on your own machine 01:42
MikeFair_ colomon: Thx
TimToady dunno if this is the leak that is causing my 15GB, but it could be, since I have the same sort of "last time through" array 01:43
also leaks with =, not just with :=
01:44 flightrecorder left, FROGGS left 01:45 flightrecorder joined
TimToady afk & 01:45
02:05 flightrecorder left, FROGGS_ left 02:14 HarryS left 02:21 kurahaupo left 02:22 PacoAir left 02:26 FROGGS joined 02:29 flightrecorder joined 02:36 am0c joined 02:42 am0c left, am0c joined
dalek ast: 6aff7bd | (Solomon Foster)++ | S32-num/expmod.t:
Add some basic tests for expmod.
 02:45
02:46 FROGGS left
colomon TimToady: is the expmod in rc yours? 02:46
02:47 flightrecorder left 02:50 yarp joined
dalek ecza: 1760451 | (Solomon Foster)++ | / (2 files):
Borrow TimToady++'s implementation of expmod.
 02:50
TimToady colomon: I believe so. 02:51
colomon That was my conclusion, as you can see. :) 02:52
sorear n: (1 .. 10**1000).roll.say 02:54
p6eval niecza v21-15-gf226bf6: OUTPUT«Inf␤»
sorear n: say 10**1000
p6eval niecza v21-15-gf226bf6: OUTPUT«1000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000…
sorear n: (1 .. 10**100).roll.say
p6eval niecza v21-15-gf226bf6: OUTPUT«2133311737390845811379582797551885492279278014655713661890031696021502921418274361246204385883586561␤»
colomon sorear! \o/ 02:55
02:55 FROGGS joined
colomon oooo, bug 02:55
sorear colomon: want to take a crack at en.wikipedia.org/wiki/Miller%E2%80...ality_test ? If you can get Range.roll working properly for bignums, it will be trivial
colomon that one probably *is* my fault.
02:56 flightrecorder joined
colomon sorear: I was planning on just importing TimToady's version as a first step. 02:56
but not tonight, I should be getting to bed.
sorear sleep well 02:59
colomon is in a great mood tonight, because someone posted "John McKenna: His Original Recordings" and "Fluters of Old Erin" to the web in their entirety. As a result, been listening to nothing but old 78 recordings all evening.
03:06 fgomez left 03:07 yarp left 03:09 FROGGS left, flightrecorder left
[Coke] wonders how old you have to be to know what a 78 is. 03:10
too young, they're from "1978". ;) 03:11
TimToady 45s are the thing now 03:12
03:13 yarp joined 03:16 benabik left 03:17 benabik joined
sorear [Coke]: they're round things that spin at 78rpm and have a groove on them encoding audio, from back in the sad days before people discovered audio could be encoded as 1s and 0s 03:18
MikeFair_ sorear: And even before '8-Tracks 03:21
03:23 fgomez joined
TimToady remembers when 33⅓ was introduced... 03:24
03:31 mucker left
TimToady and the very short reign of "HiFi" before "Stereo" became the rage 03:32
03:33 FROGGS joined
benabik wonders how many people think "HiFi" is a new kind of "WiFi" 03:33
MikeFair_ benabik: HiFi clearly must be the next gen thing after "BluRay" :) 03:41
beyond HD 03:42
03:45 FROGGS left 03:58 FROGGS joined 04:35 FROGGS left 04:36 FROGGS joined 04:46 thou joined 04:48 flightrecorder joined 05:09 raiph left
TimToady aside from the minor memory leak above, I find out why it's chewing up memory; last statement of two nested loops was a push 05:26
05:27 FROGGS left, flightrecorder left
TimToady rakudo doesn't realize the for loops are in sink context, so collects all those pushes for the current base, and doesn't realize it doesn't need them 05:27
when I change the inner push to a gather/take (or put a () after the push), it doesn't grow huge
moral: sink context can be really important for avoiding unnecessary work 05:28
the other weird thing I've figured out is that libtommath's is_prime is mostly just doing trial division by the first 256 primes, so it almost never gets into the Miller-Rabin part 05:30
I kinda wish there was a way to bypass that trial division...
when it does get into Miller-Rabin, the "random" numbers it chooses are the first N primes from the very same table...not sure if that is a good thing, numerically speaking... 05:31
so, like, if you say is-primes(1), it always uses 2 05:35
sorear probably not safe, but probably not an issue
05:35 FROGGS joined 05:36 flightrecorder joined
sorear Kolmogorov says almost all numbers are random 05:36
TimToady er, I don't follow; if it's not safe, how is it not an issue?
sorear I mean there's probably some number that will be falsely passed by tommath is_prime 100% of the time 05:37
but I'm never going to find it unless you can hook me up with an NP-oracle
TimToady one can never find that darn NP-oracle when you need her... 05:38
05:39 jaldhar joined 05:50 popl left, Chillance left
TimToady nr: print "1234"; $*OUT.flush; exit 05:51
p6eval niecza v21-16-g1760451: OUTPUT«1234Unhandled exception: Unable to resolve method flush in type TextWriter␤ at /tmp/oqg91Uva1i line 1 (mainline @ 5) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4155 (ANON @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4156 (module-CORE @ 572) ␤…
..rakudo 2e3365: OUTPUT«No such method 'flush' for invocant of type 'IO'␤ in block at /tmp/dT1bU9XBrY:1␤␤1234»
TimToady doesn't seem to be any concept of autoflush either 05:53
(and prompt still doesn't flush either in rakudo on my box, which makes it next to useless) 05:56
moritz \o morning 06:05
phenny moritz: 21 Sep 22:05Z <raiph> tell moritz there's a typo in github.com/moritz/ilbot/blob/maste...e/day.tmpl -- "<TMPL_UNLESS NAME=IS_SUMAMRY>" tried to fix and pull request via github but there's something going on that seems to break when i try.
06:09 am0c left
TimToady o/ 06:09
06:09 am0c joined
moritz phenny: tell raiph thanks, fixed 06:09
phenny moritz: I'll pass that on when raiph is around.
06:09 flightrecorder left
TimToady is-prime seems to be working pretty well 06:11
06:11 FROGGS left
moritz \o/ 06:14
except for negative numbers
r: say (-2).is-prime
p6eval rakudo 2e3365: OUTPUT«False␤»
moritz should be True, no?
TimToady uh, why? 06:15
I've never heard of negative primes...
moritz because -2 is only divisble by -2 and 1
ok, I'm wrong
never mind then
anyway, I'll do the sub forms next 06:16
should they coerce from Real and Cool?
TimToady well, I guess 06:17
moritz oh, one more question
the spec lists many multis, but usually omits the protos
TimToady though you almost want a coercion that will fail if the Real isn't an integer
moritz so the question is, should something like proto sub chr have a signature of ($) or (|) ? 06:18
thing is, (|) makes it much more extensible, if you ever want that
but it means that saying map &chr, @list; will not work as you might wish it to
because if the proto is (|), then &chr accepts arbitrarily many arguments 06:19
TimToady my inclination has been to have the more restrictive protos, since you can always shadow it with your own proto if you want to extend
and maybe that gives more for the optimizer to work with too 06:20
moritz ok
and if I shadow a proto, how do I dispatch to the original one? with OUTER?
TimToady maybe nextsame?
OUTER would probably work 06:21
06:21 yarp left
moritz considers nextsame/nextwith 06:21
TimToady presumably it'd have to be called directly from the shadowing proto 06:22
except, of course, that the proto has been compiled as an "only", so the outer proto doesn't really know about it 06:23
moritz regarding the more restrictive protos: currently it'll slow down things a bit 06:24
(I think)
because it means the signature of the proto and of the multi need to be checked
TimToady could be, unless we can use that to factor out checks from all the multis 06:25
or we figure out to ignore a proto sig that's redundant with its multis 06:26
moritz maybe if the optimizer can proof that all the candidate's signatures imply the same restriction as the proto
TimToady then it's just there for arity
but if it's just for arity, then we could probably deduce that from the multis as well 06:27
that would lead to instability in the face of imports, however
moritz arit yand autothreading 06:28
TimToady since an import could add args
anyway, I'm okay with protos being specific about arity in the core langauge, I think, and let the people who want to override that work a little harder 06:29
moritz ok 06:30
TimToady at least for the functions it makes sense to have an arity for
moritz I'll do it that way, and write a deprecation notice for the next release
sure
TimToady long term, I'd like to get to the point where introspecting arity is considered kinda wrongish 06:31
and we just bind it and see how many args actually get bound
looking at arity is kinda like checking eof
06:31 wk joined 06:35 yarp joined
cognominal r: sub a { say 'called' }; a{}; a.{}; a[]; a.[] 06:40
p6eval rakudo 2e3365: OUTPUT«called␤called␤called␤called␤»
sorear moritz: -2 is a prime in ring theory, but not in number theory :D 06:41
cognominal r: sub a { say 'called' }; a.()
p6eval rakudo 2e3365: OUTPUT«called␤No such method 'postcircumfix:<( )>' for invocant of type 'Bool'␤ in at src/gen/BOOTSTRAP.pm:825␤ in any at src/gen/BOOTSTRAP.pm:809␤ in block at /tmp/B8yI57zATl:1␤␤»
cognominal weird
TimToady r: sub a { say 'called' }; &a.()
p6eval rakudo 2e3365: OUTPUT«called␤»
TimToady as a special form, a only takes (), not .() 06:42
sorear in the more general context, it doesn't make sense to restrict to positive elements, because elements of a general ring don't have a notion of positiveness
06:42 Celelibi left
sorear but you still have primes in a sense of non-factorizability: x is prime if you cannot write x = y*z where y and z are non-invertable 06:42
(I know I'm being loose with the distinction between primes and irreducables) 06:43
cognominal apparently that channel has mutated to number theory
TimToady
.oO(one ring to rule them all)
06:43 FROGGS joined
cognominal by that Tolkien, you can unite many things. 06:45
06:49 flightrecorder joined
TimToady is wondering how niecza manages to avoid flushing problems without mentioning 'flush' anywhere 06:51
sorear TimToady: MicrosoftCamelCaseNaming 06:52
TimToady ah
stdout.AutoFlush = true; return stdout; 06:53
diakopter have your plunger handy
cognominal TimToady should learn to be insensitive 06:54
…in that case
TimToady will continue to run his quiz editor in niecza till rakudo learns to flush or Flush or whatever... 06:55
(quiz tomorrow, y'see)
moritz actually we have a pull request for something like that
sorear r: say $*OUT.^methods(:local)
p6eval rakudo 2e3365: OUTPUT«open close eof get getc lines read seek tell write opened t print slurp copy chmod Str d e f l r s w x z modified accessed changed say <anon> <anon> <anon>␤»
TimToady there's no flush in CORE 06:56
moritz TimToady: github.com/rakudo/rakudo/pull/71
TimToady and parrot only knows about "autoflush" on close, which seems like it shouldn't be optional...
moritz TimToady: a comment on the proposed API would be welcome
sorear
.oO( implement flush with ITS semantics )
 06:57
dalek kudo/nom: e095de4 | moritz++ | / (2 files):
sub forms of is-prime and expmod
moritz (you have to check the patch for that)
sorear "method DESTROY()" is obviously wrong 06:58
TimToady well, regardless of the merits of autoflushing on close, an autoflushed filehandle should be flushing it after every output 06:59
the whole un/line/full-buffering underneath that is orthogonal 07:00
moritz I thought unbuffering had the same effect as flushing
TimToady and prompt() probably needs an explicit flush in any case, unless the handle is already autoflushed 07:01
no
autoflushing is "command buffering"
buffer till we get done with this command, then flush
unbuffering usually means put out each character as generated
the distinction is probably not meaningful at the Perl 6 level however 07:02
it's the lower-level routines that would be sending every char directly to output
stderr is historically unbuffered
if your smallest command outputs a string as 'print' does, then you can't really know whether the STDIO layer is putting it out one char at a time, or waiting till the end and autoflushing 07:04
though the latter is considerably more efficient for multichar strings
mostly Perl ignores the STDIO-level distinction of un/line/full and just worries about autoflush or not 07:05
diakopter Perl including 6? 07:06
TimToady "mostly"
diakopter heh
TimToady meaning we want to provide ways to set the underlying strata, but not force people to worry about them most of the time
and that is the same for 6
sorear note that perl 5 mostly doesn't use stdio these days 07:08
TimToady yes, well...line buffering was always kind of a sad proxy for command buffering 07:09
moritz öööööööööööööööööl 5~
sorear Good morning Ronja
07:10 thou left
TimToady hopes that wasn't moritz's keyboard shorting out on puke 07:10
I also note that rakudo uses line buffering on output of $*OUT to a disk file, which could be construed as inefficient 07:11
maybe it doesn't matter so much with modern OSes though 07:12
back in the day when one VAX syscall could cost a million instructions, you tried to avoid those... 07:13
well, 10,000 anyway
diakopter I was gonna say; 1 second
TimToady whatever the number was for a minimal system call, it was significant
sorear it's still fairly bad 07:14
trapping from user mode into supervisor mode produces a rather big pipeline stall 07:15
I imagine SYSENTER et al improved things, but probably it still hurts
TimToady so we might want to look at the default buffering characteristics of output files wrt filesystems vs terminals at some point 07:16
well, looks like rakudo might get through the base 18 case in a day or so 07:17
though due to the lack of autoflushing I have to print out each progress report on separate lines...which is what originally inspired this rant :) 07:18
07:20 kaare_ joined
TimToady but it's 2/3 of the way to the high-water mark, and only using 40% of my memory, so will likely fit, unless everything doubles when it exceeds 64bit integers 07:21
diakopter quick buy more memory
TimToady I suspect the niecza blowup is related only to the size of the bigints; it actually got past the high water with less memory than rakudo is already using, but blew up partway down the descending part of the number-of-candidates curve when ints get really long 07:23
(that actual answer is supposed to be 55 digits decimal)
diakopter gets curious how perlesque would do 07:24
TimToady so probably tickling a bug in microsoft's bigint routines, but hard to reproduce, since it takes more than a day of compute time 07:25
07:26 spider-mario joined
cognominal r: $/ = 'foo' 07:28
p6eval rakudo e095de: ( no output )
cognominal TimToady, is there any use to accept $/ set to values that are not Matches? 07:29
this invite misuses
TimToady one of the tests sets it to Any, which gives STD fits 07:30
since it then thinks it's P5 code
n: $/ = 'foo' 07:31
p6eval niecza v21-16-g1760451: OUTPUT«===SORRY!===␤␤Unsupported use of $/ variable as input record separator; in Perl 6 please use the filehandle's :irs attribute at /tmp/kNtgJyozbs line 1:␤------> $/⏏ = 'foo'␤␤Parse failed␤␤»
cognominal I can understand that it is set explictely, preferably with :=, to previous matches.
07:33 FROGGS left, flightrecorder left, Psyche^ joined 07:37 Patterner left, Psyche^ is now known as Patterner 07:44 M_o_C joined
TimToady zzz & 07:45
07:46 cognominal left 07:47 cognominal joined 07:51 pmurias joined, birdwindupbird joined
pmurias jnthn: ping 07:51
07:53 cognominal left 07:59 cognominal joined 08:01 skids left, HarryS joined 08:08 cognominal left
moritz nqp: say(pir::typeof__SP(pir::getstdout__P())) 08:15
p6eval nqp: OUTPUT«FileHandle␤»
08:16 cognominal joined
pmurias should basic IO be moved into nqp rather than escaping to parrot? 08:17
moritz interesting question 08:18
I'd say +1, but I'd also wait for jnthn & pmichaud
08:18 FROGGS joined
FROGGS good morning 08:18
08:19 wamba joined 08:21 flightrecorder joined 08:23 cognominal_ joined
dalek kudo/nom: b72c434 | moritz++ | src/core/IO.pm:
flush $*OUT in &prompt
 08:24
moritz \o FROGGS
FROGGS moin moritz 08:25
moritz wow, $*IN.get seems to be broken
it doesn't return after reading one line
pmurias it's not tested?
moritz well
08:25 cognominal left
moritz it's tested, but only with streams that produces an EOF eventually 08:26
FROGGS btw, I gave up trying to get multi trait symbol importing working, I spend four days debugging World.pm and friends, but I cant find the right place where I have to tweak something
moritz :( 08:27
.oO( the World.pm is not enough )
FROGGS ya, looked at traits.pm too, there I can see the export groups 08:28
masak morning, #perl6
FROGGS but what is calling into World.pm? do I have to it the dispatcher? or earlier?
morning masak
moritz \o masak
FROGGS: Actions.pm, Grammar.pm and some few built-ins (src/core/*) call into World.pm 08:29
FROGGS: you can juts nqp::die() in there and look at the backtrace
FROGGS its a bit complex though
maybe a bit too complex for me :/ 08:30
moritz any part which you're particularly interested in?
I don't know all of World.pm, but I did work with importing a bit
08:31 GlitchMr joined
FROGGS moritz: in a module I use NativeCall together with Inline.pm 08:31
both export trait_mod:<is>
if I remove the "die, duplicate symbol trait_mod:<is>" I still dont have both symbols in the dispatcher list 08:32
NativeCall exports "is natice", "is symbol" and another one I dont know right now, my Inline.pm exports "is inline" 08:33
moritz X::Export::NameClash.new(symbol => exp_name).throw # this one?
FROGGS this: self.throw($/, 'X::Import::Redeclaration', 08:34
symbols => @clash,
source-package-name => $source_package_name,
);
moritz ah, ok
all 'is' multis share the same routine name
FROGGS the list of non-matching signatures shows only the trait_mod:<is>'s from the module that was imported at last
ya, &trait_mod:<is> 08:35
08:35 birdwindupbird left
FROGGS my problem is: where is this signature-list? 08:35
moritz that I can answer, I hope 08:36
FROGGS cool
moritz if you can get your hands on the routine object, you can call .candidates
and that's a list of candidates, and each has a .signature
though you don't need the signatures
FROGGS how can I print information about the candidates? 08:37
moritz r: say &infix:<+>.candidates.perl
p6eval rakudo e095de: OUTPUT«(sub infix:<+>($x = { ... }) { ... }, sub infix:<+>(\a, \b) { ... }, sub infix:<+>(Real \a, Real \b) { ... }, sub infix:<+>(Int:D \a, Int:D \b) { ... }, sub infix:<+>(int $a, int $b) { ... }, sub infix:<+>(Num:D \a, Num:D \b) { ... }, sub infix:<+>(num $a, num $b) …
FROGGS do I have .perl in World.pm? thats nqp, right?
moritz the object is a Perl 6 Routine
so you can call .perl
FROGGS k
moritz FROGGS: look at src/core/Routine.pm for what a Routine supports 08:38
r: sub f() { }; say &f.is_dispatcher
p6eval rakudo e095de: OUTPUT«False␤»
FROGGS k 08:39
moritz FROGGS: I think the merging should only happen if both routines have .is_dispatcher return True
r: proto f() { }; say &f.is_dispatcher
p6eval rakudo e095de: OUTPUT«True␤»
moritz r: multi f() { }; say &f.is_dispatcher
p6eval rakudo e095de: OUTPUT«True␤»
FROGGS ahh, cool
r: &trait_mod:<is>.is_dispatcher.say 08:40
p6eval rakudo e095de: OUTPUT«True␤»
moritz one way forward is to create a new proto, and attach the candidates from both conflicting objects
as a first step
in the long run, I think that we're still missing a piece, but I'll have to discuss that with jnthn 08:41
I think it should be a difference whether you export a proto or a candidate
and candidates should be auto-merged, but protos only if they are onlystar 08:42
08:43 cognominal_ left
moritz but don't worry about that yet, that's a topic for a future development 08:43
08:44 cognominal joined
FROGGS k 08:45
moritz r: say &trait_mod:<is>.onlystar 08:46
p6eval rakudo e095de: OUTPUT«False␤»
moritz r: proto foo(|) { * }; say &foo.onlystar 08:47
p6eval rakudo e095de: OUTPUT«True␤»
masak r: say {;}.is_dispatcher 08:48
p6eval rakudo e095de: OUTPUT«Can not get attribute '$!dispatchees' declared in class 'Routine' with this object␤ in any at src/gen/BOOTSTRAP.pm:579␤ in block at /tmp/Wu8lQnVML4:1␤␤»
masak hm, LTA.
moritz jnthn: ^^ why isn't &trait_mod:<is>.onlystar True? the declaration is proto trait_mod:<is>(|) { * }
masak r: say {;}.onlystar
p6eval rakudo e095de: OUTPUT«No such method 'onlystar' for invocant of type 'Block'␤ in block at /tmp/GvB4eQ4kNn:1␤␤»
08:49 cognominal left
moritz it seems that method is_dispatcher is installed in the wrong level 08:49
yes, installed in Code
08:50 cognominal joined
moritz I'll move it into Routine and spectest 08:51
if it doesn't work, we can always have the one in Code return False, and leave the more specific one in Routine 08:52
moritz loves mro
masak mro++ 08:54
moritz++
moritz it compiles with is_dispatcher in Routine 08:55
I don't think we'll see failures in the spectests then
but it's still worth checking 08:56
btw yesterday was my day in Parliament 08:57
I learned a lot
for example that I'm not suited for politics
and that the IT infrastructure of the administration is far worse than I feared
and that flying business class is something I could get used to 08:58
masak :)
FROGGS hehe
moritz the video coverage from the first session that day is already up 09:03
09:03 cognominal left
moritz I guess the one from "my" session will follow soon 09:03
FROGGS post the linky then please 09:04
moritz will do
woah, 1.7GB video for ~2h 09:05
masak it's because Parliamentary deliberations are so information-dense. 09:06
:P
moritz :-)
sorear supposes it was probably not conducted in English 09:07
moritz sorear supposes correctly
masak given that it's Germany...
moritz though the effort to get video coverage was quite impressive
with three professionally-looking cameras 09:08
masak is there an example of a country where Parliament matters are not conducted in one/the official language of the state?
I would think not.
sorear I think that by definition any language used by Parliament is an official language
masak well, for a long time Sweden didn't have an official language. 09:09
sorear although there are some funny places like the US and India where local governments can do business in languages that are not official nationally
masak I think it was like that until 2006.
moritz could imagine that the Finnish Parliament might hear international experts in English
masak oh, sure, hearing international experts in English is probably common. but that's not what I meant. 09:10
09:10 am0c left
moritz right 09:10
well, it was an expert hearing (though not internal), not a debate
09:11 cognominal joined
moritz www.bundestag.de/Mediathek/index.js...ea=details 09:11
FROGGS: ^^ that's the video coverage. I thought it was for the other hearing, but it's actually for "mine" 09:12
sorear masak: what language does the Swedish parliament debate in?
09:13 wk left
masak Swedish. 09:15
we're very English-friendly over here, but not that much ;)
dalek kudo/nom: cfd5e72 | moritz++ | src/Perl6/Metamodel/BOOTSTRAP.pm:
move is_dispatcher from Code to Routine

it did not work in Code anyway, because it uses an attribute from Routine
masak ah. Swedish was declared the "main" language of Sweden on 2009-07-01. 09:17
before that, the joke was that Swedish was one of the officially recognized languages in the European Union because it was an official language in Finland. :)
sorear offically, the US does not have a main language. nobody cared about this until about five years ago 09:18
masak I doubt that nobody cared until five years ago. 09:19
but politics is a sluggish thing, and it takes lots of effort to get any proposed law accepted.
diakopter I doubt that the amount of people who care about it changes much over time
urp 09:21
moritz r: proto a() {*}; multi a() { }; say &a.onlystar 09:22
p6eval rakudo b72c43: OUTPUT«True␤»
sorear whee, finally tracked down the bizarre set bug from this morning. 09:39
will fix it tomorrow
sleep&
FROGGS gnight
masak .oO( "[bizarre set] bug" or "bizarre [set bug]"? ) 09:45
tadzik good morning #perl6 09:59
FROGGS hi tadzik
cognominal sorear, 20 years ago, when I got the driver exam in california, I could have taken it in spanish and some other languages. I was told DMV was a mess. I did everything and got my DMV licence in 3 hours, Visibly, my informers did not ever dealt with French administrations. It was for me a model of efficient administration. Driver licenses in France are just an organized racket, random, painful and costly. 10:00
masak huh! I was contacted about providing advertising space on strangelyconsistent.org. 10:01
my gut reaction is to politely decline. 10:02
cognominal you have polite guts :)
masak well, the email making the request was polite.
10:35 wamba left 10:43 cognominal left 10:49 tgt joined 10:50 MayDaniel joined
jnthn afternoon, #perl6 10:53
tadzik good afternoon jnthn 10:54
so I barely made it in the morning. Awesome :)
moritz masak: you can politely point out that if they are interested in ad space on programming blogs, perlgeek.de/blog-en/ might be an option :-) 10:55
(for text links only)
masak I might do that. haven't replied to them yet. 10:58
they say that the ads would be unobtrusive, and I believe them. 10:59
but I still have a nagging feeling that the extent to which they will annoy/invonvenience *me* (even if slight) will exceed the monetary compensation I will get for it.
that's my rationale for my gut feeling, anyway :) 11:00
11:05 Su-Shee_ joined
dalek ast: f29377c | (Solomon Foster)++ | S32-num/expmod.t:
Add tests for sub form of expmod.
 11:07
moritz wait what, we have tests already? 11:08
colomon++
colomon moritz: I added them so I knew the niecza version was working.
Could probably use more / better tests, but they're a start, anyway.
moritz aye
rakudo passes them \o/ 11:09
colomon \o/
dalek ecza: 83d732c | (Solomon Foster)++ | lib/CORE.setting:
Loosen argument types on Int.expmod, add sub expmod.
11:09 Su-Shee left
dalek kudo/nom: c8b1982 | moritz++ | t/spectest.data:
run expmod tests by colomon++
 11:24
colomon moritz: are there is-prime tests yet? 11:44
11:50 M_o_C left
colomon is now writing is-prime tests 11:51
moritz colomon: no 11:53
diakopter halp. what does this mean: .sort(-*) 11:58
colomon diakopter: sort numerically backwards
diakopter kthxbai
colomon ie sort on numerical negation of each term 11:59
colomon is kind of shocked how fast the pure-perl6 Miller-Rabin is in Niecza. sorear++ TimToady++ 12:11
12:12 birdwindupbird joined
colomon just tested 6864797660130609714981900799081393217269435300143305409394463459185543183397656052122559640661454554977296311391480858037121987999716643812574028291115057151 12:13
dalek ast: 67af080 | (Solomon Foster)++ | S32-num/is-prime.t:
Some basic tests for is-prime.
 12:14
diakopter colomon: how long did that test take
colomon The entire is-prime.t test file (including that one) takes 4.3 seconds to fudge and execute in Niecza.
And I'm not even close to having all the optimizations in I hope to have by the end of today. 12:15
moritz nr: say is-prime 6864797660130609714981900799081393217269435300143305409394463459185543183397656052122559640661454554977296311391480858037121987999716643812574028291115057151 12:16
p6eval niecza v21-17-g83d732c: OUTPUT«===SORRY!===␤␤Undeclared routine:␤ 'is-prime' used at line 1␤␤Unhandled exception: Check failed␤␤ at /home/p6eval/niecza/boot/lib/CORE.setting line 1435 (die @ 5) ␤ at /home/p6eval/niecza/src/STD.pm6 line 1147 (P6.comp_unit @ 37) ␤ at …
..rakudo c8b198: OUTPUT«True␤»
colomon hasn't checked in niecza's is-prime yet. 12:17
moritz fwiw rakudo takes 0.87s for that one-liner here
of which 0.4..0.5s is startup time
diakopter what does .base do 12:18
moritz nr: say 255.base(16)
p6eval rakudo c8b198, niecza v21-17-g83d732c: OUTPUT«FF␤»
pmichaud good morning, #perl6
colomon \o
moritz good am, pm
pmichaud I think I would've expected the isprime opcode to go into rakudo instead of nqp.
moritz pmichaud: well, we have all the libtommath bindings in nqp 12:19
pmichaud looks closer.
moritz and all that nqp does is wrap it, and catch one special case that tommath doesn't handle well 12:20
dalek ecza: ea871cb | (Solomon Foster)++ | lib/CORE.setting:
is-prime borrowed from TimToady++.
ecza: 8e3a364 | (Solomon Foster)++ | t/spectest.data:
Turn on is-prime.t test.
pmichaud okay.
I can go with that, then.
moritz (it thinks 1 is a prime :-)
glad to hear that
jnthn o/ pmichaud 12:21
pmichaud moritz++
dalek kudo/nom: 80f73bb | moritz++ | t/spectest.data:
run is-prime tests by colomon++
 12:23
masak pmichaud! \o/ 12:25
colomon hmmm, finding all primes up to 1,000,000 is slow (niecza) 12:27
making it a good candidate for optimization, I guess. :) 12:28
dalek kudo/nom: f39ec7c | moritz++ | docs/deprecations:
schedule some deprecations for 2012.11
 12:30
spacebat_ star: <one two three>.grep({ "two" eq $_ }) 12:32
p6eval star 2012.07: ( no output )
masak spacebat_: 'say'
moritz star: say <one two three>.grep({ "two" eq $_ }) 12:33
p6eval star 2012.07: OUTPUT«two␤»
spacebat_ d'oh :)
moritz star: say <one two three>.grep('two')
p6eval star 2012.07: OUTPUT«two␤»
moritz grep doesn't just accepts blocks, but anything that you can smartmatch against
masak star: say <one two three>.grep(/^t/)
p6eval star 2012.07: OUTPUT«two three␤»
spacebat_ star: say <one two three>.grep({ "two" ~~ m:i/$_/ })
p6eval star 2012.07: OUTPUT«one two three␤»
masak star: say <one two three>.grep(/w/)
p6eval star 2012.07: OUTPUT«two␤»
masak star: say <one two three>.grep(*.chars > 4) 12:34
p6eval star 2012.07: OUTPUT«three␤»
moritz seems I've forgotten to update p6eval to star 2012.08
but now I'll just wait for this month's star release
who's doing that anyway?
spacebat_ does anyone know why <one two three>.grep({ "two" ~~ m:i/$_/ }) gives <one two three>?
2012.09 now
oh wait 12:35
masak star: say <one two three>.grep({ state $i++ == 1 })
p6eval star 2012.07: OUTPUT«===SORRY!===␤Preceding context expects a term, but found infix + instead␤at /tmp/txSqemsV1J:1␤»
spacebat_ that's just the compiler sorry
masak star: say <one two three>.grep({ (state $i)++ == 1 })
p6eval star 2012.07: OUTPUT«two␤»
masak star: say <one two three>.grep({ (state $)++ == 1 })
p6eval star 2012.07: OUTPUT«two␤»
moritz spacebat_: because ~~ sets $_ of the right-hand side to the value of the left-hand side
r: 'foo' ~~ say $_
p6eval rakudo c8b198: OUTPUT«foo␤»
pmichaud I can likely do the star release again. 12:36
spacebat_ ok guess I need to read up on ~~
pmichaud probably a good idea, since I need to figure out how to update the git subprojects and add that to the docs
spacebat_ I've been reading that recent post about loading the perl6 log with LWP::Simple
moritz ++pmichaud
spacebat_ it has a construct like that
moritz then it's likely wrong 12:37
spacebat_ yeah trying to get my head around it
@strings.grep({ $html ~~ m:i/$_/ }) 12:38
moritz yes, that's wrong
spacebat_ how would you perform that grep/match without clobbering $_
moritz well, first of all I'd this instead: 12:39
my $re = @strings.join('|'); $re = rx/<$re>/ 12:40
and then get rid of the grep entirely
but if you want to do it with grep, you just have to use a different variable than $_ 12:41
spacebat_ fair enough 12:42
moritz @strings.grep(-> $pat { $html ~~ $pat }) 12:43
spacebat_ ~~ as it was introduced into P5 was deemed too magickal and complex
is it more sane in P6?
masak yes.
moritz yes, modulo that one snag that you can't smart-match against $_ 12:44
spacebat_ w00t :)
masak thanks, mainly, to many other things in Perl 6 being more sane.
spacebat_ heh always hit the one snag
or always remember hitting it
moritz the problem with p5 is that it doesn't have a user-exposed type system
but in Perl 6 the operatioin that the smart-match performs is determined by the type
you can see how that can lead to trouble when porting the feature from 6 to 5 12:45
spacebat_ yes
moritz anyway, I'll email szabgab and point him to this discussion
so that he can fix the article
spacebat_ I was going to fix it and post a comment 12:46
moritz you can do that too
spacebat_ :)
moritz mail sent 12:47
GlitchMr > ('ab', 'bd', 'ac', 'def') »~~» /a/ 12:51
True False True False
:-)
But... there is more than one way to do it 12:52
> 'ab', 'bd', 'ac', 'def' X~~ /a/
True False True False
:-)
rn: +'e'.perl.say 12:55
p6eval rakudo c8b198, niecza v21-17-g83d732c: OUTPUT«"e"␤»
GlitchMr rn: (+'e').perl.say
p6eval niecza v21-17-g83d732c: OUTPUT«Unhandled exception: Cannot parse number: e␤ at /home/p6eval/niecza/lib/CORE.setting line 1449 (die @ 5) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3539 (ANON @ 11) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3541 (NumSyntax.str2num @ 5) ␤ at /h…
..rakudo c8b198: OUTPUT«Failure.new(exception => X::Str::Numeric.new(source => "e", pos => 0, reason => "base-10 number must begin with valid digits or '.'"))␤»
GlitchMr rn: ('e'.Num).perl.say 12:56
p6eval rakudo c8b198: OUTPUT«Cannot convert string to number: base-10 number must begin with valid digits or '.' in '⏏e' (indicated by ⏏)␤ in method Num at src/gen/CORE.setting:9869␤ in method Num at src/gen/CORE.setting:3574␤ in block at /tmp/qMVNgm21x9:1␤␤»
..niecza v21-17-g83d732c: OUTPUT«Unhandled exception: Cannot parse number: e␤ at /home/p6eval/niecza/lib/CORE.setting line 1449 (die @ 5) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3539 (ANON @ 11) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3541 (NumSyntax.str2num @ 5) ␤ at /h…
12:59 wamba joined
dalek ecza: 27543db | (Solomon Foster)++ | lib/CORE.setting:
Add deterministic shortcut rules for numbers less than 341,550,071,728,321.
 13:03
masak 341,550,071,728,321 ought to be enough for anyone. 13:04
:P 13:05
spacebat_ outside the cantor cabal
13:05 leprevost joined, wamba left
masak most positive integers are huge, in fact. 13:06
spacebat_ like, at least 90%
moritz you can a rabin-miller test up to 341e12 with only 7 tests reliably?
*can do
masak rn: say (2 ..^ 10).pick(3) 13:07
p6eval niecza v21-19-g8e3a364: OUTPUT«9 7 8␤»
..rakudo c8b198: OUTPUT«2 9 3␤»
moritz time ./perl6 -e 'say +(^50_000).grep: *.is-prime' # takes 25s on my machine
just iterating the numbers takes about 3s
colomon moritz: really? 13:10
moritz: say +(1..100000).grep({ ($_ + 10000000).is-prime }); takes 20s on my machine under niecza, and that's with expmod implemented in pure p6 13:11
up to 50_000 takes 9.2s 13:12
13:12 wamba joined
TimToady note that libtommath is doing 256 trial divisions in addition to a bunch of RM 13:12
*MR
diakopter TimToady: ahoy 13:13
colomon TimToady!
TimToady: dividing by the first 256 primes? 13:14
diakopter TimToady: how many elements in the high-water mark of base 18
it's to 949341 after 15min
TimToady 1449405 13:15
colomon: yes
diakopter 15min of bases 3-18
TimToady not too shabby
dalek c: 586deab | (Felix Herrmann)++ | lib/Int.pod:
[Int] correct return behavior of Miller-Rabin
diakopter it will be interesting to see if it crashes halfway down 18 like on mono 13:16
this is .NET
masak imagines diakopter saying that and then pushing an emissary down an infinite well 13:18
colomon rn: say (10 ** 1000).rand 13:19
p6eval rakudo f39ec7, niecza v21-19-g8e3a364: OUTPUT«Inf␤»
colomon whoops
masak close enough. :P
moritz rn: say (10 ** 1000).Num
p6eval rakudo f39ec7, niecza v21-19-g8e3a364: OUTPUT«Inf␤»
moritz that's why
rn: say (10 ** 1000).pick 13:20
p6eval niecza v21-19-g8e3a364: OUTPUT«1000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000…
..rakudo f39ec7: OUTPUT«1000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000…
diakopter is proud of the bigint pick I wrote
moritz diakopter: where is it? can we steal it for rakudo? 13:21
colomon and niecza?
TimToady hafta go do stats for quizzing now
afk &
diakopter you couuld steal it for niecza, yes, since it uses BigInteger.new(byte[]) and System::Security::Cryptography::RNGCryptoServiceProvider
gist.github.com/275b3728b6fb750ee4ee 13:22
colomon woah
13:22 leont joined
diakopter oh wait, that's the broken verison; lemme update the gist 13:23
there, updated 13:25
13:27 cognominal joined 13:28 mucker joined
diakopter the pick is actually horribly inefficient 13:29
13:29 jaldhar left, _jaldhar joined 13:30 PacoAir joined
colomon diakopter: what is that written for/in? 13:30
diakopter perlesque
we used to have a p6eval target for it 13:31
afaik, the syntax is a mostly valid subset of P6
colomon with a BigInteger type 13:32
diakopter though more properly, it's a subset of C# with perl-esque syntax
colomon basically you're taking a set of random bytes and trying to make sure it's in the range you're looking for? 13:33
diakopter yeah
13:33 cart_ joined
diakopter to be far more efficient, it needs to clear all the bits that are clear in the topmost contiguous string of 0s in the limit 13:33
including the negative 13:34
colomon I'm sure RNGCryptoServiceProvider is better, but couldn't we just use 256.rand to do that portably in p6?
13:34 Pleiades` joined
spacebat_ I'd like to syntax highlight perl6 code, is this the best place to start github.com/perl6/std/blob/master/s..._highlight 13:34
diakopter yeah, but you'd still need the raw CLR byte[]
13:36 cart left
colomon hurm. isn't niecza and rakudo's current Int.rand going to have many, many numbers it can never possibly generate if the Int is, say 10 ** 100? 13:37
diakopter yeah; it really needs something native
diakopter goes to fix the pick so it's not braindead inefficient 13:39
colomon couldn't we do something like.... (10 ** 10).rand + (10 ** 10).rand * 10 ** 10 + ... ?
diakopter might as well generate it in base 256 13:40
or higher
13:40 Pleiades` left, cart_ left
diakopter (256**1).rand+(256**2).rand*256+(256**3).rand*256**2+(256**4).... 13:42
wait. 13:43
13:43 fgomez left
colomon no, 256.rand + 256.rand * 256 + 256.ran * 65536 ... 13:44
masak right.
diakopter yeah that
:D
moritz if you just add them, you create a normal distribution
GlitchMr $int.rand is ugly for me
13:45 Pleiades` joined
masak GlitchMr: don't do that, then. 13:45
GlitchMr Shouldn't it be (^256).pick or something
13:45 cognominal left
masak that works, too. 13:45
GlitchMr Oh, .rand returns floating point
13:45 cognominal joined
masak oh, right. 13:46
moritz hm 13:47
libtommath has an mp_rand function
which I can use to make arbitrary large random numbers
but only in digit-size resolution
and digit is 48 bit on 64bit systems
13:49 Pleiades` left
moritz now I have to come up with a clever algorithm to create a bigint <= a given one 13:52
13:54 Pleiades` joined
colomon knuth? 13:57
moritz worth checking I guess :-) 13:59
GlitchMr perl6: my $a = {}; $a<a> = $a; $a.perl.say; 14:00
p6eval rakudo f39ec7, niecza v21-19-g8e3a364: OUTPUT«(timeout)»
masak GlitchMr: I believe we have an old RT ticket about that one. 14:12
GlitchMr: what would you prefer it print, ooc? 14:13
moritz it could print do { my $a = { }; $<a> = $a } 14:14
masak aye; it could do that.
though the $a would more likely be $anon_ref_001 or something. 14:15
moritz $VAR1 14:19
GlitchMr If anything, it's $a<a>
Not $<a> which is $/<a>
moritz well yes
GlitchMr perl6: $_ = 1, 2, 3, 4; .[1].print 14:20
p6eval niecza v21-20-g27543db: OUTPUT«Use of uninitialized value in string context␤ at /home/p6eval/niecza/lib/CORE.setting line 1301 (warn @ 5) ␤ at /home/p6eval/niecza/lib/CORE.setting line 266 (Mu.Str @ 15) ␤ at <unknown> line 0 (ExitRunloop @ 0) ␤ at /home/p6eval/niecza/lib/CORE.settin… 14:21
..rakudo f39ec7: OUTPUT«Index out of range. Is: 1, should be in 0..0␤ in method Str at src/gen/CORE.setting:9871␤ in method print at src/gen/CORE.setting:7379␤ in sub print at src/gen/CORE.setting:7189␤ in method print at src/gen/CORE.setting:759␤ in block at /tmp/iq57PQ7M8s:1␤␤»…
GlitchMr It works with parens. Am I doing something incorrectly?
moritz only if you expect it to print 2
GlitchMr perl6: $_ = (1, 2, 3, 4); .[1].print 14:22
p6eval rakudo f39ec7, niecza v21-20-g27543db: OUTPUT«2»
moritz GlitchMr: doc.perl6.org/language/operators#It...Precedence
GlitchMr what the
= precedence varies?
what
moritz GlitchMr: I do recommend that you start reading the synopsis if you want to learn Perl 6 14:23
14:24 xinming_ left
GlitchMr So, I should do that? 14:26
duckduckgo.com/?q=!perl+S03
14:26 xinming joined
GlitchMr std: 2 but 3 but 4 14:27
p6eval std 77327a4: OUTPUT«===SORRY!===␤"but" and "but" are non-associative and require parens at /tmp/CdCvSN0fNg line 1:␤------> 2 but 3 but ⏏4␤Check failed␤FAILED 00:00 42m␤»
GlitchMr What if I want to make something like 3 but 'string' but False
Why 'but' is non-associative anyways? 14:29
colomon std: (2 but 3) but 4 14:30
p6eval std 77327a4: OUTPUT«ok 00:00 43m␤»
GlitchMr I know that I can use parens, but why? 14:31
3 but 'string' but False sort of makes sense
diakopter colomon: I updated the gist again with the vastly improved pick 14:34
spacebat_ hmm, rakudo star doesn't seem to have a STD module, and 2012.08 doesn't like the STD.pm from github
diakopter rakudo can't handle STD yet 14:35
spacebat_ is there a nice way to get rakudo to parse some perl6 source and give the parse tree 14:36
well, I want to use rakudo to support an emacs mode for perl6 14:38
diakopter afaik, your best bet is to use STD+viv
(not rakudo)
it can output a parse tree
in yaml 14:39
spacebat_ ah, sounds promising
not sure what STD+viv is though
diakopter TimToady has updated it recently
github.com/perl6/std
GlitchMr for own own of own own then own own
CoffeeScript :-)
diakopter spacebat_: I'll be glad to help you get viv running 14:40
spacebat_ I just cloned that, will look in to how to use it
diakopter read the whole README first to get the general idea
viv by default compiles to p5, but it can also output yaml 14:41
spacebat_ cool
diakopter spacebat_: how experienced are you at writing syntax highlighters for emacs? 14:42
14:42 not_gerd joined
diakopter (curious) 14:42
spacebat_ not really at all
I've fiddled with the emacs lisp regexes of a few modes
not_gerd hello, #perl6 14:43
spacebat_ tried the cperl6-mode branch but don't like it
I kindof know my way around emacs lisp though
diakopter I hope you have fun learning/trying stuff 14:44
spacebat_ figure that emacs lisp isn't what you want to be parsing perl with
better to parse perl5/6 with perl/PPI/STD
diakopter pugs and mp6 (kp6?) compiled to some lisp at some point(s) 14:45
spacebat_ yeah nice
diakopter masak: I can't remember - did you include v6/mp6/kp6 in your history of p6 summary blog post? 14:46
spacebat_ but I like the idea of having a perl subprocess that parses the file and sends back the structure needed for highlighting
diakopter it'll be huge 14:47
even for "small" programs
not_gerd perlito has a Lisp backend as well no idea how complete that is, though...
diakopter perlito grew out of mp6/kp6 iirc 14:48
spacebat_ I figure if padre can do it, emacs can use similar machinery in a subprocess 14:49
not_gerd needs help writing grammars 14:54
how do I factor tokens expression and squares of gist.github.com/3766337 so I won't get infinite recursion?
14:55 zby_home joined
diakopter I think you'll want to use a postcircumfix there 15:00
actually, 15:03
moritz doesn't see where the infinite recursion comes from 15:04
diakopter I'd put the '[' <list> ']' in its own rule as "square tail" then refer to it in expression
with a * on it
<square_tail>*
(and if you need to add more kinds of tails, put them all in one right_tail rule as alternations) 15:05
15:05 mucker left 15:06 mucker joined
moritz oh, it's left recursion 15:06
diakopter then expression will be more complex, but it has all it needs right there - you'll just need to be careful to construct the ast properly in expression's action rule
not_gerd: you see what I'm saying? remove <expression> from squares, rename squares to squares_tail, put the 3 match options in <expression> in an alternation, followed by <squares_tail>* 15:07
moritz token expression { [<atom> | <parens> ] <postfix>?`} 15:08
token postfix { '[' <list> ']' }
something like that
diakopter that's what I said.
(twice)
not_gerd diakopter, moritz: thanks, I got it 15:09
15:10 zby_home left, zby_home_ joined
diakopter moritz must have ignored me 15:10
moritz diakopter: no, I haven't 15:13
diakopter: I just put it into code
diakopter: you did too, but not as concise maybe
diakopter oh 15:14
15:15 sivoais left
dalek rlito: f9a20c8 | (Flavio S. Glock)++ | t5/01-perlito/17-hash-autovivify.t:
Perlito5 - add autovivification test
 15:16
15:19 flightrecorder left 15:21 FROGGS left, Gothmog_ left, Gothmog_ joined 15:27 sivoais joined 15:34 Celelibi joined 15:44 raiph joined 15:48 wamba left
raiph moritz: re *IN.get: irclog.perlgeek.de/perl6/2012-09-19#i_6002246 15:50
phenny raiph: 06:09Z <moritz> tell raiph thanks, fixed
moritz raiph: thanks 15:51
15:59 kaare_ left 16:00 kaare_ joined, wamba joined 16:06 yarp_ joined 16:09 yarp left, yarp_ is now known as yarp 16:11 mucker left
raiph moritz: more re *IN.get: concluding at irclog.perlgeek.de/perl6/2012-09-19#i_6002337 16:17
(er, well, properly golfed at rather than concluding at; discussion continues after that line) 16:19
16:20 Circlepuller_ joined
Ulti is there any libraries for IRC knocking around for Rakudo? 16:21
16:22 Circlepuller left, yarp left
timotimo GlitchMr: re your panda blogpost: in zsh, you'd put stuff like $PATH in .zshenv, no? 16:24
not_gerd Ulti: according to modules.perl6.org/ , there's IRC::Utils and Net::IRC::Bot, but they might have bitrotten 16:26
GlitchMr timotimo: hmmm... possibly 16:31
I haven't used zsh
I'm using my totally incompatible with anything shell
timotimo hm, not so sure about putting the panda and perl6 path in front of all others.
GlitchMr Why not? 16:32
This way you can avoid system's Perl 6
Let me guess, the only reason is performance?
Isn't this optimized anyways?
timotimo hm, i read that in some article some time 16:34
but it *may* have been just an argument against having . in the path
masak this thought seems to pop up everywhere nowadays: twitter.com/richhickey/status/2495...7718670336
GlitchMr Well, `.` in PATH couldn't be optimized
masak something like Make Everything Lexically Scoped.
GlitchMr But otherwise, it could store stuff in hash tables and use inotify to know when some PATH directory updates 16:35
timotimo the argument was more about security and unexpected things though, but it's been too long now.
GlitchMr I wouldn't ever put . to PATH
timotimo neither would I, nor the author of that article, if i remember that correctly
GlitchMr But, I would like to see actual benchmarks. Weren't hash tables O(log n)? 16:36
Oh, wait, they weren't 16:37
It's... average of O(1 + n / k)
timotimo supposed to be constant time
access, that is.
GlitchMr What was "k" anyways?
en.wikipedia.org/wiki/Hash_table
timotimo probably number of buckets
GlitchMr According to Wikipedia, it isn't
There is reason why Perl 5 randomizes hash order 16:38
dalek kudo/nom: e015e40 | moritz++ | / (5 files):
remove LAZY, which has long been deprecated
 16:39
timotimo languages should always select a random hash function from a global family (is that what they're called) anyway!
ah, universal hash families is what they're called 16:42
16:42 integral left
dalek p: 8ae6f20 | moritz++ | src/ (2 files):
implement bigint rand op
 16:42
GlitchMr Of course, it doesn't have to be actually hash table 16:43
It could be some alternative approach such as en.wikipedia.org/wiki/Judy_array 16:44
16:44 thou joined
GlitchMr I don't know what bash actually does, but caching would make sense 16:44
colomon rn: say (10 ** 1000).ln * 2 16:45
p6eval rakudo f39ec7: OUTPUT«No such method 'ln' for invocant of type 'Int'␤ in block at /tmp/B3r75KPSwO:1␤␤»
..niecza v21-20-g27543db: OUTPUT«Inf␤»
colomon rn: say (10 ** 1000).log * 2
p6eval rakudo f39ec7, niecza v21-20-g27543db: OUTPUT«Inf␤»
colomon not a glorious showing here!
moritz I know :(
GlitchMr Hmmm... I actually wonder what is actual performance of PATH in bash 16:46
16:46 whiteknight joined
GlitchMr perl -e'print join ":", ("/bin", "/usr/bin") x 10000' 16:47
Perhaps...
I meant
export PATH=$(perl -e'print join ":", ("/bin", "/usr/bin") x 10000')
16:47 tgt left
GlitchMr The answer being: 16:49
glitchmr@feather ~> perl
-bash: /usr/bin/perl: Argument list too long
wow - very long ENV could cause failures to run applications - makes sense
moritz garbage in, garbage out :-) 16:50
GlitchMr Anyways, with 2000 paths in PATH 16:51
glitchmr@feather ~> time q
-bash: q: command not found
real 0m0.012s
16:51 thou left
GlitchMr With 0 paths in PATH 16:51
glitchmr@feather ~> time q
-bash: q: command not found
real 0m0.012s
wait...
I CAN STILL RUN python WITH 0 PATHS? 16:52
What?
glitchmr@feather ~> time q
-bash: q: No such file or directory
real 0m0.004s
Ok, real stats
(before you will ask why time works - it's shell builtin) 16:53
16:54 felipe left
GlitchMr It's probably builtin so it could be used to time shell piping 16:54
moritz does 'time' includes the search in $PATH? 16:55
GlitchMr Well, it looks like it does
It was slower with 2000 paths
But I ran tests with CAKE variable instead of PATH and I don't have any slowdown 16:57
Ulti thanks not_gerd
16:57 birdwindupbird left
GlitchMr 0.01s is rather fast 16:57
But when running multiple programs or running loops... 16:58
Probably not really
moritz iirc $PATH lookups are cached anyway
16:59 integral joined, integral left, integral joined
GlitchMr I'm almost sure it's just hash performance 16:59
Hashes aren't O(1) in accessing performance
Ulti not_gerd: Net::IRC::Bot had an update two months ago so might be good
GlitchMr I could think of O(1) algorithms, except that algorithm is so wasteful that you would need Turing machine to run it or something 17:00
those*
are*
them*
I have found something - lwn.net/Articles/419826/ 17:01
17:01 thou joined
Ulti awesome panda install Net::IRC::Bot and my do nothing bot works :) 17:07
masak cool. 17:08
not_gerd bye, #perl6 17:09
17:09 not_gerd left
Ulti I'm writing a bot which tracks emotional content of what people are saying and keeps a score of how negative/positive they are and how negative and positive they are to other users 17:10
17:11 thou left
moritz sounds quite interesting 17:11
Ulti there is a nice blob of open data where someone hand annotated the emotional impact of a couple of thousand words from Twitter posts 17:14
arnsholt How are you resolving sentiment? 17:15
SVM classifier or something?
Ulti just looking up the word content in a dictionary and summing the scores from the database
heh no
nothing complicated
arnsholt Right. You'll probably want a negation resolver for that approach though
Ulti so a word like "bastard" has a -5 score and something like "amazing" would be +5
arnsholt Right, that one 17:16
A friend of mine used that to evaluate the negation resolution system he did for his master's thesis
moritz and what would "not bad" give?
Ulti at the moment the score of bad ;)
arnsholt That's the tricky part =)
Ulti but check for not prefixed isn't that hard 17:17
moritz
.oO( nothing wrong with that)
Ulti just invert the sign on the score if not is prefixed
arnsholt So what my friend did was this: sum scores of all words, negated words have changed sign
Ah, but negation usually has longer scope than a single token =)
Ulti moritz this is just a fun bot ;P
moritz Ulti: yes, I figured :-) 17:18
arnsholt His results showed that for positive sentiment, negation resolution didn't help the system a lot, but for negative sentiment it helped quite a bit
masak moritz: I read your "sounds quite interesting" above and immediately saw the autopun potential in that ;)
Ulti I already have a markov parody generator that will swap out the next word generated from a thesaurus based on altering emotional content, it works quite well
arnsholt interesting
17:19 leprevost_ joined
Ulti but yeah short of a perfect English grammar engine I'm not going to be able to do a great job 17:19
arnsholt In his data set it looked like positive sentiment was almost always expressed with positive words, but for negative it was quite common to see things like "this isn't a very good movie"
Ulti also noun use compared with adjective use makes a big difference
17:19 leprevost left
arnsholt Nah, don't need a perfect parser =) 17:20
He used a relatively straightforward approach
Ulti using negative words to describe something is more negative than just a noun which could be negative
masak so, "that amazing bastard" would have a score of exactly 0? :P 17:21
Ulti neuro.imm.dtu.dk/wiki/AFINN <-- the data if anyone wants some fun
masak: yeah something like that
clearly the scoring function is going to be key, might be better to average and just go with whatever has the majority 17:22
masak Ulti: that reminds me of someone who tried to find the extremes of hapiness scoring in (I think) The Simms.
Ulti so if amazing is only +1 or something but bastard is -5 its probably reasonably negative even if its said in a nice way
masak: lol 17:23
masak Ulti: they concluded that the highest happiness score was becoming a father, and that it was exactly canceled out by being on fire four times.
I wish I remembered the URL for that.
Ulti hahaha I bet a real world study would be similar 17:24
masak :P
moritz "highest happiness score was becoming a father" -- I can confirm that :-)
masak but not the 4x-on-fire thing, I hope.
moritz I've never tried that part, and don't intend to 17:25
Ulti mattoates.co.uk/work/parody.php <-- the thing I already have using some stolen data I could get into trouble for having
moritz: I bet being a father of a teen will adjust for the happiness without being on fire 17:26
or worse your teenager sets you on fire
mst I think I only set other teenagers on fire 17:29
Ulti I mostly blew things up with the minimum of fire
tadzik good evening 17:31
moritz \o tadzik 17:37
raiph o/ tadzik
dalek kudo/nom: 59dd40f | moritz++ | / (2 files):
make Int.rand return a random Int if the invocant is too big to become a Num
 17:38
17:44 dalek joined, ChanServ sets mode: +v dalek
colomon moritz: how does nqp::rand_I work? 17:49
afk # 4yo chasing....
moritz colomon: see github.com/perl6/nqp/commit/8ae6f20892 17:53
colomon ?errr.... if you use mod like that, won't you get an uneven distribution? Or am I missing something here 17:55
?
moritz colomon: it won't be perfectly uniform 17:56
colomon probably better than returning Inf... 17:57
moritz colomon: but note that the length of the random number is 1 digit longer than the modulus
colomon: and 1 digit is 48bit on 64bit systems
colomon oh
that does seem relatively harmless 17:58
moritz so the relative non-uniformity is limited to 2**-48
and maybe 2**-20 on 32bit platforms
I didn't bother coming up with a faster and/or more precise algorithm 18:00
18:03 leprevost_ left
moritz oh, it seems I've broken something with that patch 18:06
18:07 leont left
colomon is trying to figure out how libtommath does it, with no luck so far. 18:11
moritz how it does what? 18:12
dalek kudo/nom: 7d6aa02 | moritz++ | / (2 files):
Revert "make Int.rand return a random Int if the invocant is too big to become a Num"

This reverts commit 59dd40fa6b6eec1688e2b8a1f4a35cdad7d50e64. It broke stuff, even though I thought I had tested it. Back to the debugging board.		 18:13
18:15 tgt joined
masak the debugging board looks like a drawing-board, but it has more hexdigits on it :) 18:20
moritz correct
colomon moritz: how it does the random numbers for mp_prime_miller_rabin 18:21
moritz no idea on that one
sorear good * #perl6 18:25
colomon o/ 18:26
18:26 whiteknight left
sorear colomon: according to TimToady++, libtommath cheats and uses its internal list of the first 256 primes as "sufficiently random" numbers for miller-rabin 18:31
colomon oh
sorear moritz: are you still interested in cleverer approaches for uniform random bigints?
timotimo: Niecza uses universal hashing :D
colomon sorear: I certainly am interested in cleverer approaches.
moritz sorear: yesish :-) 18:32
timotimo as it should, i'd say. 18:33
colomon likes the idea of slipshod probabilistic algorithms even less than he likes the idea of probabilistic algorithms 18:34
18:39 fgomez joined
diakopter don't you hate it when you've been debugging code for an hour, and then you finally realize you're the one who wrote it? 18:41
dalek ast: b093af3 | (Solomon Foster)++ | S32-num/log.t:
Add test for the log of a very large Int.
colomon diakopter: indeed, that's a terrible, terrible feeling.
18:42 fgomez left
dalek ecza: 6287229 | (Solomon Foster)++ | lib/Builtins.cs:
Add BigInteger handling in ln.
 18:42
moritz doesn't really want to write a bigint log algorithm
colomon it's good for the soul! 18:43
colomon actually just used BigInteger.Log
sorear colomon: (Fat)Rat log ....
colomon sorear: yeah, that's clearly needed too. 18:44
moritz well, I guess I'll just implement log2
colomon but one baby step at a time.
masak .oO( log2 is all you need )
sorear depending on how many ulps you want bigint log is not so hard, just cut off all the low-order digits, turn the top 53 bits into a double, and pass it off to FLOG
moritz masak: I know 18:45
masak I suspected you knew ;)
moritz my idea is simply shift off digits until the result can be packed into a double
sorear FLOGP1 if you're feeling perverse
colomon has just been ordered to read Paddle-to-the-Sea
moritz just finished his reading duty and put the small one to bed 18:46
18:46 leprevost_ joined
sorear colomon: where did you get the magic numbers in github.com/sorear/niecza/commit/27...8dbf69adce ? 18:48
18:49 birdwindupbird joined
sorear moritz: actually what I was going to propose diakopter++ beat me to. 18:49
18:50 kaare_ left
sorear moritz: round up to the nearest power of 2, pick random numbers that are less than that power of 2 (easy to do with a random bit source), redo; if > $limit 18:50
expected iteration count will clearly not exceed 2
18:50 fgomez joined
sorear expected iteration count can be made arbitrarily close to 1 by changing the round-up rule 18:51
18:51 Su-Shee_ is now known as Su-Shee
moritz sorear: sounds sensible 18:51
18:52 felipe joined
moritz (though strictly speaking it's not an algorithm, becaues it's not guarantueed to terminate :-) 18:52
18:57 Chillance joined 19:02 Circlepuller joined 19:03 fgomez left, fgomez joined 19:04 benabik left, benabik joined, Circlepuller_ left
cognominal hum, forgot how to test a missing optional argument versus undefined 19:05
S06: You may check whether an optional parameter was bound to anything by calling C<VAR($param).defined>. 19:06
masak I don't think that ever worked in any implementation.
a perhaps more important question is: what makes you care? :)
MikeFair_ colomon: In what context? Sounds like it ought to get initialized by something if not provided
s/colomon/cognominal 19:07
:)
masak cognominal: you could set the default to a unique object that can never be passed in from the outside.
MikeFair_ cognominal: or perhaps it's your function/equation that needs to actually initialize it to a default if it is undefined (in which case the success of undefined is a missing optional parameter) 19:08
masak r: class A {}; constant DEFAULT = A.new; sub foo($param = DEFAULT) { say $param === DEFAULT ?? "wasn't passed" !! "was passed" }; foo(); foo(A.new)
p6eval rakudo 59dd40: OUTPUT«wasn't passed␤was passed␤»
masak cognominal: ^
cognominal it came in the context of an article I write. To explain the setter/getter generated for a public attribute. 19:09
missing optionnal argument means getter 19:10
interesting that the terms getter and setter don't appear in S12 19:11
but accessor is there
r: sub a($v?) { say VAR($v).defined } a 1 19:15
p6eval rakudo 7d6aa0: OUTPUT«===SORRY!===␤Confused␤at /tmp/otD8lMBymY:1␤»
cognominal r: sub a($v?) { say VAR($v).defined }; a 1
p6eval rakudo 7d6aa0: OUTPUT«===SORRY!===␤CHECK FAILED:␤Undefined routine '&VAR' called (line 1)␤»
cognominal r: sub a($v?) { say $v.VAR.defined }; a 1 19:16
p6eval rakudo 7d6aa0: OUTPUT«True␤»
colomon sorear: en.wikipedia.org/wiki/Miller%E2%80...ality_test
cognominal r: sub a($v?) { say $v.VAR.defined }; a
p6eval rakudo 7d6aa0: OUTPUT«False␤»
cognominal so it is there alright but as method macro, not as a function macro. 19:17
masak cognominal: the terms getter/setter aren't in S12, because that's not what they are, really. 19:20
cognominal: notice that actual setting of the value is done with '='
cognominal if I get things right, the setter/getter implicitely generated for a rw attribute C<$.attr> is equivalent to : method attr($v?) { $v.VAR.defined ?? $.attr = $v !! $.attr } 19:21
sorear colomon: what about it
colomon sorear: that's where the magic numbers came from.
sorear oh! 19:22
thanks.
colomon hopes wikipedia knows what it is talking about.
cognominal masak: the setting with = is another way or the only way?
diakopter colomon: I got my prime program to test primality at roughly 30,000 very large bigints/second
colomon diakopter: define very large? 19:23
diakopter 50 decimal digits. I guess not enormous.
cognominal S12: Public attributes have a secondary sigil of "dot", indicating the automatic generation of an accessor method of the same name"
masak cognominal: hm, you can always manually create setters that take an argument. but I haven't seen people do this, and I wouldn't consider it advisable or idiomatic.
diakopter afk a while & 19:24
cognominal I wonder where I have picked that...
masak r: class A { has $!x; multi method x() { $!x }; multi method x($new_val) { $!x = $new_val } }; given A.new { .x(42); say .x() } 19:25
p6eval rakudo 7d6aa0: OUTPUT«42␤»
masak do not do this :)
jnthn The accessor just returns the attribute. If you say "is rw", then it doesn't decontainerize it, which means that any scalar container remains and the assignment just works.
cognominal I was talking about generated accessors
masak cognominal: I know you were.
cognominal: but you seemed to be under the impression that they worked as above.
jnthn Right, the above is how geneated ones work :) 19:26
sorear diakopter: post-TWIRL, I'd say we want 200 or 300 digit primes for RSA
19:26 MikeFair_ left 19:31 fgomez left 19:32 leprevost_ left
diakopter sorear: I was wrong about the 50 digits 19:36
that's how long the winning prefix is
I'm instrumenting now to find out an average size per base 19:37
masak learns about TWIRL
I like hypothetical hardware devices :)
diakopter sorear: also, the point of the program is not to generate primes. it's to run TimToady's test 19:38
19:44 birdwindupbird left
diakopter hm; maybe 50 was too high. 19:44
well no
19:44 fgomez joined 19:46 birdwindupbird joined 19:50 FROGGS joined 19:51 flightrecorder joined 20:02 GlitchMr left
diakopter phenny: tell TimToady interestingly, stems of even and odd lengths alternate between 1x and 4x rate of primality testing 20:11
phenny diakopter: I'll pass that on when TimToady is around.
TimToady bases also tend to alternate, due to baked in primes 20:13
phenny TimToady: 20:11Z <diakopter> tell TimToady interestingly, stems of even and odd lengths alternate between 1x and 4x rate of primality testing
20:13 bruges_ left
TimToady figgered out how to tether to my phone... 20:13
*his
20:15 bruges joined
colomon TimToady: is there a reason you didn't use 2 * n.ln ** 2 as the upper bound for a in your Miller-Rabin? 20:18
sorear masak: yes, TWIRL is one of my all-time favorite papers :D 20:19
colomon: 2 * n.ln ** 2 is for the deterministic but GRH-conditional Miller test
colomon sorear: ah 20:21
20:38 MikeFair_ joined
masak sorear: only read the Wikipedia article so far. but then I'll read the paper, too. 20:52
21:05 zby_home_ left 21:08 raiph left
diakopter TimToady: my fast run is almost caught up to my less fast run at 1.5m on the base 18 downslope 21:19
4cores++ 21:20
my poor Thinkpad fan...
21:22 cognominal left 21:24 cognominal joined 21:27 cognominal_ joined 21:29 cognominal left 21:37 cognominal_ left 21:41 mtk joined 21:46 cognominal_ joined 21:52 birdwindupbird left 22:18 cognominal_ left 22:26 cognominal joined 22:27 MayDaniel left 22:33 tgt left
diakopter TimToady: how far has any of your runs gotten? finished base 18? 22:38
22:39 flightrecorder left, cognominal_ joined, flightrecorder joined 22:41 cognominal left 22:52 spider-mario left, cognominal_ left 22:57 leont joined 23:02 Circlepuller_ joined 23:04 Circlepuller left 23:10 MikeFair_ left 23:28 doy left, revdiablo joined 23:29 doy joined 23:38 Guest29702 joined, raiph joined
raiph phenny, tell pmichaud fyi: gist.github.com/3768134 23:38
phenny raiph: I'll pass that on when pmichaud is around.
23:41 PacoAir left 23:46 kid51 joined 23:47 onski joined
onski nr: 'a, b, c and d' ~~ /\w ** ', '/ and say ~$/ 23:48
p6eval niecza v21-21-g6287229: OUTPUT«Potential difficulties:␤ Unsupported use of atom ** ', ' as separator; nowadays please use atom+ % ', ' at /tmp/qnvZImduaw line 1:␤------> 'a, b, c and d' ~~ /\w ** ', '⏏/ and say ~$/␤␤a, b, c␤»
..rakudo 7d6aa0: OUTPUT«===SORRY!===␤Quantifier quantifies nothing at line 2, near " ', '/ and"␤»
onski nr: 'a, b, c and d' ~~ /\w % ', '/ and say ~$/ 23:49
p6eval niecza v21-21-g6287229: OUTPUT«===SORRY!===␤␤Unrecognized regex metacharacter % (must be quoted to match literally) at /tmp/lY2bedDSxL line 1:␤------> 'a, b, c and d' ~~ /\w %⏏ ', '/ and say ~$/␤␤Unable to parse regex; couldn't find final '/' at /tmp…
..rakudo 7d6aa0: OUTPUT«===SORRY!===␤Preceding context expects a term, but found infix / instead␤at /tmp/9AqigSvliB:1␤»
onski nr: 'a, b, c and d' ~~ /\w +% ', '/ and say ~$/ 23:50
p6eval rakudo 7d6aa0, niecza v21-21-g6287229: OUTPUT«a, b, c␤»
onski nr: 'a, b, c and d' ~~ /\w + % ', '/ and say ~$/ 23:51
p6eval rakudo 7d6aa0, niecza v21-21-g6287229: OUTPUT«a, b, c␤»
23:51 wamba left 23:52 flightrecorder left, FROGGS left
onski nr: 'a, b, c and d' ~~ /\w + % [,\s]/ and say ~$/ 23:54
p6eval rakudo 7d6aa0: OUTPUT«===SORRY!===␤Unrecognized regex metacharacter , (must be quoted to match literally) at line 2, near "\\s]/ and s"␤»
..niecza v21-21-g6287229: OUTPUT«===SORRY!===␤␤Unrecognized regex metacharacter , (must be quoted to match literally) at /tmp/FucmZNVxD5 line 1:␤------> 'a, b, c and d' ~~ /\w + % [,⏏\s]/ and say ~$/␤␤Unhandled exception: Unable to resolve method ast i…
onski nr: 'a, b, c and d' ~~ /\w +% [\,\s]/ and say ~$/ 23:55
p6eval rakudo 7d6aa0, niecza v21-21-g6287229: OUTPUT«a, b, c␤»
sorear welcome, onski 23:56
onski good morning! sorear.
nr: say 'nr++ !' 23:59
p6eval rakudo 7d6aa0, niecza v21-21-g6287229: OUTPUT«nr++ !␤»
23:59 onski left

Powered by Cro and the Raku® Programming Language.

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

Thank you!