»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'p6: say 3;' or rakudo:, or /msg camelia p6: ... | irclog: irc.perl6.org or colabti.org/irclogger/irclogger_logs/perl6 | UTF-8 is our friend!
Set by moritz on 22 December 2015.
xenu code examples at perl6.org have messed up formatting 00:26
i can fix it, i have a commit bit
but the question is 4 spaces or 2 spaces?
before commit that broke it, first example had two spaces and third one had four spaces 00:27
(second doesn't need any indenation)
indentation*
MasterDuke average it out then, 3
xenu nah, (4+2+0)/3 == 2 00:28
MasterDuke well, my non-joking vote is 4 00:29
dalek href="https://perl6.org:">perl6.org: 5617d84 | (Tomasz Konojacki)++ | source/index.html:
normalize code snippets indentation (4 spaces)
 00:56
xenu went with 4 spaces, i prefer it that way too 00:56
BenGoldberg m: gather { take 42 }; 43; 01:15
camelia rakudo-moar 8ede35: OUTPUT«WARNINGS for <tmp>:␤Useless use of constant integer 43 in sink context (line 1)␤»
BenGoldberg Why doesn't gather in sink context warn?
timotimo m: gather { say "hi"; take 1; say "boo"; take 2; }; 99; 01:31
camelia rakudo-moar 8ede35: OUTPUT«WARNINGS for <tmp>:␤Useless use of constant integer 99 in sink context (line 1)␤hi␤boo␤»
timotimo ^- because sinking a gather executes its contents
Hotkeys p6 doesn't have TCO does it? 03:20
or rather rakudo 03:21
nicq Is anyone on ATM? 03:57
Anyone know if there is an effort to add Rakudo with the moarVM to a package repo for Ubuntu or other distro? 03:59
tx0h hello 04:20
gfldex m: say "ohai tx0h!" 04:20
camelia rakudo-moar 8ede35: OUTPUT«ohai tx0h!␤»
tx0h is there a way to check if a dbiish connection is still alive? i like to keep my connection persistent open but it may happen that the database server needs to restart but the $dbh isn't connected after the restart (of cause) 04:22
konobi query with SELECT 1 every so often =0) 04:23
tx0h ok, but is this select cheaper than setting up a new connection? 04:24
konobi yup 04:27
tx0h oh, but a do('select 1') also throws an error, error handling on dbiish isn't documented. 04:33
geekosaur .u 61612 04:39
yoleaux No characters found
AlexDaniel please don't get angry about FAQ issues that I submit. I find it funny, hope you have a good laugh too. 04:49
tx0h ok, try..CATCH works well 04:54
i try to do my first perl6 web project :-) 04:56
stmuk AlexDaniel: your comments are useful but I'm not keen on things being described as "weird". 05:14
AlexDaniel stmuk: Couldn't find any better word. Didn't try harder because I thought that the word choice there is not important 05:20
stmuk github seems totally broken anyway :/
AlexDaniel yeah, down
Roamer` hmm, is there any appreciable difference between for %hash{*}:kv and for %hash.kv? 05:42
TimToady the latter is likely to be a bit faster 05:43
(I could be wrong about that, of course.) 05:45
psch m: my %h = :1a, :2b, :3c; say %h{*}.kv eqv %h.kv 05:46
camelia rakudo-moar 8ede35: OUTPUT«False␤»
psch m: my %h = :1a, :2b, :3c; say %h{*}.kv.perl; say %h.kv.perl
camelia rakudo-moar 8ede35: OUTPUT«(0, 1, 1, 3, 2, 2).Seq␤("a", 1, "c", 3, "b", 2).Seq␤»
Roamer` erm, psch, {*}:kv, not {*}.kv
psch oh
my bad
TimToady oh, {*} is a list
psch m: my %h = :1a, :2b, :3c; say (%h{*}:kv).perl; say %h.kv.perl 05:47
camelia rakudo-moar 8ede35: OUTPUT«("a", 1, "c", 3, "b", 2)␤("a", 1, "c", 3, "b", 2).Seq␤»
Roamer` TimToady, would that be because the {*} has to build a list of the pairs first and then iterate over them, and the .kv just starts iterating?
yeah, looks that way, thanks to you both
TimToady :kv may have to check each key to see that it exists, unless it's smart about * producing all existing keys 05:48
TimToady it it's smart enough, it could just translate {*}:kv to .kv though 05:49
ufobat good morning p6 :-) 05:51
TimToady o/ 05:52
tx0h what i really don't like are the nuance differences between p5 and p6. the space between if and ( for instance, or the asterisk in %*ENV... 06:03
TimToady it's all in the interests of namespace management for extensibility 06:04
keywords are not functions, so you may not call a keyword as if it were one
TimToady that let's us add keywords or functions independently 06:05
*lets
TimToady likewise you don't have to know that ENV is special, the * tells you that it's a dynamic/global variable 06:05
in p5 you have to memorize the list of special symbols
tx0h but you have to agree that these differences are annoying. especially if you you program in both languages parallel 06:07
TimToady everything is annoying, if you let it :)
tx0h i hope for a epic6 or such alike module for eclipse, that would be great. 06:10
stmuk I always found if( annoying in perl 5 06:33
tx0h stmuk: you didn't got forced to don't use a whitespace 06:36
tx0h use or not use whitespaces is part of freedom of expression. thats why i don't like the python block concept, where whitespaces are part of the syntax. now p6 follows this idiocracy 06:38
stmuk github.com/FROGGS/p6-Slang-Tuxic may be of interest 06:44
tx0h yepp, there are some approaches 06:50
ty
this already shows that there is a problem (even if it is a psychological) 06:51
TimToady yawns
tx0h sry 06:56
Roamer` I guess there wouldn't really be a way to uninstall a single installed module (or even a module and everything that depends on it) from a repository 07:51
is there then a way to blow away everything that has been installed by panda?
of course I can rakudobrew nuke and then build again, but is there a faster way?
DrForr That's more of a package-manager thing. 07:52
moritz Roamer`: I think zef supports uninstallation, or something 07:56
Roamer` moritz, ahh, I guess it's time I actually tried a different package manager. Okay, thanks 07:59
lizmat TimToady: so $supply.emit --> Nil, what about sub emit ? 08:04
Making that also --> Nil breaks some expectations, I'm afraid
however, I would be all for it, as it would allow for easier / better optimisations 08:05
stmuk Roamer`: zef has module uninstall 08:21
as moritz says
stmuk me-- # not fully reading scrollback 08:22
ismailarilik Hi! 10:12
I has seen this error message:
===SORRY!=== Unable to write bytecode to 'C:\Users\ismail\Öğretmen\lib\.precomp\03419F43BFFCE0AEDE2CD622AE91E8513382EF49.1468939074.63395\A3\A3416250288174506BE179C4BE847ECCB7196C9A.bc'
ismailarilik I have two related files: lib\model.pm6 and bin\komut-arayüzü.p6 10:14
ismailarilik The first line of lib\model.pm6: unit module Model; 10:14
The first line of bin\komut-arayüzü.p6: use Model; 10:15
ismailarilik I run bin\komut-arayüzü.p6 file on the project directory with this command: perl6 -I "lib" bin\komut-arayüzü.p6 10:16
Where am I wrong?
psch ismailarilik: the problem is that the perl6 executable can't open the file for writing 10:19
psch ismailarilik: that's more likely to be a permission problem in your operating system than a bug in your Perl 6 code 10:19
ismailarilik The project directory is in my home directory
C:\Users\ismail\Öğretmen>perl6 -I "lib" bin\komut-arayüzü.p6
Is it related with a encoding problem? 10:20
psch possibly
i don't know how windows/ntfs encodes paths 10:21
ismailarilik Okay, I will use ASCII for files and directories and try again
psch in any case, the error you're getting is from the VM backend and happens if the call to fopen fails
ismailarilik Command has worked on this directory: C:\Users\ismail\Ogretmen 10:23
So it is related with an encoding problem.
Thanks a lot. =)
psch ismailarilik: what's your filesystem? 10:27
ismailarilik I guess it is NTFS 10:28
How I can learn it precisely?
psch right click on the partition and click "Properties" 10:29
your system encoding probably also plays a role
ntfs internally encodes as utf16, and we encode the filename as utf8, so that might be it already 10:30
we do have a dev on windows, but i assume they're pretty anglocentric in path and filenames :)
ismailarilik I have looked up 10:36
It is NTFS
I created all these files, not fetched from a VCS host, for ex. GitHub. 10:37
mniip $ perl6 test.pl6 10:58
Segmentation fault (core dumped)
I suppose that's not something that's supposed to happen
gfldex ismailarilik: git and shares don't mix well. Make sure you don't got any anything, including the temp dir on a share, before you write a bug report. 11:00
masak mniip: you suppose correctly
gfldex mniip: please rakudobug
jnthn Well, unless you did something silly with nativecall :) 11:01
mniip nah I was doing a some pretty primitive messing around 11:02
and I think I accidentally multi'ed a builtin
jnthn hm, most of 'em are multis. But yeah, file it :)
mniip bpaste.net/show/25676c934f42 11:06
gfldex gist.github.com/gfldex/491d637082d...7adcd2746c 11:08
m: gist.github.com/gfldex/491d637082d...7adcd2746c
camelia rakudo-moar 7f9623: OUTPUT«(signal SEGV)»
gfldex m: multi cross(&f, @ ($x, *@xs), @ys) { flat map({&f($x, $_)}, @ys), cross(&f, @xs, @ys); }; multi cross(&f, @ (), @ys) { (); } 11:09
camelia rakudo-moar 7f9623: OUTPUT«(signal SEGV)»
gfldex m: multi cross(&f, @ ($x, *@xs), @ys) { flat map({&f($x, $_)}, @ys), cross(&f, @xs, @ys); }; multi cross(&f, @, @ys) { (); }
camelia rakudo-moar 7f9623: OUTPUT«(signal SEGV)»
gfldex m: multi cross(&f, @ ($x, *@xs), @ys) { flat map({&f($x, $_)}, @ys), cross(&f, @xs, @ys); }; multi cross(&f, @, @ys) {}
camelia rakudo-moar 7f9623: OUTPUT«(signal SEGV)»
gfldex m: multi cross(&f, @ ($x, *@xs), @ys) {}; multi cross(&f, @, @ys) {} 11:10
camelia rakudo-moar 7f9623: OUTPUT«(signal SEGV)»
gfldex m: multi cross() {}; multi cross() {}
camelia rakudo-moar 7f9623: OUTPUT«(signal SEGV)»
gfldex m: multi cross() {}; 11:11
camelia rakudo-moar e87752: OUTPUT«(signal SEGV)»
mniip haha
gfldex that may be the shortest segfault we ever had :)
masak no, I think we've had shorter ones 11:16
diakopter++ holds some hard-to-beat records in that department 11:17
gfldex mniip: please rakudobug the golfed version
mniip I guess I can golf it a but further 11:18
m: multi cross{} 11:19
camelia rakudo-moar e87752: OUTPUT«(signal SEGV)»
masak src/core/List.pm:my &cross = &infix:<X>; 11:20
that's the declaration
hm...
m: my &foo = &infix:<X>; multi foo{}
camelia rakudo-moar e87752: OUTPUT«5===SORRY!5=== Error while compiling <tmp>␤Redeclaration of routine foo␤at <tmp>:1␤------> 3my &foo = &infix:<X>; multi foo{}7⏏5<EOL>␤»
masak m: my &foo = &infix:<X>; { multi foo{} }
camelia ( no output )
masak ok, hypothesis disproved 11:21
mniip um
crashes over here
masak not here
This is Rakudo version 2016.07.1-57-g0419194 built on MoarVM version 2016.07-3-gc01472d 11:22
mniip This is Rakudo version 2016.07.1 built on MoarVM version 2016.07
masak bisectable: my &foo = &infix:<X>; { multi foo{} }
bisectable masak: On both starting points the exit code is 0 and the output is identical as well
masak: Output on both points:
masak bisectable says it works always :)
bisectable masak: On both starting points the exit code is 1 and the output is identical as well
masak: Output on both points: ===SORRY!=== Error while compiling /tmp/NsiSKwsZ9p␤Preceding context expects a term, but found infix : instead␤at /tmp/NsiSKwsZ9p:1␤------> says it works always :<HERE>)
masak oh come on, bisectable 11:23
bisectable: help
bisectable masak: Like this: bisectable: good=2015.12 bad=HEAD exit 1 if (^∞).grep({ last })[5] // 0 == 4 # RT128181
masak bisectable: url?
bisectable masak: On both starting points the exit code is 1 and the output is identical as well
masak: gist.github.com/1417c4b529e8c813d6...e45f8e4df6
masak finds github.com/perl6/bisectbot
mniip ! 11:24
it crashes if you enter it as two lines into the repl
lizmat fwiw, I'm testing a fix on the segv 11:26
masak aha -- metacpan.org/source/HINRIK/Bot-Bas...ot.pm#L669 means that the ":" is optional after "bisectbot" 11:29
I think I have what I need to put together a patch
lizmat 998a1ef has a fix for the "multi cross" segv 11:32
mniip gfldex, perhaps I'm eligible for "least time spent with perl to make it crash" because I literally only started figuring out the language yesterday 11:44
El_Che mniip: you figured out perl? :) 11:46
masak yes, do tell :P 11:47
lizmat mniip++ # but now it doesn't crash anymore :-)
mniip El_Che, I can't say I did that yet 11:48
so far I'm not seeing the large scale idea
gfldex mniip: quite a few things are not implemented yet. You can't have autothreading if your threading stuff produces odd behaviour including segfaults. 12:00
jkramer Ahoy! 12:03
jkramer How can I re-export used modules from a module? Or just some stuff from a module? Is there something like 'use SomeModule is export;'? 12:04
gfldex jkramer: see: docs.perl6.org/language/modules#EXPORT 12:05
dalek c: bd72f04 | (Wenzel P. P. Peppmeyer)++ | doc/Language/modules.pod6:
add sub EXPORT to index
 12:08
mniip lizmat, hmm, are those builtin modules somehow special? 12:14
is there something preventing us from doing a similar setup with something different from cross
ah I see you created an issue on moarvm 12:16
[Coke] table of contents hide/show is broken for me on chrome, os x, docs.perl6.org/type/AST 12:18
Ulti if I wanted a sliding window over a list is there a nice builtin for it? so the list (1,2,3,4) I would want returned for a window of 2 ((1,2),(2,3),(3,4)) 12:21
rotor but on a position rather than value is sort of that thing
[Coke] ok, reload makes it work, but now the hide is silly - it just blanks it out, doesn't free up any space. 12:22
[Coke] finds/opens a ticket. :|
[Coke] bumps github.com/perl6/doc/pull/767 - my recommendation is to revert for now. 12:24
jnthn m: say (1,2,3,4).rotor(2, -1)
camelia rakudo-moar aa5e49: OUTPUT«===SORRY!===␤Cannot have elems < 1, did you mean to specify a Pair with => -1?␤»
jnthn m: say (1,2,3,4).rotor(2 => -1)
camelia rakudo-moar aa5e49: OUTPUT«((1 2) (2 3) (3 4))␤»
jnthn Ulti: ^^ 12:25
Ulti thats neat 12:27
I just came up with one that invokes a scary error! :D
m: say (2,3,4).combinations.grep: 1 < *.elems < 3
camelia rakudo-moar aa5e49: OUTPUT«===SORRY!===␤QAST::Block with cuid 1 has not appeared␤»
Ulti wat
m: say (2,3,4).combinations.grep: 1 < *.elems
camelia rakudo-moar aa5e49: OUTPUT«((2 3) (2 4) (3 4) (2 3 4))␤»
Ulti not what I wanted 12:28
but its an interesting error
thanks jnthn
jnthn wat
Yeah, indeed it is
Ulti I'll take your solution and give you a bug ;)
feels fair
jnthn Dammit :P
otoh, at least it's not another concurrency bug :)
Ulti yeah and it came ready golfed 12:29
jnthn :)
m: 1 < *.elems < 3
camelia rakudo-moar aa5e49: OUTPUT«===SORRY!===␤QAST::Block with cuid 1 has not appeared␤»
jnthn Golfeder
Ulti golfest
shillo Hey, is this a known problem? Pushing into large arrays (>4M elemenets) seems to have square complexity
For instance, small arrays: 12:30
perl6 -e 'my @a; @a.push($_) for ^1000000' 0.99s user 0.06s system 99% cpu 1.052 total
perl6 -e 'my @a; @a.push($_) for ^2000000' 2.08s user 0.08s system 99% cpu 2.163 total
perl6 -e 'my @a; @a.push($_) for ^4000000' 4.72s user 0.12s system 99% cpu 4.845 total
.. and then larger ones:
perl6 -e 'my @a; @a.push($_) for ^8000000' 12.25s user 0.22s system 99% cpu 12.482 total
jnthn shillo: Yes, every time it resizes it has to copy. It doens't know how many elements you're going to have up front.
shillo perl6 -e 'my @a; @a.push($_) for ^16000000' 36.75s user 0.53s system 99% cpu 37.346 total 12:31
shillo perl6 -e 'my @a; @a.push($_) for ^32000000' 124.53s user 1.17s system 99% cpu 2:05.91 total 12:31
yeah, but if you allocate double size from the original, that gives you linear amortized
shillo ... and it'd happen for small arrays, too, unless you start resizing into smaller sizes than double 12:31
jnthn I don't think it doubles every time. 12:33
I think it does that up to a point
And then goes in smaller increments
jnthn checks
jnthn github.com/MoarVM/MoarVM/blob/mast...ray.c#L318 12:34
We *could* change it to just double every time, though that is of course a trade-off. 12:35
shillo yeah, I think it's the right choice because if you've gone up to 10M, what's the chance you'll stop anytime soon? But YMMv. 12:36
shillo ... OTOH, one way to optimize would be to know what's the chunk size that malloc can accomodate without its own overhead 12:36
... Moar uses system malloc, though? Or its own?
jnthn System malloc once you're at that size. 12:37
shillo ok :)
mniip at that size it's time to call mmap though 12:38
shillo ... by choosing I mean something that's double, then round to closest that fits malloc prefered chunk size (which tends to be mmap page size - malloc headers size)
... the mallocs I looked at do just that, mniip. They just substract a few words for headers. 12:39
mniip I wrote an allocator that did nothing but mmap once 12:40
shillo a number of mallocs will mmap when they run out, and will return the completely empty mmaped pages to the OS
mniip it was fun to see lua consume hundreds of megabytes for simple tasks 12:41
shillo Oh, just pure mmap? Fun :) 12:42
mniip yeah, whole page for even 1 byte allocations
shillo jnthn: about doubling always, statistically you'll only be wasting 25% memory on average. Because the first half arrived from copy, and will be full, then (if the array sizes are random) the second half will end at a random spot 12:44
jnthn I'm not sure what the full history on that code is. I know that I took it from the QRPA implementation that pmichaud did on Parrot, but I don't know what that was inspired by. I've no idea what Perl 5 does, though somebody here probably does. :)
shillo jnthn: So the fill will be 50%-100%, with not-that-much waste.
jnthn I'd be fine with a patch to change it, if somebody wants to try tweaking it and do measurements to see if it's an improvement. 12:45
timotimo the problem about "always double" is when you reach half the size of your user's RAM
you'll try to double, but hit the ceiling, and stay at only half 12:46
jnthn tbh, if you're dealing with that much data you maybe want to think your problem through a tad more than using a dynamic array. ;)
timotimo right; like for the code above you'd probably want a native array to conserve some more memory
jnthn Yeah, the heap analyzer uses native arrays to good effect :) It's good if your data is integers. 12:47
Or floats
timotimo also, pushing a bunch of boxed values into the same array is pessimal for our GC, as the amount of objects that go on to the gen2 is far too large 12:47
shillo .... just looking at benchmark, it's not quadratic, in any case.
timotimo: Well, yes, but with quadratic complexity, you'll run out of time long before you run out of RAM
jnthn timotimo: Yeah, there is that, though GC only runs so often :) 12:48
timotimo does realloc allow you to recover when it couldn't malloc enough space?
jnthn But yeah, pre-tenuring is an interesting thing for the future.
timotimo yup
using @a.push($_) for ^1_000_000 gives me 46 GC runs, 3 of which are full 12:49
shillo jnthn: I think I remember you made GC always double the last generation size, so GC is not the problem
timotimo 14.62% spent in GC :)
shillo (... and I read the GC source, so...)
jnthn shillo: Yes, but the first generation is fixed in size and because everything gets a second stay in GC before promotion, then you'll end up with a collection that frees little, then one that promotes a lot, and vice versa 12:50
um, vice versa wasn't what I wanted
shillo yeah, I know
jnthn But anyway, that pattern.
shillo ... but that's normal for generational GCs when you allocate a lot
jnthn Yup
shillo ... I don't see that as a problem
timotimo imgur.com/TBSr6fQ 12:51
gregf_ where does the sub/method 'cross' get picked from?
timotimo "picked"?
shillo jnthn: Mind you, I don't see a series of pushes to be the best pattern for most applications, but it will be idiomatic for people coming from either perl5 or C++/STL 12:52
gregf_ i thought that the same sefault should get thrown for evey global function thats overridden :|
er, picked == "called" # :/ sorry, my bad
s/evey/every/
s/sefault/segfault/ 12:53
jnthn shillo: Yeah, the interesting work in this area is dynamic profiling of allocation sites, to see which things tend to end up getting promoted reliably and then compiling their allocations to always be directly in the old generation.
timotimo gregf_: you should be able to figure it out by looking at the commit that fixed the problem
shillo jnthn: Interesting. :)
timotimo github.com/rakudo/rakudo/commit/998a1ef168
gregf_ timotimo: cool, ta 12:54
shillo timotimo: I'm still not sure it's good to slow down very badly for arrays of ~10M elements in order to not run out of memory with arrays around 200M-1000M elements. 12:55
jnthn We could just pick a much larger threshold
8K elements isn't really very many given the memory size of modern computers :) 12:56
shillo timotimo: I mean, if you look at this as 'How much data I can handle before I need to change my program?', well. 12:56
timotimo If realloc() fails, the original block is
left untouched; it is not freed or moved.
^- we should be able to compare the previous pointer value with the returned value and try smaller increases before giving up and throwing our hands in the air
timotimo if that's something you consider worth trying, anyway 12:57
shillo What exactly is ssize in that code? Because it doesn't seem to be a number of elements 13:00
... Had to go into milions before it started to slow down
jnthn storage size
The number of elements we can hold
jnthn When we unshift, we ++ start, and -- elems 13:01
So unshift doesn't involve a memcpy
And can be cheap
perlpilot sounds just like Perl 5
shillo Uh, so why doesn't it slow down much sooner?
jnthn Cache boundaries maybe? 13:01
jkramer What's the equivalent of $some_object->$method_name_in_variable in P6? 13:02
timotimo $some_object."$method_name_in_var"()
jkramer Thanks :)
timotimo YW
jnthn Once you get into the millions, then the memory associated with the array won't fit in any of the CPU caches any more
shillo True 13:03
jnthn And trips to main memory cost up to 100 times more than cache hits
timotimo and when you're copying stuff over, you're probably holding two arrays in cache
jnthn (For L1...for L3 "only" 5-10 times I guess...)
timotimo unless memcpy is clever enough somehow to skip writing to the cache and somehow writes directly to main memory?
or maybe it causes some DMA coprocessor to run?
i really have no idea how modern machines do such things 13:04
shillo Wait, no, I mean, it doesn't slow down at all. I just tried replacing 8192 with 81920000. The smallish case doesn't speed up even slightly.
jnthn How are you measruing?
*measuring
shillo Or wait, I'm blind. It does. :p 13:05
timotimo measruing wrnog, clearly
jnthn ah :)
shillo perl -e 'my @a; push @a, $_ for 0..20000000' 1.38s user 0.18s system 99% cpu 1.563 total
timotimo clraelry?
shillo No, that's perl5
Duh, I pulled a wrong line from history
timotimo m: $_ = "measuring wrong, clearly"; s[ <:Letter>+ ] = $/.comb.pick(*).join; .say 13:07
camelia rakudo-moar aa5e49: OUTPUT«geusrnami wrong, clearly␤»
timotimo m: $_ = "measuring wrong, clearly"; s:g[ <:Letter>+ ] = $/.comb.pick(*).join; .say
camelia rakudo-moar aa5e49: OUTPUT«agsmirnue ngwor, elyracl␤»
timotimo that's .. a bit radical :)
shillo jnthn: I think that realloc was the wrong place to look
timotimo m: $_ = "measuring wrong, clearly"; s:g[ <:Letter> ** 1..3 ] = $/.comb.pick(*).join; .say
camelia rakudo-moar aa5e49: OUTPUT«maeusrgin worgn, leclray␤»
shillo ... no speedup in any cases
perl6 -e 'my @a; @a.push($_) for ^4000000' 4.76s user 0.12s system 99% cpu 4.883 total 13:08
perl6 -e 'my @a; @a.push($_) for ^8000000' 12.29s user 0.24s system 99% cpu 12.536 total
jnthn No shenanigans like "forgot to make install"?
shillo (that's with if (ssize < 819200000) {)
No, but I only did make install on Moar 13:09
timotimo you know when i said "14.62% spent in GC"?
shillo ... I assume I don't need to rebuild nqp/rakudo?
jnthn That should be enough
timotimo i 10x'd the amount of pushes and gc went up to 35.02%
timotimo The average nursery collection time was 23.4ms. The average full collection time was 145.93ms. 13:09
the last full collection took 371 ms
that wouldn't cause the time to go quadratic, but it's kinda bad :) 13:11
unmatched} m: sub foo { say <foo bar ber>.map: * ~ "_" x $++ }; foo; foo; foo 13:12
camelia rakudo-moar aa5e49: OUTPUT«(foo bar_ ber__)␤(foo___ bar____ ber_____)␤(foo______ bar_______ ber________)␤»
unmatched} m: sub foo { say <foo bar ber>.map: { $_ ~ "_" x $++ } }; foo; foo; foo
camelia rakudo-moar aa5e49: OUTPUT«(foo bar_ ber__)␤(foo bar_ ber__)␤(foo bar_ ber__)␤»
perlpilot timotimo: What's it look like for unshift rather than push?
perlpilot guesses it's worse 13:12
timotimo for GC? should be the same
shillo Oh, *facepalm*
unmatched} So the WhateverCode doesn't get treated as its own block for the anon vars?
jnthn Doubt it
timotimo unmatched}: there's no curlies, so there's no lexical scope, nor a block
moritz correct
shillo ... this is fast because realloc reserves a *lot* more than you asked
unmatched} Thanks 13:13
shillo ... and in most cases doesn't trigger a copy
timotimo it's valuable to keep that as a simple rule
shillo I put a printf("%ld %ld %ld\n", ssize, slots, slots_old); after realloc (and saved slots into slots_old) and got that it copies... rarely. Basically when you double the size, until it hits a limit size 13:14
I mean, when you hit double the size of the last copy, until a limit set in realloc implementation. 13:15
timotimo cool
literal is there a way to force Panda to use https as the git protocol when building Rakudo Star? 13:16
timotimo i think we have GIT_PROTOCOL support in some places
however, rakudo star is supposed to install stuff from the files we ship along with it 13:17
literal both NQP and Rakudo default to https, but Panda doesn't for some modules
timotimo without hitting the 'net
literal my build is dying because it's trying to clone a git:// repository from github (moritz/svg.git)
shillo timotimo: Not really cool, because this relies on a particular malloc implementation to work in linear time for even smallish arrays. 13:18
timotimo oh
well, fair enough
timotimo installed 8 gigs of swap space so the next 10x can hopefully finish 13:19
gfldex literal: you can force git to do that, ask google how 13:20
perlpilot pretty sure panda groks GIT_PROTOCOL 13:21
timotimo here comes the swap 13:23
my music is suffering from it already
tadzik pretty sure panda coined it :) 13:24
dalek c/revert-767-animated_css: feb4798 | Altai-man++ | / (2 files):
Revert "Css-based TOC animation"
 13:25
[Coke] perlpilot: yes it does, I use it all the time 13:26
dalek c: feb4798 | Altai-man++ | / (2 files):
Revert "Css-based TOC animation"
 13:27
c: 5f09019 | Altai-man++ | / (2 files):
Merge pull request #774 from perl6/revert-767-animated_css

Revert "Css-based TOC animation"
dalek c: f151451 | Altai-man++ | html/css/style.css:
Delete remainder of github.com/perl6/doc/pull/774
 13:28
timotimo wow, the 100_000_000 items array push run finished 13:29
[Coke] wonders how he can get back to a working doc build on os x. :|
ismailarilik gfldex: I don't think I understand you...
timotimo 80% time spent in GC :)
the last full run it did took a measly 13.8 seconds 13:30
probably due to swapping?
so i suppose i can throw that test result into the bin
gfldex ismailarilik: just make sure you don't got anything on a share because rakudo can't do anything about problems that arise from shares 13:31
timotimo OK, with 40_000_000 items GC is down to 64%, which is fine ... 13:35
shillo timotimo: ... and I just remembered the real tradeoff with doubling-before-calling-realloc: if you reserve your own space by doubling, and realloc reserves its own space by, basically, ignoring your request, 13:36
timotimo yup
shillo You end up with two independent chunks of wasted space, only one of which is usable by your application for array growth
... which is why I think that the correct thing to do here is to double always (until you run out) but have a known malloc implementation so you know which numbers to double to 13:37
... so that realloc doesn't waste extra space for you 13:38
... OR
shillo ... double manually but call malloc / memcpy on your own 13:38
... and hope malloc won't reserve space like that.
shillo (most mallocs are very tight and actually won't) 13:38
timotimo btw, for 40_000_000 elements but using a native array, you get 3% GC time and only 19 seconds (instead of 3:58 minutes) on my machine 13:39
perlpilot Perl 5 grows its arrays by doubling last time I checked ... I wonder how it does that.
shillo perl5, 160,000,000, 11 sec 13:40
... normal array
perl -e 'my @a; push @a, $_ for 0..160000000' 10.26s user 1.06s system 99% cpu 11.331 total
timotimo well, perl6 is known to be a bit slower than perl5 for many things :) 13:41
oh, that was also with --profile, which adds a bit of extra overhead to every call and allocation
i get that in 49 seconds. 5x slower than perl5 isn't that bad, really. 13:42
FWIW, when i use "my int @a = ^160_000_000", it's done in only 2.5 seconds 13:43
shillo pypy -c 'a=[] for i in xrange(160000000): a.append(i)' 1.92s user 1.04s system 99% cpu 2.973 total
timotimo oh come on, now. that's just unfair! :) 13:44
pypy is 100% space technology from space
shillo Yeah. pypy is evil :)
It also hates threading, otherwise it'd be a go-to solution for implementing.. anything. 13:45
(meaning, any language)
timotimo we lose about 30% time through invocation overhead for the push + for thing
shillo ... oh, today I benchmarked push specifically to see if it gained anything from declared return Nil
timotimo i count frame_invoke, fixed_size_alloc, try_return, allocate_frame, alloc_zeroed, fixed_size_free, frame_destroy, args_proc_init, args_set_result_obj, and frame_find_invokee_multi_ok as those things 13:46
travis-ci Doc build passed. Altai-man 'Revert "Css-based TOC animation"' 13:47
travis-ci.org/perl6/doc/builds/148033822 github.com/perl6/doc/commit/feb479880cd5
shillo ... and also, apology to everyone for bringing up a microbenchmark - I realise this is not the kind of thing that should be the primary opt target :) 13:47
timotimo nah, it's fine
it's a bit surprising to me that spesh doesn't inline the push call there 13:49
shillo Like I said the other day, the only more practical code I wrote is slowed down by IO.basename, and that one waits on regexp search/replace. I suspect that p6 web frameworks that use grammars for routing will probably wait on the parsers to finish
timotimo yeah, probably 13:50
invocation of regexes isn't optimized enough
timotimo huh, it's not even turning the invoke into a fastinvoke 13:53
[Coke] so, I can't get a clean install of Pod::To::BigPage due to various dependency failures on sergot's and cosimo's modules. 13:55
gfldex [Coke]: disable the test that is getting IO::Socket::SSL involved and see if that fixes your problem. If it does, send a PR. 13:57
shillo timotimo: how do you look into spesh output? By turning on spesh log or? 14:02
timotimo yes, MVM_SPESH_LOG=/tmp/foo.txt
and then just start reading from the bottom :D
unmatched} So I got this line of code: dd [$_, $s.channels] for |$s.channels; And it outputs [("#zofbot", "#perl6"), Array[<anon>].new($("#zofbot", "#perl6"))] How come it doesn't flatten the @.channels attribute's value? I want to iterate over each item 14:15
timotimo if it's an array, you probably want to .List (or .list?) that 14:16
unmatched} dd [$_, $s.channels] for |$s.channels.List; gives the same result. 14:17
unmatched} is really pissed off.
Just spent 20 minutes wrestling with a :%arg not being passed as an Associative and now this ~_~ 14:18
timotimo oh
gfldex unmatched}: looks like $s.channels is 2 levels deep
we do have .deepmap but we don't got .deepiterator. I found myself wanting the latter a few times. 14:20
unmatched} gfldex++ thanks that was it.
Man, the % and @ sigils are warps of Perl 6. 14:21
I'm gonna avoid them from now on whenever I can
unmatched} *warts 14:21
jkramer I think I came up with the most cleverest epic web framework ideas ever, made possible by perl6! dpaste.com/2BSBVFB 14:26
gfldex m: sub deep (@a) { gather { @a.deepmap({take $_}); } }; my @l = [1,[2,3,[4,5,6]]]; .say for deep @l;
camelia rakudo-moar aa5e49: OUTPUT«1␤2␤3␤4␤5␤6␤»
gfldex unmatched}: ^^^
Juerd jkramer: Translating user input to internal identifiers. What could possibly go wrong? 14:27
jkramer I don't want to brag, but I think I'm a genius
Juerd jkramer: It should at least match against some kind of whitelist to make sure that those methods were intended to be called externally.
jkramer Juerd: it does 14:28
Juerd rereads
jkramer Juerd: return $.bad-request unless $method ~~ any(<get post head put delete>); 14:28
Juerd Oh, it does
Juerd thinks of something better to nag about 14:28
I got it! Whitelists suck! :)
jkramer :D
Juerd Instead, would there be a way to indicate this at the method itself? 14:29
Juerd multi method get (...) is web { ... } 14:30
gfldex jkramer: matching against a string literal with a signatuer is quite clever indeed. You may actually have invented that.
jkramer Juerd: I was playing with traits first, but I think since we know what HTTP methods exist it's easier/more conclusive to do it like this. Also having a get method that's supposed to be called and other get methods that are not doesn't really make sense
Juerd jkramer: Good point 14:31
jkramer Also I want to use traits for stuff like get(...) is memcached later. It's gonna be awesome :D
Juerd I like the idea
I wonder if some kind of relative paths could be used
Juerd So you can load (mount?) an application and stick it on a custom path prefix 14:32
jkramer Yeah, I'm thinking about that right now. Something like "class Foo does Blurgh has Prefix(...)" or whatever you can do with traits
Also I need a better name
Juerd Well, if you use instances instead of classes, you can mount the same application twice with different parameters 14:33
unmatched} jkramer: it's neat. Another point of awesomeness with that interface: multi method get ('blog', |c) { self.blog: |c }; multi method blog ('read', $id) {...}; multi method blog ('create', ... ) { ... }
Similar to Mojolicious's under() 14:34
Juerd I wonder, can roles have parameters?
timotimo of course they can 14:34
we have roles so that we can have parameterized types
psch m: role R[::T] { method foo { T.new() } }; class A does R[Int] { }.foo.say
camelia rakudo-moar aa5e49: OUTPUT«0␤»
Juerd timotimo: But can I load a role twice in the same class? :) 14:35
jkramer unmatched}: What is |c? Never seen that
unmatched} jkramer: a capture
hug[tab][tab] Capture
damn bot
timotimo the role bodies will probably conflict, but if the role bodies are empty that should work
well ... it might 14:36
psch m: role R[::T] { multi method foo(T $a) { $a.succ } }; class A does R[Int] does R[Str] { }.foo("foo").say
camelia rakudo-moar aa5e49: OUTPUT«fop␤»
unmatched} huggable: Capture 14:36
huggable unmatched}, Argument list suitable for passing to a: docs.perl6.org/type/Capture
psch m: role R[::T] { multi method foo(T $a) { $a.succ } }; my $a class A does R[Int] does R[Str] { }; $a.foo("foo").say; $a.foo(1).say 14:36
camelia rakudo-moar aa5e49: OUTPUT«5===SORRY!5=== Error while compiling <tmp>␤Two terms in a row␤at <tmp>:1␤------> 3ti method foo(T $a) { $a.succ } }; my $a7⏏5 class A does R[Int] does R[Str] { }; $a␤ expecting any of:␤ infix␤ infix stopper␤ …»
psch m: role R[::T] { multi method foo(T $a) { $a.succ } }; my $a = class A does R[Int] does R[Str] { }; $a.foo("foo").say; $a.foo(1).say 14:36
camelia rakudo-moar aa5e49: OUTPUT«fop␤2␤»
psch Juerd: ^^^ looks good from here 14:37
tadzik unmatched}: that reminds me of my old idea of using grammars for routing :D 14:38
unmatched} :D 14:39
tadzik: sounds cool. How did it work out?
tadzik ttjjss.wordpress.com/2010/11/ 14:40
Hacker News liked it :P
unmatched} tadzik++ that's awesome :D 14:41
tadzik :)
gregf_ m: say [1,"a"].map(*.succ) 14:42
camelia rakudo-moar aa5e49: OUTPUT«(2 b)␤»
tadzik holy batman that was over 5 years ago
jkramer When I have an attribute in a role, can I set a default value for that attribute in a class that does the role? 14:51
Just overriding has $attribute = 123 gives me an error. 14:52
timotimo you'll have to make the default value a parameter of the role, or make your BUILD do the right thing 14:53
jkramer How does the syntax for the role parameter look like?
timotimo see psch's eval a quarter-hour ago
the ::T was a type capture, so instead of giving you the value, it'll give you the type 14:54
so just use $defval instead
psch m: role R[$def] { has $.a = $def }; class C does R[10] {}.new.a.say
camelia rakudo-moar aa5e49: OUTPUT«10␤»
jkramer Ah right, I though that was for type parameters only. Thanks! 14:55
timotimo nope, ::T is a thing in any signature, and role parameters go via signatures 14:57
you should be able to do anything a regular signature can do in there
psch m: role R[&c:(Int $, Str $ --> Bool)] { has &.c = &c }; class C does R[sub (Int $a, Str $b --> Bool) { $a < 10 && $b.chars < 10 }] { }; C.new.c()(1, "foo").say 15:00
camelia rakudo-moar aa5e49: OUTPUT«True␤»
psch hadn't known that its a Signature in the brackets 15:01
dalek osystem: c4040ad | (Brian Duggan)++ | META.list:
added oauth2-client-google
 15:02
osystem: 75a8370 | Altai-man++ | META.list:
Merge pull request #230 from bduggan/oauth2-client-google

added oauth2-client-google
jkramer I have an error that I don't really understand. This line: @path.first($.prefix.arity) ~~ $.prefix (where $.prefix is a Signature :('post') and @path is ['post', 1] gives me this: Cannot convert string to number: base-10 number must begin with valid digits or '.' in '⏏post' 15:03
psch m: <foo bar baz>.first(2).say
camelia rakudo-moar aa5e49: OUTPUT«Cannot convert string to number: base-10 number must begin with valid digits or '.' in '3⏏5foo' (indicated by ⏏)␤ in block <unit> at <tmp> line 1␤␤Actually thrown at:␤ in block <unit> at <tmp> line 1␤␤»
timotimo the @path.first($.prefix.arity) runs a smart match against the number over all the things
yeah, what psch showed
psch m: <foo bar baz>.first("bar").say 15:04
camelia rakudo-moar aa5e49: OUTPUT«bar␤»
psch m: <foo 1>.first("1").say 15:05
camelia rakudo-moar aa5e49: OUTPUT«1␤»
jkramer Ah damn, I should start reading the actual description of methods in the docs
timotimo perhaps
jkramer .head is what I wanted
yoleaux jkramer: Sorry, that command (.head) crashed.
timotimo ah, hehe
gfldex .oO( Make the docs telepathic. ) 15:06
perlpilot that's one way to head-desk 15:07
jkramer Not sure if this is the best way to do it, but I added signature prefixes: dpaste.com/1PHD6CN 15:10
I guess next I should add nested routes, ie. route classes inside other route classes with dispatching to nested routes using their prefixes or something. 15:11
gfldex jkramer: you may be able to do that and more with MOP and .assuming 15:13
ab6tract jkramer: nice :) 15:17
gfldex jkramer: also, while considering different solutions you may want to measure the request/s you can achieve. It doesn't help you much the have A Most Elegant Webframework that is slow as fuck.
mst actually, for real world use I find benchmarking requests/second of a raw framework is stupid 15:19
it doesn't help you much to have spent hours micro-optimising your web framework when it turns out the bottleneck isn't in the framework 15:20
tadzik "our JS UI toolkit doesn't do jack shit but hell is it fast!"
mst gfldex: anyway, if I wanted to ricer, I'd build a gentoo stage 1 install and then write golang :D
gfldex tadzik: "our JS UI toolkit does literally everything but wont finish until tomorrow" Does that sound better? 15:24
jkramer gfldex: Currently there's no way of benchmarking anyway because HTTP::Server::Async crashes after a few requests :) Guess I should use something else or use something more generic/pluggable 15:27
mst gfldex: yes, because then you profile it to find the parts that are *actually* slow 15:28
jkramer Right now it's just a POC of the idea anyway, don't know how far I'll take it
mst jkramer: your example has a 'get /foo/new/123' that changes data
jkramer: GET requests that change data: not even once. not even in examples. please. 15:29
jkramer Pff yeah whatever mom 15:30
timotimo Mom Said To == mst 15:33
jkramer ;) 15:36
jkramer Can I use type parameters for Hash to restrict the types of both the keys and the values? I tried Hash[Int, Str] but it does weird stuff 15:57
jkramer Although I was just guessing the syntax 15:57
timotimo you can, but the parameters are the other way around from what you expect 15:58
hoelzro it would nice if we had named parameters for parameterized roles 15:59
timotimo we ... we don't?
hoelzro eg. Hash[:key(Int), :value(Str)]
hoelzro shrugs
maybe we do
perlpilot thought we did
hoelzro oh, awesome!
jkramer timotimo: I tried both ways but complains about the values type either way
perlpilot (I had exactly the same reaction in my head as timotimo)
hoelzro shows how much Perl 6 I've done recently =S
I can't wait until this house thing is done - then I can hack again \o/
jkramer m: my Hash[Int,Str] %x; %x{123} = 'foo' 16:00
camelia rakudo-moar aa5e49: OUTPUT«Type check failed in assignment to %x; expected Hash[Int,Str] but got Str ("foo")␤ in block <unit> at <tmp> line 1␤␤»
timotimo jkramer: can you write a small test case for your stuff?
jkramer m: my Hash[Int,Str] %x; %x{'foo'} = 123
camelia rakudo-moar aa5e49: OUTPUT«Type check failed in assignment to %x; expected Hash[Int,Str] but got Int (123)␤ in block <unit> at <tmp> line 1␤␤»
timotimo aha!
you're declaring that you want a hash of hashes of int and str
m: my %x is Hash[Int, Str]; %x{'foo'} = 123 16:01
camelia rakudo-moar aa5e49: OUTPUT«5===SORRY!5=== Error while compiling <tmp>␤Can't use unknown trait 'is ' in a variable declaration.␤at <tmp>:1␤------> 3my %x is Hash[Int, Str]7⏏5; %x{'foo'} = 123␤ expecting any of:␤ TypeObject␤ default␤ dyn…»
jkramer Oh shit, wrong sigil I guess
timotimo okay, is is wrong there
gfldex m: my Int %h{Str}; %h{'abc'} = 10; dd %h;
camelia rakudo-moar aa5e49: OUTPUT«Hash[Int,Str] %h = (my Int %{Str} = :abc(10))␤»
timotimo yeah, what gfldex just showed is the proper way
jkramer Ah, thanks!
gfldex we seam not to doc the value time in docs.perl6.org/type/Hash 16:04
gfldex or am i just blind? 16:04
hoelzro do we have a "
ugh, finger slipped
hoelzro do we have a "Perl 5 -> Perl 6 pitfalls" section on docs.perl6.org? 16:04
I noticed we have a "traps" section
TimToady isn't a pitfall a kind of trap? 16:05
gfldex hoelzro: there are some pitfalls stated in the 5to6 sections
MahBot gfldex, Message forwarded 16:05
unmatched} oops 16:05
gfldex EUNKNOWNBOT 16:06
MahBot Hey, everybody! ZoffixW over at local/#perl6 says I ♥ you! 16:06
unmatched} :D Multi-server tunnel with Perl 6 IRC::Client :D 16:06
timotimo cute. 16:07
unmatched} This is the code that dun it: gist.github.com/zoffixznet/0d37221...c61091bf0d 16:08
hoelzro TimToady: sure enough =) 16:08
gfldex: should the traps section link to that Perl 5 traps section? 16:09
hoelzro oh good, the trap I thought of is already in there =) 16:12
gfldex hoelzro: there is no explicit Perl 5 trap sections. Links would go all over the place.
hoelzro ok
unmatched} This trap gets me frequently: 16:13
m: my %h = :42foo :45bar; dd %h
camelia rakudo-moar aa5e49: OUTPUT«Hash %h = {:foo(42)}␤»
unmatched} m: class { method bar (:$foo, :$bar){ dd[ $foo, $bar ] } }.bar: :42foo :45bar 16:14
camelia rakudo-moar aa5e49: OUTPUT«method bar␤»
unmatched} wat? 16:15
The space is missing between dd and [, but what's with the output? :S
dalek osystem: 8866514 | (Peter Pentchev)++ | META.list:
Add Getopt::Std - POSIX getopt(3) options processing.
 16:16
osystem: 1393b78 | (Steve Mynott)++ | META.list:
Merge pull request #231 from ppentchev/roam-getopt-std

Add Getopt::Std - POSIX getopt(3) options processing.
jnthn m: dd
camelia rakudo-moar aa5e49: OUTPUT«block <unit>␤»
jnthn m: sub foo() { dd } 16:17
camelia ( no output )
jnthn m: sub foo() { dd }; foo
camelia rakudo-moar aa5e49: OUTPUT«sub foo␤»
unmatched} Ah. jnthn++
jnthn Heh, TIL. :)
timotimo i didn't even know about that
jnthn Did anyone else know about that? :P
I guess lizmat must have :)
dalek c: e19bb99 | (Wenzel P. P. Peppmeyer)++ | doc/Type/Hash.pod6:
show value constraints on Hash
 16:18
gfldex i'm very pleased with the example i found for that one
timotimo great :) 16:22
unmatched} .&dd shouldn't really be in the docs. It's Rakudo-specific helper 16:26
m: subset Powerfull of Int where * > 9000; my Powerfull %h{Str}; %h{'Goku'} = 9001;
camelia ( no output )
unmatched} :S
oh 16:27
m: subset Powerfull of Int where * > 9000; my Powerfull %h{Str}; %h{'Goku'} = 900;
camelia rakudo-moar aa5e49: OUTPUT«Type check failed in binding to assignval; expected Powerfull but got Int (900)␤ in block <unit> at <tmp> line 1␤␤»
gfldex unmatched}: the ecosystem may disagree with you when it comes to dd 16:30
TimToady the ecosystem does not define Perl 6 16:32
mst unmatched}: Powerful
TimToady and roast doesn't test dd
unmatched} gfldex: no, the ecosystem uses a Rakudo-specific helper that's not part of the Perl 6 specification
dalek c: 9b1ca52 | (Zoffix Znet)++ | doc/Type/Hash.pod6:
s/Powerfull/Powerful/

  mst++		 16:36
gfldex TimToady: the ecosystem defines how Perl 6 is used. I'm very unhappy with all those says in the docs because it gives the impression that say is a good default. Data will be lost, tears will be shed. 16:46
[Coke] I agree that dd shouldn't be in the doc site. 16:52
I'm OK with say being used for things that have trivial .gists, but would be fine if more complicated things used put.
(I'm also ok if we just use put everywhere) 16:53
unmatched} put's not really the same thing tho 16:56
m: put IO::Socket::Async
camelia rakudo-moar aa5e49: OUTPUT«Use of uninitialized value of type IO::Socket::Async in string context␤Any of .^name, .perl, .gist, or .say can stringify undefined things, if needed. in block <unit> at <tmp> line 1␤␤»
unmatched} m: say IO::Socket::Async
camelia rakudo-moar aa5e49: OUTPUT«(Async)␤»
unmatched} This is curious: 16:57
stmuk say :$foo.perl is a little like dd
unmatched} m: put -∞..∞; say -∞..∞
camelia rakudo-moar aa5e49: OUTPUT«*..*␤-Inf..Inf␤»
timotimo it not being the same thing is the whole point, isn't it?
unmatched} Sure, but it's not suitable to replace all of `say`s in the docs. 16:58
timotimo not all of it, of course
[Coke] m: my $a = -∞..∞; say ~$a; say $a;
camelia rakudo-moar aa5e49: OUTPUT«*..*␤-Inf..Inf␤»
timotimo i thought the problem was that say tends to abbreviate lists and such?
[Coke] unmatched}: we can't *just* replace say with put, no
dalek c: 2bd91a5 | (Zoffix Znet)++ | doc/Type/Hash.pod6:
Simplify example
 17:03
[Coke] yay, got make test working again locally for perl6/doc 17:05
dalek c: 267c0ad | (Zoffix Znet)++ | doc/Type/Hash.pod6:
Add missing comment marker
gfldex m: subset Powerfull of Int where * > 9000; my Powerfull %h{Str}; try { %h{'Goku'} = 900; CATCH { say $_ } } 17:15
camelia rakudo-moar aa5e49: OUTPUT«Type check failed in binding to assignval; expected Powerfull but got Int (900)␤ in block <unit> at <tmp> line 1␤␤Type check failed in binding to assignval; expected Powerfull but got Int (900)␤ in block <unit> at <tmp> line 1␤␤Actually throw…»
gfldex the reason I use dd there is that I got the fully quallified name of the exception for free 17:16
unmatched}: if you remove the type of the exception from the example please add that type in the descriptive text 17:18
m: subset Powerfull of Int where * > 9000; my Powerfull %h{Str}; try { %h{'Goku'} = 900; CATCH { say $_.WHAT } } 17:21
camelia rakudo-moar aa5e49: OUTPUT«(Binding)␤Type check failed in binding to assignval; expected Powerfull but got Int (900)␤ in block <unit> at <tmp> line 1␤␤»
unmatched} .^name
gfldex m: subset Powerfull of Int where * > 9000; my Powerfull %h{Str}; try { %h{'Goku'} = 900; CATCH { when X::TypeCheck::Binding { put .^name } } }; say 'alive'; 17:23
camelia rakudo-moar aa5e49: OUTPUT«X::TypeCheck::Binding␤alive␤»
dalek c: 46e69e3 | (Zoffix Znet)++ | doc/Type/Hash.pod6:
Restore exception message into code example
 17:25
unmatched}
.oO( a "chat" section on perl.org that lets you chat on this channel... )
 18:02
gfldex unmatched}: i had the wild idea to have webchat.freenode.net/ on docs.perl6.org 18:03
"if you have any further questions ..."
or some easy way to seach the irclogs 18:04
ENEEDMORETEA
mst that would be wesome
sena_kun m: say say Int.^methods(:all).pick.name;
camelia rakudo-moar aa5e49: OUTPUT«sech␤True␤»
mst awesome
DBIx::Class pioneered that with mibbit, it's been very nice
unmatched} Yeah, I like the "Questions? Chat with us" button (on the left, middle) here: metacpan.org/pod/Moose 18:05
sena_kun m: say Int.^methods(:all).pick.name;
camelia rakudo-moar aa5e49: OUTPUT«fmt␤»
mst actually got the 'other' box added to mibbit specifically so irc.perl.org could use it
unmatched} :o
mst sadly they threw an epic tantrum at the freenode opers a few years back and got very banned
unmatched} wow 18:06
mst culminating in them saying "no, we won't guarantee that we won't mine users' chat history to serve better converting ads"
unmatched} I was thinking of more a rolling chat. You don't have to log in or anything. You just load the site and you can see humans talking and you can talk too 18:07
mst hm
mst oooh, what you'd want for that is something like a convos instance, except one that maintained a bot, so it only had to actually connect a nick when you started trying to speak 18:07
unmatched} I've never used convos. I'm thinking it doesn't even need to connect a nick. The bot would make up a nick based on IP address and would prefix messages with that nick, talking on behalf of that person. 18:08
I think #go-lang has this sort of thing
Well, maybe not the via their site, but I've seen a bot talk on behalf of people 18:10
mst oh, you mean more like the XMPP bridge in #tcl
unmatched} Probably 18:11
dalek osystem: ba55f31 | (Matt Oates)++ | META.list:
Add Math::FourierTransform

First features for FourierTransform		 18:12
Ulti ^ was surprisingly slick to write in P6 18:14
looks nicer than the python just because maths functions for complex numbers are already there and you don't need a complex numbers import 18:15
unmatched} My first full draft of IRC::Client blog post is 5325 words >_<
Ulti blag golf
unmatched} github.com/MattOates/Math--FourierTransform 18:16
TimToady flog galb
m: complex 18:18
camelia rakudo-moar aa5e49: OUTPUT«5===SORRY!5=== Error while compiling <tmp>␤Undeclared routine:␤ complex used at line 1␤␤»
TimToady pity we don't have native complex yet 18:19
sena_kun m: say ((1, 2), <a b>).flatmap(&uc).join('|'); # 1|2|A|B
camelia rakudo-moar aa5e49: OUTPUT«1 2|A B␤»
TimToady oh, I guess Complex is compact nums though already
Ulti is it? I assumed it was two Rats or something underneath? 18:20
TimToady well, most complex computations don't come out exact anyway 18:21
so Rats would be mostly wasted
and turn into nums anyway
Ulti yeah I noticed my code didn't really match other langs functions 100%
should probably do an actual numeric test using sin() 18:22
feels like maths functions someone should write a general testing framework for everyone to test against
mantovani (stupid, stupid, stupid question) why does you guys created MoarVM and just used jvm as backend ? 18:29
and not just used jvm*
a friend of my ask it, and I didn't know what to answer 18:30
mantovani oh I found, sorry 18:31
perl6advent.wordpress.com/2014/12/...6-in-2014/
japhb mantovani: A few different reasons, but for *me* the most important is that JVM has really slow startup. Sucks for interactive tools.
mantovani "MoarVM provides much lower memory usage, faster startup time, and is significantly faster than parrot at runtime"
japhb Parrot has gone "biologically stable" 18:32
unmatched} Also, we can modify MoarVM for Perl 6 needs :)
mantovani since perl 6 run on jvm I thought it satisfy all perl 6 needs 18:33
geekosaur well, it sort-of runs... 18:34
unmatched} Well, you can drive a screw in with a hammer, but it's much better to have a screw driver.
mantovani understood
arnsholt To expand a bit on it, the JVM conception of classes and dispatch and such isn't quite the same as the Perl 6 one 18:35
So it's not just a matter of compiling to JVM classes and invoking JVM methods directly on those objects 18:36
So the Rakudo/JVM runtime code has to do a fair bit of legwork to translate between the different models, leading to increased overhead
mantovani so the jvm implementation was just to learn/for fun ? 18:38
dalek c: a8bb893 | (Wenzel P. P. Peppmeyer)++ | html (2 files):
add link to webchat
 18:39
unmatched} mantovani: no, to run on JVM. Some organizations, for example, require the projects must run on JVM, not on some arbitrary platform
[Coke] a big selling point for running on the JVM for me was that I could run it in a managed j2ee container. this became much less relevant with docker.
mantovani so jvm implementation will be stable as MoarVM ?
[Coke] but up to about six months ago, rakudo on jvm was going to be the only way to get perl 6 running here at work. 18:40
mantovani: that's the goal, but we're not there.
arnsholt Also, the JVM backend was implemented before Moar was started
In the beginning, there was Parrot
mantovani ok, because I come here in the channel few days ago ... because the best of java are the libraries and since perl 6 run it on jvm should not be hard use jar libraries in perl as I do in scala. 18:41
arnsholt Then jnthn++ implemented multi-backend, and Rakudo ran on both Parrot and JVM
Then, later on (a year or two, I think) MoarVM was implemented
sena_kun Is `yada` method name an abbreviation or it just comes from "いやだ"?
unmatched} sena_kun: I'd think this www.dictionary.com/browse/yada--yada?s=t 18:42
sena_kun unmatched}, thanks. How interesting. 18:43
arnsholt mantovani: I think we have at least partial JVM interop 18:47
But the exact translation of Perl 6 method invocations to Java invocations are a bit tricky, because method resolution is different in Java and Perl 6
sena_kun m: my ($a, $b) = SetHash.new(1, 2, 3), SetHash.new(2, 4); say $a (^) $b; 18:48
camelia rakudo-moar aa5e49: OUTPUT«set(4, 3, 1)␤»
sena_kun Is this elements order machine-specific?
timotimo our hashes don't keep order around 18:50
perlpilot sena_kun: sets aren't ordered
gfldex pitty the irclog doesn't show that the webchat was used to join. Would be interresting to see if the link on docs.perl6.org actually lures folk in.
sena_kun Thanks. 18:50
hoelzro gfldex: we could change the link to redirect to the webchat, and see how many people hit that redirect link 18:51
psch arnsholt: mmd on Java classes is mostly solved. we do lack marshalling for Positional and Associative to corresponding Collection classes though
arnsholt: what i'm anticipating to be the hardest part is translating between interfaces and Roles though
unmatched} m: gist.github.com/zoffixznet/729acd1...9619eba91f 18:52
gfldex hoelzro: it's in the apache logs. But so is every spider and bot that hits the site. I somehow doubt the bots would solve the captcha to talk to us.
arnsholt psch: Oh, cool! I've been away long enough that so many of the things that were hard to fix have been figured out \o/ 18:53
camelia rakudo-moar aa5e49: OUTPUT«(timeout)Starting irc.freenode.net␤Starting localhost␤irc.freenode.net: IO::Socket::Async<139743539025240>␤Got stuff! :kornbluth.freenode.net NOTICE * :*** Looking up your hostname...
unmatched} I guess expecting redo inside a promise to redo a loop is expecting too much?
I'm trying to pop open a bunch of sockets and reconnect them when they disconnect :/
hoelzro ahhh 18:54
psch j: use java::lang::System:from<JavaRuntime>; System.get_out.println("foo")
camelia rakudo-jvm cd19db: OUTPUT«foo␤»
unmatched} gfldex: match IPs between chat logs and apache logs
psch arnsholt: fields aren't wrapped in Proxies yet, which is also hard, but Perl 6 level hard, not codegen hard. hence 'get_out' above :) 18:54
gfldex unmatched}: the irclogs that i can see don't got ips 18:55
unmatched} Maybe someone here has more detailed logs
sena_kun m: my $n = bag "a" => 0, "b" => 1, "c" => 2, "c" => 2; say $n.keys.perl; # ("a" => 0, "b" => 1, "c" => 2, "c" => 2).list 18:57
camelia rakudo-moar aa5e49: OUTPUT«(:c(2), :b(1), :a(0)).Seq␤»
sena_kun Wow. 18:58
gfldex @gateway/web/freenode/ip is the string to grep for
unmatched} Hm, I'm gonna try to reimagine my problem as a supply with socket disconnection events 18:59
unmatched} m: react { whenever supply { start { emit 42 }; } { .say } }; sleep 2 19:12
camelia ( no output )
unmatched} I can't emit into a supply from within a Promise?
avar Does anyone here know offhand whether Inline::Perl6 embedded in a perl without multiplicity or threading is capable of : perl5 -> perl6 -> *call subs in perl5* 19:13
TimToady m: my $n = ("a" => 0, "b" => 1, "c" => 2, "c" => 2).Bag; say $n.perl 19:14
camelia rakudo-moar aa5e49: OUTPUT«("c"=>4,"b"=>1).Bag␤»
TimToady sena_kun: ^^^ looking for that? 19:15
this is why the composer is different from the coercer
sena_kun TimToady, not really. I'm just checking correctness of example's output in our docs. And it seems our real compiler behavoir is slightly different from what documentation says. 19:17
TimToady m: my $n = bag "a" => 0, "b" => 1, "c" => 2, "c" => 2; say $n.perl
camelia rakudo-moar aa5e49: OUTPUT«(:c(2)=>2,:b(1)=>1,:a(0)=>1).Bag␤»
TimToady well, keys should never list "c" => 2 twice 19:17
sena_kun TimToady, it really should. That's because I check and fix it now. :) 19:18
Or rather "Because of that", oh. 19:20
unmatched} Solved my problem with a Channel. Rough version: gist.github.com/zoffixznet/24d292d...f8d2754e03 19:40
harmil_wk m: sub infix:<↑>($a, $b) is assoc<right> {$a ** $b}; sub infix:<↑↑>($a, $b) is assoc<right> { [↑] $a xx $b }; sub infix:<↑↑↑>($a, $b) is assoc<right> { [↑↑] $a xx $b }; sub infix:<↑↑↑↑>($a, $b) is assoc<right> { [↑↑↑] $a xx $b }; say 3↑↑3 19:45
camelia rakudo-moar aa5e49: OUTPUT«WARNINGS for <tmp>:␤Useless use of $a in sink context (lines 1, 1, 1)␤7625597484987␤»
harmil_wk Am I missing something, there? Seems like there should be no warning.
unmatched} bisect: m: sub infix:<↑>($a, $b) is assoc<right> {$a ** $b}; sub infix:<↑↑>($a, $b) is assoc<right> { [↑] $a xx $b }; sub infix:<↑↑↑>($a, $b) is assoc<right> { [↑↑] $a xx $b }; sub infix:<↑↑↑↑>($a, $b) is assoc<right> { [↑↑↑] $a xx $b }; say 3↑↑3 19:47
[Coke] it doesn't warn for me locally on 2016.07.1
bisectable unmatched}: Exit code is 0 on both starting points, bisecting by using the output 19:48
unmatched}: bisect log: gist.github.com/369412d2b8ff352b2e...b35f89846f
unmatched}: (2016-07-26) github.com/rakudo/rakudo/commit/5334cb7
unmatched} harmil_wk: would you please open a ticket for it?
harmil_wk Sure
unmatched} huggable: rakudobug
huggable unmatched}, To report a bug, email detailed description and any test cases to [email@hidden.address] or use perl6 query on rt.perl.org ; see github.com/rakudo/rakudo/#reporting-bugs
harmil_wk I've been pretty prolific with the bug reports of late. I'm worried I'm going on a list ;) 19:49
unmatched} :)
[Coke] harmil_wk: you'll never catch up with masak, so dont' worry. :)
unmatched} harmil_wk: oh and would you include the link to the commit that bisectable gave above? 19:50
TimToady: if you're around and have time, seems it points to the sinkiness propagation: github.com/rakudo/rakudo/commit/5334cb7
[Coke] unmatched}: note that, in general, we want files attached to tickets, not links to pastebots.
but I realize that's a pita in general. 19:51
unmatched} [Coke]: I mean the link to the commit
unmatched} & bus
[Coke] \o/ Spinning up a perl 6 docker container at work for a CI solution.
unmatched}: WHOOPS, my bad, sorry
AlexDaniel [Coke]: generally, is there any need to provide bisect log at all? If one doubts that this is the right commit, he can always build on commit^ and check it himself. 19:52
[Coke] on the ticket? not required, but can't hurt, I'd wager. 19:53
[Coke] finds -13- perl 6 docker containers on docker registry. huh 19:54
harmil_wk rakudobugged it. Also, can I just say that being able to define ↑↑↑↑ so trivially makes me feel awe, joy and horror all at the same time. 19:55
AlexDaniel commit 5334cb7 sub infix:<↑>($a, $b) is assoc<right> {$a ** $b}; sub infix:<↑↑>($a, $b) is assoc<right> { [↑] $a xx $b }; sub infix:<↑↑↑>($a, $b) is assoc<right> { [↑↑] $a xx $b }; sub infix:<↑↑↑↑>($a, $b) is assoc<right> { [↑↑↑] $a xx $b }; say 3↑↑3 19:56
committable AlexDaniel: ¦«5334cb7»: WARNINGS for /tmp/Mzi8nSUmqf:␤Useless use of $a in sink context (lines 1, 1, 1)␤7625597484987
AlexDaniel commit 5334cb7^ sub infix:<↑>($a, $b) is assoc<right> {$a ** $b}; sub infix:<↑↑>($a, $b) is assoc<right> { [↑] $a xx $b }; sub infix:<↑↑↑>($a, $b) is assoc<right> { [↑↑] $a xx $b }; sub infix:<↑↑↑↑>($a, $b) is assoc<right> { [↑↑↑] $a xx $b }; say 3↑↑3
committable AlexDaniel: ¦«5334cb7»: 7625597484987
AlexDaniel just wanted to make sure because there's some “timed out after 10 seconds” warning in the bisect log…
AlexDaniel MasterDuke: oh uh! Did I break it? :) 19:58
[Coke] hoelzro: what is your docker perl6 instance? mj41's? 19:59
oh, I can just build my own, let me do that. :)
gfldex we have map and deepmap but grep and <insert-missing-sub-here>. Is there a reason for the gap? 20:45
masak gfldex: do you see a use case? my guess is someone did for deepmap 20:46
gfldex masak: we get the question why flat doesn't flatten recursively quite often. And my suspicion is that they actually ask for deepgrep (because deepmap is already there). 20:48
jnthn Wouldn't they want deepflat? :) 20:50
mst oooh, deepgrep would be really cool
masak so, deepgrep would traverse things in pre-order, omitting whole subtrees when it got falsy from its callback? 20:51
gfldex we do have flatmap, but that doesn't recurse
jnthn Right, nor does flat
Well
It does until it sees a Scalar container
gfldex m: sub deep (@a) { gather { @a.deepmap({take $_}); } }; my @l = [1,[2,3,[4,5,6]]]; .say for deep @l; 20:52
camelia rakudo-moar aa5e49: OUTPUT«1␤2␤3␤4␤5␤6␤»
jnthn Which is fine, but then an Array is a bunch of Scalar containers, so .flat on an array is just identity
masak .oO( "Finally, just deepzip your Supply with a HyperBag. As you can see, it's easy once you see the trick!" )
gfldex deep would be a recurseing iterator
jnthn Something along those lines would deal with the boilerplate. deep isn't verb, alas. 20:53
gfldex we don't got that in core either (and lizmat would like make it faster then I did)
geekosaur spelunk? :p 20:54
jnthn So it'd need to be something like visit or walk or something better. :)
gfldex visit or I would have to change a module :-> 20:55
jnthn Oh no, naming...time to run away. :-)
gfldex for descend @a {}
dalek c/literal_examples: 405fd02 | Altai-man++ | doc/ (30 files):
Almost all example outputs were changed to literal versions
 20:57
masak descend++ 20:58
gfldex m: sub deep (@a) { gather { @a.deepmap({take $_}); } }; say [+] deep (1,2) Z (3,4);
camelia rakudo-moar aa5e49: OUTPUT«10␤»
gfldex masak: ^^^ usecase (because we dont got |Z)
gfldex slippy metaops might kill most of the usecases tho 21:00
avar Suggestion for a namespace to package perl6 modules under perl5 namespaces? 6Wrap::* ? 21:29
Maybe just Six::* 21:30
kurahaupo a namespace in perl5 that starts with a digit is going to be entertaining to use 21:31
timotimo how about roman numeral 6? 21:36
avar I think I'll go for Six::*, easy to say, not in use 21:37
Ulti geekosaur: spelunk more like dive :P 21:44
or maybe delve
AlexDaniel sena_kun: wow, what a lovely job you did 21:50
sena_kun AlexDaniel, thanks. We have one more question to resolve, look my last comment on the issue page. 21:51
timotimo where do i have to look to see what you're talking about? 21:52
sena_kun Let's see... 21:53
AlexDaniel github.com/perl6/doc/issues/761 github.com/perl6/doc/pull/775
sena_kun Ah, already ansered. 21:54
dalek c: 405fd02 | Altai-man++ | doc/ (30 files):
Almost all example outputs were changed to literal versions
 21:56
c: 34f2b67 | (Wenzel P. P. Peppmeyer)++ | doc/ (30 files):
Merge pull request #775 from perl6/literal_examples

Almost all example outputs were changed to literal versions
AlexDaniel sena_kun: we need more options 21:59
sena_kun AlexDaniel, I'll update my answer then. Perhaps it's reasonable to close this issue(since it's solved) and file new? 22:01
AlexDaniel sena_kun: yes
sena_kun AlexDaniel, okay.
timotimo cool 22:03
sena_kun New issue - github.com/perl6/doc/issues/776 22:08
harmil_wk Have to go eat, but thought this would amuse: github.com/ajs/perl6-Math-Arrow 22:46
AlexDaniel sena_kun: not sure if I understand your point. camelia already runs the latest rakudo, and you can use star: to run it on the latest Star release. I do agree that you should not depend on camelia for writing docs (you probably want to configure your setup so that you can type such characters), but I still did not get what you mean… 23:10
sena_kun AlexDaniel, eeeh, it seems my grammar gets worse and worse with time... I'll try to explain now. 23:11
But firstly, one question. I cannot have camellia-like output if I only have rakudo built on my machine? 23:12
My point is that if I cannot, then it will be hard to type characters like newline and "<<"/">>" everytime. If I have camelia on my localhost with default rakudo - it's easy to just copy-paste output. But if not, then I need to type such characters everytime. 23:14
timotimo oh, you don't have a compose key or something?
AlexDaniel sena_kun: I am kinda playing devil's advocate here, but why don't you just PM real camelia on freenode? :) 23:15
timotimo might not always have internet connectivity
AlexDaniel and to fix that you are ready to run camelia locally with a your own local irc server just to not have to type unicode characters? I don't know, it sounds weird… 23:16
-a
sena_kun timotimo, nope. And I, to be honest, don't want it too much. But it's my problems and I doubt it relevant to question.
Just as a pure example, I suppose. I cannot be docs writer, so it's just hypothetical here. 23:17
It's only plain old me, so if majority thinks unicode formatting is cool - I'm okay with it. 23:18
*think
Ah.
sena_kun I'll make a note to my comment now. 23:20
timotimo i'm a bit annoyed that i can't easily type most emoji on my keyboard
AlexDaniel m: my $output = “hello world\ntest\nfoo”; say “OUTPUT: «$output.subst(:g, “\n”, ‘␤’)»”
camelia rakudo-moar aa5e49: OUTPUT«OUTPUT: «hello world␤test␤foo»␤»
AlexDaniel sena_kun: what about this ↑ ?
sena_kun AlexDaniel, hmm. Is it formatting function for an author to use or you propose to lie it on repository shoulders? If first, I'm okay with it. 23:23
AlexDaniel sena_kun: I mean that you can replicate what camelia does with a simple one-liner, it's not a big problem 23:24
sena_kun AlexDaniel, ah, that's what it is. Yes, then it's no problem. And even if you're without internet connection, you can just copy characters from some another page. Okay, this point is solved. 23:25
sena_kun I've updated my comment. 23:26
AlexDaniel I'm still against «» just because a lot of examples in the docs tend to have (), and «()» is OK but less than awesome. But I really think that typability of some characters should not be the main deciding factor in this case 23:27
sena_kun Sorry, guys, but I'm done already for today. Good night, #perl6. 23:28

Powered by Cro and the Raku® Programming Language.

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

Thank you!