samcv maybe i just need to make my function faster. that is part of it at least.hm 00:02
timotimo, do you know if it cache's the result of nqp::fc and returns the same fc string if i repeatedly request the foldcase version for the same string? 00:04
timotimo no, it won't cache that 00:07
i'll be playing vidya game with a friend now :)
samcv ok 00:08
timotimo and after that probably sleep 00:12
samcv \o/ ok i got it. woo 00:38
only about 16% slower than nqp::index 00:41
from initial estimates
and the speed of worst-case caseinsensitive matching goes from 2.26970566 to 1.2345366s 00:43
comparing from a release a month or two ago
timotimo samcv: is that a fake number? 1.234? :) 01:32
samcv no 01:37
real number :) 01:38
but if i made up a number it could very well be that. heh
Geth roast: 389db59ff9 | (Samantha McVey)++ | S03-operators/repeat.t
[unfudge] Huge repeat counts become neg. numbers RT #128035
 05:49
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=128035
Geth nqp/foldcase: 4 commits pushed by (Samantha McVey)++ 06:08
samcv nqp-jvm: use NQPHLL;say(nqp::getcomp('jvm').language) 06:10
camelia Error occurred during initialization of VM
Could not allocate metaspace: 1073741824 bytes
samcv nqp: use NQPHLL;say(nqp::getcomp('jvm').language)
camelia Cannot call method 'language' on a null object
at <tmp>:1 (<ephemeral file>:<mainline>)
from gen/moar/stage2/NQPHLL.nqp:1587 (/home/camelia/rakudo-m-inst-2/share/nqp/lib/NQPHLL.moarvm:eval)
from gen/moar/stage2/NQPHLL.nqp:1790 (/home/camelia/rakudoā€¦
samcv nqp: use NQPHLL;say(nqp::getcomp('moar').language)
camelia moar
[TuxCM] This is Rakudo version 2017.03-53-g5ca924df2 built on MoarVM version 2017.03-31-g8f9325b8 06:18
csv-ip5xs 3.069
test 12.619
test-t 4.897 - 4.928
csv-parser 12.989
samcv o/ [TuxCM]
[TuxCM] o/ 06:19
samcv [TuxCM], is there a way i can dl yoru perl6 test and bench it locally
hopefully i can bump that speed up with 1.7x :i speed boost coming very soon
also would like to just profile it in general
[TuxCM] all url's have been posted here repeatedly
modules.perl6.org/ 06:20
github.com/Tux/CSV
samcv the test you actually run though. i know those url's
but the benchmarks
the actual thing you run or whatever
[TuxCM] there is a README.speed
samcv ok thanks you. will read that 06:21
lizmat Files=1181, Tests=55992, 196 wallclock secs (11.64 usr 4.68 sys + 1159.57 cusr 112.20 csys = 1288.09 CPU) 07:36
www.flickr.com/photos/wendyga/3365...619026503/ # Camelia on stage in Brussels at ARW concert 08:35
www.flickr.com/photos/wendyga/3331...619026503/ # closeup 08:36
masak :) 08:55
camelia is entering into her rock phase...
Zoffix huggable: csv 10:05
huggable Zoffix, github.com/Tux/CSV (see `bench csv` for how to run bench)
Zoffix huggable: bench csv
huggable Zoffix, export PATH=`pwd`/install/bin:$PATH; cd CSV; for i in $(seq 1 10000); do echo 'hello,","," ",world,"!"'; done > /tmp/hello.csv; time perl6 -Ilib -MText::CSV test-t.pl </tmp/hello.csv
Zoffix samcv: ^ that's it. You can ignore the export path part 10:06
NeuralAnomaly: status 10:13
NeuralAnomaly Zoffix, [āœ˜] Next release will be in 2 weeks and 4 days. Since last release, there are 13 new still-open tickets (13 unreviewed and 0 blockers) and 56 unreviewed commits. See perl6.fail/release/stats for details
Zoffix Damn time flies.
jnthn swats a time fly 10:15
Zoffix I wanna cram all the IO updates into 2017.04 release and somehow thought there were still 4 weeks until the release :) 10:19
So 6.d is additive, but can we replace entire classes, say with `class IO::Path {ā€¦}` or will it cry about redeclaration? 10:47
guess I can try and see...
(not gonna do it, but pondering whether I can write `FastIO` module that ditches $*SPEC with the assumption that it can later be just copy-pasted into 6.d or whatever language) 10:50
lizmat my question is really: what do we need $*SPEC for? 10:53
basically preprocessing paths before you get an absolute path that you can feed to the OS
Geth rakudo/nom: 42f34fb8ae | (Pawel Murias)++ | 2 files
Move tests for is_approx to a separate file

It's deprecated so it's tested separately.
Zoffix lizmat: but also to stringify them 10:57
m: "../".IO.relative("/tmp/foo/bar").say
camelia ../../../home/camelia/..
Zoffix m: "../".IO.relative("/tmp/foo/bar").resolve.say 10:58
camelia No such method 'resolve' for invocant of type 'Str'
in block <unit> at <tmp> line 1
Zoffix m: "../".IO.relative("/tmp/foo/bar").IO.resolve.relative.say
camelia ..
Zoffix And modify them... .child/.parent/.concat-wtih
m: my $*SPEC = IO::Spec::Win32; "../".IO.relative("/tmp/foo/bar").say 10:59
camelia ..\..\..\home\camelia\..
Zoffix My biggest worry is switching to unorthodox path separators on Windows. Sure they work with the OS, but people will also feed them to all sorts of stuff. 11:00
geekosaur define "unorthodox" 11:01
Zoffix And the headlines that will follow: "Slow ass Perl 6 optimized by using wrong Windows paths and now it broke XYZ"
Zoffix needs to get ready to work 11:02
lizmat Zoffix: my approach to this was: from outside s/\\/\// :-)
and on output, if Win, then change all / to \
output like shelling out 11:03
geekosaur well. there's the "orthodox" that says that all paths shall be unixlike, and there's the "orthodox" that says that Windows paths use \ as the separator and / is compatibility shim 11:04
and you can;t win with the "using wrong Windows paths" because half the programs you run into will only work right with one or the other 11:05
so no matter which you use, you will be wrong some of the time
"wrong"
lizmat goes afk again 11:06
Zoffix m: '/tmp/foo3\lulz2'.IO.mkdir; say grep *.contains('lulz2'), dir '/tmp/' 11:18
camelia ("/tmp/foo3\lulz2".IO)
Zoffix And on Windows that'd be a folder within a folder 11:19
geekosaur yes? 11:22
Zoffix doesn't understand the question 11:23
geekosaur doesn;t udnerstand your point unless it's "but it mixed path separators we must hide this from the little kids"
...if it confuses them then perhaops they should not use an OS which treats them both the same *usually* 11:24
Zoffix No, my point was that if abandoning $*SPEC and preprocess paths the same on all OSes, the above case is problematic 11:26
calling .parent() on it should give different results on Windows and *nix 11:27
geekosaur scans back to see what that's about 11:28
uh
no, getting rid of $*SPEC for *buggy* speed up is terrible idea
do NOT do this 11:29
or do it and stop advertising Windows compatibility
Zoffix Right
But there's still a lot we can gain by making SPEC internal and getting rid of the dynvar and all the methods that just return '.', '..' 11:31
Which I'll attempt to do in a module 11:33
Geth roast: cc5ee3d7f0 | (Pawel Murias)++ | S24-testing/10-is-approx.t
Test is-approx.
 11:36
geekosaur the dynvar maybe. I'd be careful about the others unless you are certain Perl 6 will never be ported to an OS with different conventions 11:37
yes, I'm sure it's faster, just like it would be to ignore how path separators work on Windows
Zoffix The port would take care of adding back the . and .. :) 11:39
geekosaur ...what? 11:40
(this looks worrisome anyway, how do you deal with ... uh, which OS used a sequence of // to back up instead of a .. ?) 11:41
Zoffix People porting to such an OS would add support for it.
Just like now they'd have to add a specific ::Spec::* module for it 11:42
Woodi I totally don't get that problems with paths... a) during VM compile time, installation and precompilation require *correct* paths from current OS; b) in runtime user should provide correct paths... what is missing ? 12:18
Zoffix Woodi: "/tmp/foo3\lulz2" is *correct* path on both Linux and Windows. But that `\` in it has different meaning depending on the OS. 12:21
geekosaur which is confused by the fact that the CLI pretty much requires you to use \ on Windows, but the OS API accepts either (because the CLI considers / the start of a command line option) 12:23
Woodi then only one thing can be done: stick to common case && provide specialised "use Windows::Quirks;" 12:29
geekosaur what is the "common case"? 12:31
(hint: there is no correct answer. you don't get to force one on Windows.)
Woodi '\' on *nix, '/' on Win, ':' on Mac ? and user need to provide sensible paths 12:32
geekosaur er?
that's completely wrong...
Woodi what ? yes, '/' on Unix :)
geekosaur ':' was only used in MacOS 9 and earlier 12:33
Woodi geekosaur: now you are lost in details
afk
geekosaur uh
again, you don;t get to assert that
Zoffix :)
geekosaur an OS X API will not parse ':' as a path separator. you don't get to assert otherwise unless you want to give up compatibility with anything else 12:34
and, Windows APIs accept either / or \. again, you don;t get to assert otherwise. the OS gets to assert it, and it does assert it; you must comply, not try to argue otherwise 12:35
Zoffix
.oO( resistance is futile... )
stmuk_ wibble 12:48
Woodi b. I'm proposing limiting user input to common cases. in case someone want more complicated things he needs to use Pro:SpecificOS module. same story as with object-relational mappers - at some level of control you need to drop sugar. 13:12
Zoffix he? 13:13
I'm not sure how you plan on limiting the presence of 'foo\bar' directory when doing `dir` for example
Woodi but details: on Linux/Unix if someone provide "/tmp/foo3\lulz2".mkdir - not a problem. on Windows "/tmp/foo3/lulz2".mkdir is error, "\tmp\foo3\lulz2".mkdir is ok. on windows, "/tmp/foo3\lulz2".mkdir is error - command starts with switches 13:15
do we plan to support case when user do not know what os he use ? 13:16
Zoffix And it seems to come from a viewpoint that the user manually types up all the paths into their code, when more often than not, the paths would come from somewhere else. So what you'd have is most people writing programs that crash on certain input and more skilled programmers always including Pro::SpecificOS modules simply to prevent those crashes.
Woodi: we already do :/
Woodi Zoffix: require Pro::Virtualos :) 13:17
geekosaur woodi, so you want perl 6 not to accept paths that other programs can legally produce on Windows?
but it sounds like this will have to be a learning experience because multiple people think this somehow makes sense. 13:18
Woodi geekosaur: if path are invalid for subsystem we use internally then yes.
geekosaur the path is not invalid for Windows
Woodi geekosaur: but it's nor "common case". and it is resolved by including Pro:: :) 13:19
geekosaur the path is invalid for a cheerful charlie path parser that deliberately violates the platform specification in the name of speed
this is not acceptable
and again, you do not get to say what the common case is
the actual common case on Windows? you don;t normally get paths that *mix* them, but you will often get paths that use one or the other
Woodi cc is: *nix path separator is '/', on Windows '/'. 13:20
perlpilot
.oO( the "common case" is fubar :) )
geekosaur but even so, I think you do not get to declare "common case" unless you want to not interoperate with other Windows programs
Woodi you know what on Window :)
geekosaur which is your call, I guess, if you like telling Windows users to go away
Woodi geekosaur: if someone want to operate he include Pro::Windows 13:21
geekosaur no
oh never mind
yes, please do that
Zoffix heh
That's not gonna happen./
geekosaur you will evidently need to learn the hard way what is acceptable to Windows users
so yes, please do that
Woodi geekosaur: "Windows user" is Explorer user by definition - he use path picker :) 13:22
geekosaur ...
I'm out. please do what you want.
Woodi geekosaur: I would go, you pls stay, seriously 13:23
geekosaur no, I am out of this conversation. you want to dictate to Windows users what is correct, you dictate to Windows users what is correct.
you find out how fare it gets yuou, *then* we can consider having this conversation. 13:24
perlpilot This whole "common case" thing reminds me of character encodings. Either you know what encoding you need to use or you guess. The "common case" here sounds like guessing. And, of course, guessing is rife with pain and suffering. 13:27
Woodi just dividing problem space on cases. I do not say what's correct becouse it would deny the reality 13:28
Zoffix lizmat: so how did camelia manage to get on stage of a concert?
lizmat: also is that picture private or can I tweet it? 13:29
C:\temp>perl6 -e "say dir 'foo/bar'" 13:31
("foo/bar\ber".IO "foo/bar\meow".IO)
Zoffix is disappointed with those path separator mixage in output :() 13:32
Especially since it's double-quoted and wrong
At least .perl is right 13:33
timotimo good catch on the backslashes there
perlpilot Zoffix: pretty sure woolfy posted on FB about how they gave a camelia to each of ARW before their show in Brussels 13:35
Zoffix Ah, now I see it :)
timotimo i don't know who ARW are :| 13:36
perlpilot timotimo: Yes 2.0
Zoffix doesn't know who either are :)
geekosaur wouldn't this be something like 5.0? 13:37
perlpilot whatever, I can't keep track of how many incarnations their are/were :)
geekosaur :)
en.wikipedia.org/wiki/Yes_(band) and it looks like 4.0 13:38
Yes, ABWH, Cinema/"Yes 2.0", ARW 13:39
timotimo yes, yes.
perlpilot Also, en.wikipedia.org/wiki/Anderson,_Ra...nd_Wakeman
Zoffix is amused en.wikipedia.org/wiki/No_(band) is not a 404 :)
SmokeMachine Zoffix, lizmat: what %hash{"exists", "does not exist" ; "exists", "does not exist"}:exists should return? (True, False, False) or (True, False, False, False)? I mean: once I found that the first level doesn't exists, should I return False and finish, or should I return 1 False for each key on the next levels? 14:01
Zoffix Wouldn't that be (False, False)? 14:07
m: my %h = a => { c => 42 }; b => { c => 42 }; dd %h{"a", "b"; "b", "b"}
camelia WARNINGS for <tmp>:
Useless use of "b => { c => 42 }" in sink context (line 1)
(Any, Any, Any, Any)
Zoffix m: my %h = a => { c => 42 }, b => { c => 42 }; dd %h{"a", "b"; "b", "b"}
camelia (Any, Any, Any, Any)
Zoffix Oh
I already forgot how these are parsed :) 14:08
m: my %h = a => { c => 42 }, b => { c => 42 }; dd %h{"a", "a"; "c", "c"}
camelia (42, 42, 42, 42)
Zoffix m: my %h = a => { c => 42 }, b => { c => 42 }; dd %h{"a", "a"; "c", "b"}
camelia (42, Any, 42, Any)
Zoffix m: my %h = a => { c => 42 }, b => { c => 42 }; dd %h{"a", "a"; "b", "b"}
camelia (Any, Any, Any, Any)
Zoffix Oh, ok.
SmokeMachine: so I'd think it'd be (False, False, False, False) 14:09
SmokeMachine iy would try: %a{"exists", "exists"}, %a{"exists", "do not exists"}, %a{"do notexists", "exists"}, %a{"do not exists", "do not exists"}
s/,/;/g
Zoffix Basically mirror the output of normal lookup.
m: my %h = 42 ZR=> <a b c d e>; dd %h<a b c z>:exists
camelia (Bool::True, Bool::False, Bool::False, Bool::False)
Zoffix huh 14:10
m: my %h = 42 ZR=> <a b c d e>; dd %h<a b c z>
camelia (42, Any, Any, Any)
Zoffix m: my %h = 42 XR=> <a b c d e>; dd %h<a b c z>:exists
camelia (Bool::True, Bool::True, Bool::True, Bool::False)
Zoffix Right
lizmat Zoffix: Wendy and I were at a Meet & Greet with Jon Anderson, Rick Wakeman and Trevor Rubin. We each gave them a Camelia. Trevor apparently put his on his monitor. The rest is history 14:14
timotimo cool
lizmat Zoffix: it's a public picture, you can tweet it :)
SmokeMachine Zoffix: another thing is this:
lizmat Rabin
argh
lizmat is not awake yet 14:15
SmokeMachine m: my %a{List} = (1, 2) => 42; my $l = List.new: 1,2; say %a{$l} # should it work?
camelia (Any)
Zoffix m: my $l = (1, 2); my %a{List} = $l => 42; say %a{$l} 14:17
camelia 42
Zoffix m: my %a{List} = (1, 2) => 42; my $l = (1, 2); dd %a.keys[0].WHICH; dd $l.WHICH 14:19
camelia ObjAt.new("List|69236784")
ObjAt.new("List|69237696")
Zoffix SmokeMachine: don't think so, 'cause it uses .WHICH doesn't it?
And when it is the same object, the lookup works fine.
lizmat SmokeMachine: it should mirror the datastructure returned without :exists
is what I think :-) 14:20
Zoffix m: my %a{List} = [1, 2] => 42; my $l = [1, 2]; say %a{$l}
camelia (Any)
Zoffix m: my $l = [1, 2]; my %a{List} = $l => 42; say %a{$l}
camelia 42
SmokeMachine m: my $l = [1, 2]; my %a{List} = $l => 42; say %a{1;2} 14:21
camelia Type check failed in binding to 'key'; expected List but got Int (1)
in block <unit> at <tmp> line 1
lizmat afk again&
Zoffix awww dammit 14:31
my awesomely twittable example is hitting a bug
Or is it...
m: gist.github.com/zoffixznet/0823179...a08a1ca914
camelia [["Make", "R::I::SET_LINE_ENDING_ON_HANDLE", "4.1x", "Faster"], Mu, Mu]
[["Make", "IO::Spec::Unix.catdir", "3.9x", "Faster"], Mu, Mu]
[["Make", "IO::Spec::Unix.split", "36x", "Faster"], Mu, Mu]
[["Make", "IO::Spec::Unix.rel2abs", "35\%", "faster"],ā€¦
Zoffix Why are those Mus there? Why are $type and $speedup aren't set? 14:32
Making them with: my ($sha, $, $, *@rest ($, $type, $speedup, $)) := \ |.words;
Works fine if I make it a sub's signature, so I guess it is indeed a bug :( 14:33
m: sub ($sha, $, $, *@rest ($, $type, $speedup, $)) {dd [ @rest, $type, $speedup ];}(|'fa9aa47 [io grant] Make R::I::SET_LINE_ENDING_ON_HANDLE 4.1x Faster'.words) 14:34
camelia [["Make", "R::I::SET_LINE_ENDING_ON_HANDLE", "4.1x", "Faster"], "R::I::SET_LINE_ENDING_ON_HANDLE", "4.1x"]
Zoffix m: my ($sha, $, $, *@rest ($, $type, $speedup, $)) := \ |'fa9aa47 [io grant] Make R::I::SET_LINE_ENDING_ON_HANDLE 4.1x Faster'.words; dd [ @rest, $type, $speedup ];
camelia [["Make", "R::I::SET_LINE_ENDING_ON_HANDLE", "4.1x", "Faster"], Mu, Mu]
Zoffix Rakudobugged: rt.perl.org/Ticket/Display.html?id=131071 14:38
And another issue is hitting another example :/ 14:51
Perl 6 isn't lovin' me today
m: my %h .= push: a => 42, a => 72; dd %h; 14:52
camelia Hash %h = {}
Zoffix why is it empty? 14:53
m: my %h; %h .= push: a => 42, a => 72; dd %h;
camelia Hash %h = {}
Zoffix m: my %h; %h .= push: $_ for a => 42, a => 72; dd %h;
camelia Hash %h = {:a($[42, 72])}
Zoffix Ah 14:56
m: my %h .= append: "a" => 42, "a" => 72; dd %h;
camelia Hash %h = {:a($[42, 72])}
Zoffix But still doesn't work in this form 14:57
m: gist.github.com/zoffixznet/92d0cf7...8f699dd91b
camelia Hash %h = {:Bar($["1.2", "4.2", "10.2"]), :Foo($["4.1", "1.3"]), :Meow("1.3")}
Zoffix ...
I guess my perl6 is too old
And I bet it's lizmat++'s fix to the %h = %h, %blah that fixed it 14:58
And I can't tweet this example either 'cause the fix is in 2017.03 but everyone's on 2017.01 Rakudo Star :P 15:00
Unlucky day for me
Zoffix tweets it anyway
u: āØ 15:27
unicodable6 Zoffix, U+2368 APL FUNCTIONAL SYMBOL TILDE DIAERESIS [So] (āØ)
Zoffix That gotta be made into an op :P
[Coke] I vote we call it "the moustache" 15:28
Zoffix I'm working on my March grant report ATM. Will publish it in ~4 hours 15:30
[Coke]: how do you publish it on TPF's site? Do you just scrape the HTML page? 'cause I got all of these reports in Markdown format, if that makes things easier 15:37
[Coke] hokay.
I just scrape the HTML. I am not sure if I can use the markdown, but I'm happy to try
Zoffix OK 15:38
dogbert17 .tell AlexDaniel if possible could you add gist.github.com/dogbert17/3eb15ea5...1d4966f7b2 to RT #131003 16:03
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=131003
yoleaux2 dogbert17: I'll pass your message to AlexDaniel.
nine
.oO(Will UNIX conventions still be in use in 100 years?)
 16:25
nine read a bit of the backlog
Zoffix Perl 6 IO TPF Grant: Monthly Report (March, 2017): blogs.perl.org/users/zoffix_znet/20...-2017.html 16:27
[Coke] Zoffix++ 16:34
Zoffix
.oO( s/Seqqy/Sequential/ )
 17:01
TimToady
.oO( class HyperSeq does Things )
 17:03
Zoffix This reminds me what I wanted to ask TimToady
timotimo my $dog will WORK(:4food) 17:04
TimToady sips coffee and hopes it will help
timotimo Zoffix: report looks good! i haven't read the previous one, though :(
Zoffix m: Seq.new: class :: does Iterator { method pull-one {}; method sink-all { say "sank" } }.new
camelia sank
Zoffix m: my $x = Seq.new: class :: does Iterator { method pull-one {}; method sink-all { say "sank" } }.new; $x
camelia WARNINGS for <tmp>:
Useless use of $x in sink context (line 1)
Zoffix m: my $x = Seq.new: class :: does Iterator { method pull-one {}; method sink-all { say "sank" } }.new; sink $x
camelia WARNINGS for <tmp>:
Useless use of $x in sink context (line 1)
sank
Zoffix m: my $x = Seq.new: class :: does Iterator { method pull-one {}; method sink-all { say "sank" } }.new; $x.sink 17:05
camelia sank
jnthn Well, as previously mentioned, Sequential isn't the best name for something that might work in parallel and produces results out of sequence. :-)
TimToady suggested Batchable earlier, if you consider Seq to be a single batch
jnthn Hmm
I wonder whether HyperSeq and RaceSeq should even have Seq in their name also :) 17:06
Zoffix TimToady: ^ is that by design that a class in a variable doesn't get sunk? Makes sink-all kinda not as useful. I first noticed that with IO::Handle's iterators where such a call on sinkage would cause the handle's closing....
TimToady but I was semi-serious with "Things", in the sense of "ThingsToDoNotNecessarilyInOrder"
jnthn haha...does Things :P
But yeah, it kinda works I guess :) 17:07
It's better than Seqqy :P
TimToady does WorkPolicy maybe
Zoffix jnthn: in that case clearly the name should be uenSeqtial
TimToady does Consequential :) 17:08
Zoffix "Seque: A Programming Language for Manipulating Sequences"
From www.dictionary.com/browse/seque
TimToady maybe they should be named Don'tCheat, CheatButGetAwayWithIt, CheatAndDon'tCare 17:09
.oO(Cordia, Concordia, Discordia)
 17:11
Zoffix is done skimming through gist.github.com/jnthn/6a80a9712fb3...0e46fca6d7 17:13
Over my head :) I think poor name suggestions is all I got :)
perlpilot Zoffix++ (just read your report) 17:16
TimToady
.oO( class HyperSeq does WhatNow )
 17:17
.oO( class HyperSeq does HowNow )
 17:18
.oO( class BrownCow does HowNow )
TimToady should be forbidden from free-association this early in the day...
.oO(we could redefine "Parallax" to mean how far you relax things in the parallel direction...)
 17:21
Zoffix :o
TimToady
.oO(and justify it because it tells you how far away the other end of the work is :)
 17:22
TimToady goes back to backlogging hoping his brane follows along after 17:23
TimToady 's brane is still far too gleeful about the antisociality of bending yet another existing word into a perletzel 17:28
Zoffix: re irclog.perlgeek.de/perl6-dev/2017-...i_14324664 I would say that callwith/nextwith work like 'next' in the middle of a loop (the dispatcher's loop, as it were, so continues the current dispatch cascade), while samewith is a complete redispatch, more like restarting the entire loop with different arguments, as I understand it 17:46
m: multi fact(0) { 1 }; multi fact($n) { $n * samewith $n - 1 }; say fact(42) 17:49
camelia 1405006117752879898543142606244511569936384000000000
TimToady m: multi fact(0) { 1 }; multi fact($n) { $n * callwith $n - 1 }; say fact(42)
camelia Use of Nil in numeric context
in sub fact at <tmp> line 1
0
TimToady samewith gets back to the 0 case successfully due to the redispatch
callwith/nextwith cannot get back to the 0 case because it's already been rejected in the current dispatch 17:50
Zoffix Then what's the difference between callwith/nextwith? 17:51
TimToady nextwith doesn't come back to this function, like a tail call
Zoffix oh 17:52
TimToady makes no difference here, since it's the last statement
jnthn nextwith is like return callwith
TimToady but hopefully optimized
Zoffix I'm now confused about which ones I was confused about :)
TimToady m: multi fact(0) { 1 }; multi fact($n) { samewith($n - 1) * $n }; say fact(42) 17:54
camelia 1405006117752879898543142606244511569936384000000000
TimToady samewith is not a tail call, but could be considered a strong hint to do TCO if it's last
nextwith is a forced tail call, but only of the rest of the current dispatch 17:55
we don't really have a way to force a tail-call redispatch yet
Zoffix [callwith] with new args, call from scratch and return here;
[callsame] with same args, call from scratch and return here;
[samewith] with new args, use the same dispatch chain to make a call and return here;
[nextwith] with new args, use the same dispatch chain to make a call and don't return here; 17:56
[nextsame] with same args, use the same dispatch chain to make a call and don't return here;
I think I got it :)
TimToady uh no
Zoffix bah :D
TimToady call* reuse the current dispatch chain
samewith creates a new dispatch chain 17:57
"reuse" in the sense of using the rest of the current chain, not in the sense of a restart
or equivalently, samewith can reuse the current chain but only by starting again from the front, which is probably what really happens 17:58
but I think of it as a fresh dispatch
call* and next* can only complete the current dispatch 17:59
Zoffix OK, after playing with some code I think I got it.
I need to write a blog post about it :)
Geth rakudo/nom: 19eeddbf35 | (Elizabeth Mattijsen)++ | t/spectest.data
Mark integration/eval-and-threads.t as slow

So it won't affect the total wallclock time as much as it did before.		 18:02
Zoffix m: foo (Any [$x]) {dd $x}([42]) 18:04
camelia 5===SORRY!5=== Error while compiling <tmp>
Missing infix inside []
at <tmp>:1
------> 3foo (Any [7ā5$x]) {dd $x}([42])
expecting any of:
bracketed infix
infix
infix stopper
Zoffix It complains about the sig, right?
m: multi foo (Int $x) { say "Int [$x]"; foo Any }; multi foo(Cool $x) { say "Cool [$x]"; nextsame; say "were back!" }; multi foo(Any [$x]) { say "Any [$x]" }; foo "x"
camelia Cool [x]
Zoffix But not in this case...
TimToady your first one was just a syntax error, missing 'sub' 18:07
Zoffix Ah. OK. Thanks. 18:09
TimToady m: sub foo (Any [$x]) {dd $x}([42])
camelia Int $x = 42
Zoffix Now I got it: *with = use these new args; *same = just use the same args; call* = call next candidate and return; next* = call next candidate and don't return; same* = restart from scratch
We're clearly missing samesame :)
TimToady yes, we clearly are :P 18:10
loop {} works better for that though
[Coke] there's more than same way to same it. 18:11
TimToady same old, same new...
lizmat: re irclog.perlgeek.de/perl6-dev/2017-...i_14325105 but then what about phasers that might return their value, or might also just be used for a side effect? 18:15
m: say { 42; ENTER say "bye" }()
camelia WARNINGS for <tmp>:
Useless use of constant integer 42 in sink context (line 1)
bye
True
TimToady m: say { 42; ENTER "bye" }()
camelia WARNINGS for <tmp>:
Useless use of constant string "bye" in sink context (line 1)
Useless use of constant integer 42 in sink context (line 1)
bye
TimToady huh, I guess that's a useless use of useless use, at least for "bye" 18:16
TimToady found hisself a buggy! 18:17
m: say { INIT "bye" }() 18:18
camelia WARNINGS for <tmp>:
Useless use of constant string "bye" in sink context (line 1)
bye
TimToady m: say { BEGIN "And yet I live!" }() 18:19
camelia And yet I live!
TimToady m: say { CHECK "And yet I live!" }()
camelia And yet I live!
TimToady m: say { INIT "And yet I live!" }()
camelia WARNINGS for <tmp>:
Useless use of constant string "And yet I live!" in sink context (line 1)
And yet I live!
TimToady curious
m: say { INIT now }() 18:20
camelia Instant:1490725251.568083
TimToady I guess we'd've discovered this bug earlier if we carped about 'now' in sink context
m: now; now; now; 18:21
camelia ( no output )
TimToady m: pi; pi; pi;
camelia WARNINGS for <tmp>:
Useless use of constant value pi in sink context (lines 1, 1, 1)
Zoffix m: rand; rand; rand;
camelia ( no output )
TimToady well, obviously 'now' updates the clock, just as 'rand' updates the RNG :P
if nobody ever asked "What time is it?", time would just quit... 18:23
TimToady wonders if someone can construct such a clock from trees falling in the forest that no one hears... 18:25
Zoffix Yes, but no one would hear when its alarm goes off... 18:27
lizmat m: my $a; { $a = LEAVE 42 }; say $a # TimToady: clearly, not all phasers are created equal 18:33
camelia WARNINGS for <tmp>:
Useless use of constant integer 42 in sink context (line 1)
(Any)
lizmat m: my $a; { $a = ENTER 42 }; say $a 18:34
camelia WARNINGS for <tmp>:
Useless use of constant integer 42 in sink context (line 1)
42
TimToady m: my $a; { $a := LEAVE 42 }; say $a
camelia WARNINGS for <tmp>:
Useless use of constant integer 42 in sink context (line 1)
Nil
TimToady but yes, clearly some phasers are better at time travel than others :) 18:35
lizmat m: my $a; for ^1 { $a = FIRST 42 }; say $a # indeed 18:36
camelia WARNINGS for <tmp>:
Useless use of constant integer 42 in sink context (line 1)
(Any)
TimToady that one just looks like a bug to me, assuming FIRST is an ENTERoid 18:39
m: for ^1 { say "HERE"; FIRST say "THERE" }
camelia THERE
HERE
TimToady which it is
lizmat back to schlepping& 18:40
Zoffix Hm. Quitting drinking has sure dulled my writing skills... 19:01
I guess I can find solace in my sharpened programming skills...
samcv hello everyone 19:20
Zoffix \o
m: multi foo (Int $x) { say "Int [$x]"; $x.flip }; multi foo (Cool $x) { say "Cool [$x]"; my $v = callwith Mu; say "We're back! The return value is $v"; $x.flip }; multi foo (Any $x) { say "Any [$x]"; $x.flip }; foo "I ā™„ Perl 6!"; 19:27
camelia Cool [I ā™„ Perl 6!]
Type check failed in binding to '$x'; expected Any but got Mu (Mu)
in sub foo at <tmp> line 1
in sub foo at <tmp> line 1
in block <unit> at <tmp> line 1
Zoffix Is it a bug that it crashes?
timotimo there's no candidate that would accept Mu, why would you expect callwith Mu to give anything but an exception? 19:28
Zoffix m: multi foo (Int $x) { say "Int [$x]"; $x.flip }; multi foo (Cool $x) { say "Cool [$x]"; my $v = callwith Mu; say "We're back! The return value is $v"; $x.flip }; multi foo (Date $x) { say "Any [$x]"; $x.flip }; foo "I ā™„ Perl 6!"; 19:30
camelia Cool [I ā™„ Perl 6!]
Use of uninitialized value $v of type Any in string context.
Methods .^name, .perl, .gist, or .say can be used to stringify it to something meaningful.
in sub foo at <tmp> line 1
We're back! The return value is
jnthn If you use callwith, it's on you to ensure that the argument types are acceptable
Zoffix But there's no ^ candidate there now and it doesn't crash.
jnthn Sure, then there's isn't a candidate
So there's nothing to fail to bind 19:31
Zoffix So it just tries the *next* next not the next matching?
jnthn Yes
Zoffix Oh
jnthn You are iterating a predestined (based on original arguments) set of candidates
callwith and nextwith don't cause that candidate set to be reevaluated.
TimToady (that's what samewith is for) 19:32
jnthn Indeed.
TimToady nextwith is useful only when you know you have a more generic candidate later in the list 19:33
jnthn Heh, probably "predetermined" woulda been a better word :)
TimToady I often have determination to get to my destination. 19:34
Zoffix m: class A is Int {}; class B is A {}; multi foo (A) { say "A"; nextsame }; multi foo (B) { say "B"; }; multi foo (Int) { say "Int" }; foo A; 19:36
camelia A
Int
Zoffix How come that didn't go to "B"? :/
jnthn Because B is narrower than A 19:37
Zoffix Oh doh
m: multi foo (IntStr) { say "IntStr"; nextwith "x" }; multi foo (Int) { say "Int"; }; multi foo (Cool) { say "Cool" }; foo <42>; 19:38
camelia IntStr
Type check failed in binding to '<anon>'; expected Int but got Str ("x")
in sub foo at <tmp> line 1
in sub foo at <tmp> line 1
in block <unit> at <tmp> line 1
Zoffix OK. Now I get it :D
jnthn++ thanks
samcv Zoffix, do you find HARNESS6 faster than harness5? 19:41
i have not timed them. but am curious
Zoffix samcv: nope 19:46
Don't see my timings for HARNESS6 in the logs, but it's a lot slower and seems to be missing some tests. 19:48
I think it's like 300-400 seconds for the stresstest, while the stresstest on harness5 runs in ~120 seconds
That's on 24-core box with the rakudo's max threads var bumped to something higher than normal. 19:49
samcv what's the normal max thread var? 19:53
jnthn Bumping Rakudo's max threads higher will probably only hurt performance for stresstest6 19:56
samcv: It defaults to 16
Thing is, the thread scheduler is none too smart so will quickly end up with the maximum number of threads
But harness6 doesn't really need that many 19:57
Even if your TEST_JOBS is far higher
timotimo i have here a test in the spec tests that has been eating a full cpu core for many minutes now 19:58
t/spec/S19-command-line/repl.rakudo.moar (Wstat: 256 Tests: 13 Failed: 1) 19:59
Failed tests: 10
TODO passed: 7
Zoffix jnthn: doesn't TESTJOBS affect how many tests it runs at the same time? So if max threads is 16 and your TESTJOBS is 30, how can it do all of them? 20:06
timotimo, does that test hang normally? I noticed Proc::Async.kill that doesn't-hang uses sometimes doesn't kill 20:07
timotimo yeah, when i run it i get a test failure and then it hangs 20:08
ok 9 - variables persist across multiple lines of input - its subtest "stdout is correct" fails, the "stderr is correct" succeeds, and then it hangs
when run without a harness, that is
Zoffix I'll take a look in ~1hr 20:09
[Coke] multi-threaded tests don't play nice with a high TEST_JOBS, right. 20:10
jnthn Zoffix: It uses Proc::Async, so it's non-blocking
Zoffix: Tests are run as subproccesses 20:11
Zoffix Ah. OK.
jnthn Zoffix: So one thread in the harness can service a bunch of running spectests
Zoffix cool
jnthn bbl
Geth rakudo/nom: 27f546907a | TimToady++ | src/Perl6/Actions.nqp
suppress useless use on INIT/ENTER/FIRST

Would be nice to determine this in 2nd pass, but for now, it's better to allow false negatives than false positives.		 20:12
rakudo/nom: 41ac4b4a3d | TimToady++ | src/Perl6/Actions.nqp
Catch useless use of 'now' now, now...
Zoffix timotimo: does it hand when you type the code manually in repl? It runs string "my \$x = 42;\nsay qq/The value is \$x/;\n", 20:34
remove backslashes and \n s are separate lines
Well, I lied. I'm too tired to look at it today, so maybe fudge it for now? 20:35
timotimo doesn't hang 20:36
Zoffix is off for rest of night
timotimo gnite Zoffix :) 20:37

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!