»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'p6: say 3;' or rakudo:, std:, or /msg camelia p6: ... | irclog: irc.perl6.org | UTF-8 is our friend!
Set by masak on 12 May 2015.
raiph m: sub foo(Int $x) { say $x; }; my Str $f = "Hi!"; foo $f; 00:03
camelia rakudo-moar a32c14: OUTPUT«5===SORRY!5=== Error while compiling /tmp/y6gFxJE7Ac␤Calling foo(Str) will never work with declared signature (Int $x)␤at /tmp/y6gFxJE7Ac:1␤------> 3Int $x) { say $x; }; my Str $f = "Hi!"; 7⏏5foo $f;␤»
raiph Sgeo: ^^
Sgeo Ah 00:04
raiph m: sub foo(Int $x) { say $x; }; foo "bar"; 00:05
camelia rakudo-moar a32c14: OUTPUT«5===SORRY!5=== Error while compiling /tmp/ugkiamRWwN␤Calling foo(str) will never work with declared signature (Int $x)␤at /tmp/ugkiamRWwN:1␤------> 3sub foo(Int $x) { say $x; }; 7⏏5foo "bar";␤»
00:09 Peter_R joined 00:34 jjido left, shinobi-cl left
japhb m: my Str $f = 12; 00:44
camelia rakudo-moar a32c14: OUTPUT«Type check failed in assignment to '$f'; expected 'Str' but got 'Int'␤ in block <unit> at /tmp/nX6o8GshRQ:1␤␤»
japhb m: say "Boo!"; my Str $f = 12;
camelia rakudo-moar a32c14: OUTPUT«Boo!␤Type check failed in assignment to '$f'; expected 'Str' but got 'Int'␤ in block <unit> at /tmp/4enJ6sMDI7:1␤␤»
japhb Oh interesting, I thought we had that one at compile time already.
m: say "Boo!"; my Str $f := 12; 00:45
camelia rakudo-moar a32c14: OUTPUT«Boo!␤Type check failed in binding; expected 'Str' but got 'Int'␤ in block <unit> at /tmp/AXkyhR1m8b:1␤␤»
japhb m: say "Boo!"; my Str $f ::= 12;
camelia rakudo-moar a32c14: OUTPUT«Boo!␤Type check failed in binding; expected 'Str' but got 'Int'␤ in block <unit> at /tmp/_5Apymdzvo:1␤␤»
japhb Huh
Well, after combining a few minutes a week for several weeks, I can happily report: Experiment at putting "plain OO" sugar on top of CQS + event sourcing: SUCCESS. 01:03
diakopter Run command wrote to STDERR: install/bin/perl6 --optimize=3 -e 0 01:06
Compiler rakudo-moar is failing for test zero, continuing to next compiler/test.
japhb diakopter: If you cd into the build tree for rakudo-moar/nom (or rakudo-moar/glr or whatever), and run that exact command, what happens? 01:10
diakopter well I gotta fixup my path in a new terminal
japhb No! 01:11
just open a new terminal, cd to .../perl6-bench/components/rakudo-moar/nom/ and run the command it printed.
('install/bin/perl6 --optimize=3 -e 0')
diakopter oh 01:13
WARNINGS:
Useless use of constant integer 0 in sink context (line 1)
"Useless use"; I love it 01:14
japhb So yeah, not a whole lot I can do about that. 01:16
Actually: If we still agree that -e should disable strict, shouldn't it disable warnings and deprecations as well?
geekosaur that's an infamous perl5 warning too :) 01:21
(well, void context instead of sink, but "useless use...")
ShimmerFairy There's no warnings pragama implemented yet, afaik, so there's nothing for it to disable :) 01:22
deprecations would count under warnings. I always thought that perhaps the warning pragma took options, like 'no warnings :deprecations' for instance
01:35 njmurphy joined 01:43 njmurphy left 01:46 leont left, aborazmeh joined, aborazmeh left, aborazmeh joined 01:53 llfourn joined 01:55 colomon left 01:56 colomon joined 01:57 llfourn left 02:00 colomon left 02:09 mr-foobar left 02:12 colomon joined, gfldex left 02:17 colomon left, colomon joined 02:18 mr-foobar joined, cognominal joined 02:19 ggoebel joined, BenGoldberg joined
skids m: sub foo (Any:D $x) { say $x; }; my Failure $f = Failure.new; foo($f); # Is actually a successful call (Failure.new is DEFINITE), it is the "say $x" that explodes. 02:21
camelia rakudo-moar a32c14: OUTPUT«Failed␤␤Actually thrown at:␤ in sub foo at /tmp/0Zhjo1PNgf:1␤ in block <unit> at /tmp/0Zhjo1PNgf:1␤␤»
skids Sgeo: ^^
Sgeo m: sub foo (none(Failure) $x) { say $x; }; my Failure $f = Failure.new; foo($f); 02:22
camelia rakudo-moar a32c14: OUTPUT«5===SORRY!5=== Error while compiling /tmp/cHkW23YfPe␤Invalid typename 'none' in parameter declaration.␤at /tmp/cHkW23YfPe:1␤------> 3sub foo (none7⏏5(Failure) $x) { say $x; }; my Failure $f␤»
Sgeo I thought junctions were allowed as types?
TimToady no, use a where
Sgeo Does where use implicit ~~ or just call a closure? 02:23
m: sub foo (Any:D $x where {$_ !~~ Failure}) { say $x; }; my Failure $f = Failure.new; foo($f);
camelia rakudo-moar a32c14: OUTPUT«Constraint type check failed for parameter '$x'␤ in sub foo at /tmp/05SdFsxKXe:1␤ in block <unit> at /tmp/05SdFsxKXe:1␤␤»
skids m: sub foo (Any:D $x) { say $x; }; my Failure $f = Failure.new; $f.defined; foo($f); # Also get this to work first before playing with the more explody variety.
camelia rakudo-moar a32c14: OUTPUT«Failed␤␤»
TimToady it's a smartmatch 02:24
Sgeo So what's the more concise way to write what I wrote?
m: sub foo (Any:D $x where none(Failure)) { say $x; }; my Failure $f = Failure.new; foo($f);
camelia rakudo-moar a32c14: OUTPUT«Constraint type check failed for parameter '$x'␤ in sub foo at /tmp/nqIXvgXbMu:1␤ in block <unit> at /tmp/nqIXvgXbMu:1␤␤»
02:34 cognominal left
skids Well, you could avoid DRY with: 02:37
m: subset Pedantic of Any where { $_ !~~ Failure and $_.defined }; sub foo(Pedantic $x) { 42.say }; my Failure $f = Failure.new; foo($f);
camelia rakudo-moar a32c14: OUTPUT«Constraint type check failed for parameter '$x'␤ in sub foo at /tmp/a0ZBaw7QQB:1␤ in block <unit> at /tmp/a0ZBaw7QQB:1␤␤»
02:38 cognominal joined 02:40 noganex_ joined 02:42 noganex left
skids Or if you prefer to see the actual failures instead of typecheck failures: 02:44
m: subset Pedantic of Any where { ($_ ~~ Failure and $_.throw) and $_.defined }; sub foo(Pedantic $x) { 42.say }; my Failure $f = Failure.new; foo($f);
camelia rakudo-moar a32c14: OUTPUT«Failed␤␤Actually thrown at:␤ in any at src/gen/m-Metamodel.nqp:2867␤ in block <unit> at /tmp/i7QDD9UxyC:1␤␤»
02:53 mjgardner joined 02:56 cognominal left 03:12 nys left 03:25 llfourn joined 03:29 mjgardner left 03:32 mr-foobar left 03:33 llfourn left 03:40 mr-foobar joined
dalek ast: d54ea63 | skids++ | S09-hashes/objecthash.t:
Add tests for RT#118031 which might be closeable.
 04:11
Sgeo Subsets just put me in mind of pattern matching, is there a way for a subset to return a value to be bound into a variable at use? 04:13
skids What, like an identity value? 04:18
raiph m: subset PositiveInt of Int where +*; my PositiveInt $foo = 42; say $foo 04:19
camelia rakudo-moar a32c14: OUTPUT«42␤»
raiph m: subset PositiveInt of Int where +*; my PositiveInt $foo = -1; say $foo 04:20
camelia rakudo-moar a32c14: OUTPUT«-1␤»
skids m: say so +(-1) 04:21
camelia rakudo-moar a32c14: OUTPUT«True␤»
raiph m: subset PositiveInt of Int where * > 0; my PositiveInt $foo = 0; say $foo
camelia rakudo-moar a32c14: OUTPUT«Type check failed in assignment to '$foo'; expected 'PositiveInt' but got 'Int'␤ in block <unit> at /tmp/SzsYPZ3q7S:1␤␤»
raiph Sgeo: I don't think I understood your comment... 04:22
04:23 llfourn joined
Sgeo Meh I'm overreaching I think 04:24
Although... in a given/when, can the thing given to when return a value that could then be used by the block?
I'm assuming not since that's not how ~~ works
Can ACCEPTS return a non-boolean, and what, exactly, happens if it does? 04:26
raiph m: subset PositiveInt of Int where * > 0; given 0 { say "nay" when not PositiveInt }; given 1 { say "yay" when PositiveInt };
camelia rakudo-moar a32c14: OUTPUT«nay␤yay␤»
raiph m: subset PositiveInt of Int where * > 0; given 1 { say "nay" when not PositiveInt }; given 0 { say "yay" when PositiveInt }; 04:27
camelia rakudo-moar a32c14: OUTPUT«nay␤»
Sgeo m: subset PositiveInt of Int where * > 0; say PositiveInt.WHAT 04:28
camelia rakudo-moar a32c14: OUTPUT«(PositiveInt)␤»
Sgeo So you can smartmatch on subsets?
raiph m: subset PositiveInt of Int where * > 0; for -1,0,1 { say $_, $_ ~~ PositiveInt } 04:31
camelia rakudo-moar a32c14: OUTPUT«-1False␤0False␤1True␤»
raiph subsets are just a (sub) type 04:33
m: subset PositiveInt of Int where * > 0; given 1 { say "nay" when none(PositiveInt) } 04:40
camelia ( no output )
raiph goodnight #perl6 04:44
04:44 raiph left
skids Well, there's always: 04:45
m: given 1 { when $f = Int { $_.say; $f.say } }
camelia rakudo-moar a32c14: OUTPUT«5===SORRY!5=== Error while compiling /tmp/SZfk7XX7k6␤Variable '$f' is not declared␤at /tmp/SZfk7XX7k6:1␤------> 3given 1 { when 7⏏5$f = Int { $_.say; $f.say } }␤»
skids hrm 04:46
oh.
m: given 1 { my $f; when $f = Int { $_.say; $f.say } } 04:47
camelia rakudo-moar a32c14: OUTPUT«1␤(Int)␤»
04:49 llfourn left 04:50 ChoHag joined
ChoHag Does a function with a signature still have access to the signature as a whole in a single variable? 04:50
04:52 rilly joined
ChoHag No. Does a function with a signature still have access to the *capture/argument list* as a whole in a single variable? 04:52
04:54 rilly is now known as rilly42
skids ChoHag: if you want that, use a capture in the sig: 04:54
04:54 xinming_ joined
skids m: my sub a (|c ($a, $b)) { c.perl.say }; a(1,2); 04:55
camelia rakudo-moar a32c14: OUTPUT«\(1, 2)␤»
04:55 aborazmeh left 04:57 xinming left
ChoHag Thanks that's got it. 04:58
Sgeo m: my sub a (|c ($a, $b)) { c.perl.say }; my $foo = 10; a(1,2, $foo); 05:23
camelia rakudo-moar a32c14: OUTPUT«Too many positionals passed; expected 2 arguments but got 3 in sub-signature of parameter c␤ in sub a at /tmp/AjGMD1zJ2v:1␤ in block <unit> at /tmp/AjGMD1zJ2v:1␤␤»
Sgeo m: my sub a (|c ($a, $b, $c)) { c.perl.say }; my $foo = 10; a(1,2, $foo);
camelia rakudo-moar a32c14: OUTPUT«\(1, 2, 10)␤»
05:23 llfourn joined
Sgeo m: my sub a (|c ($a, $b, $c is rw)) { c.perl.say }; my $foo = 10; a(1,2, $foo); 05:23
camelia rakudo-moar a32c14: OUTPUT«\(1, 2, 10)␤»
Sgeo oops, c and $c are different variables right? 05:24
05:28 kaare__ joined
ChoHag m: sub foo (Str @things) { @things.perl.say }; my Str @thongs = <a b c d>; foo(@thongs); 05:34
camelia rakudo-moar a32c14: OUTPUT«Array[Str].new("a", "b", "c", "d")␤»
ChoHag m: sub foo (Str @things) { @things.perl.say }; my Str @thongs = <a b c d>; foo(@thongs[2..*]);
camelia rakudo-moar a32c14: OUTPUT«Type check failed in binding @things; expected 'Positional[Str]' but got 'Parcel'␤ in sub foo at /tmp/Pt1jiBUXlx:1␤ in block <unit> at /tmp/Pt1jiBUXlx:1␤␤»
ChoHag Why can't I pass an array slice? Does the ongoing GLR affect this? 05:35
rilly42 m: sub foo (Str @things) { @things.perl.say }; my Str @thongs = <a b c d>; foo(@thongs[2..*] but Positional[Str]); 05:38
camelia rakudo-moar a32c14: OUTPUT«("c", "d")␤»
ChoHag So that answers question 3 - what's an effective workaround? 05:39
But is this really desired behaviour? 05:40
dalek ast: 00a9f8a | skids++ | S06-currying/misc.t:
Add test for RT#123938 which may be closeable
 05:43
skids Well ISTR Parcel is not long for this earth, so something will likely change there. 05:46
05:46 jjido joined 05:47 elohmrow joined 05:48 elohmrow left
ShimmerFairy Sgeo: from earlier, ~~ can return a non-boolean value; the most often instance of this is "string" ~~ /regex/ :) 05:49
Sgeo And what do when etc. do with this value? 05:50
ShimmerFairy coerce to boolean
05:50 elohmrow joined
Sgeo There's no way for when's block to access the value itself? 05:51
ShimmerFairy You can access the value passed to 'given' as $_ , but the result of the comparison isn't available AFAIK (and fwiw, I've never heard anyone need it inside 'when' blocks) 05:53
ChoHag Really? You wouldn't want the Match object inside the when? 05:56
05:56 cognominal joined
ChoHag Sgeo: Docs suggest there's a $/ -- regex match. Does that have what you want? 05:57
05:57 cognominal left
ChoHag doc.perl6.org/language/variables#Pr..._variables 05:57
Sgeo ChoHag, that's only useful for regexes though, what if I have my own thing that has an ACCEPTS that returns a useful non-bool?
ShimmerFairy ah yes, forgot about $/ . Latest regex match is put there 05:58
ChoHag But back on the "isn't available", it seems odd that it wouldn't be and odder to suggest it doesn't need to be.
06:01 aborazmeh joined, aborazmeh left, aborazmeh joined
ShimmerFairy I can't think of a place where implicit smartmatching is done and the result of that implicit smartmatch is put somewhere for the user to mess with. 06:02
Sgeo I can think of equivalents in other languages
awwaiid panda broken? :( . Just tried to upgrade via rakudobrew to moar-2015.07.2, and then rakudobrew build-panda, but panda is mad at me 06:04
ChoHag How do you break out of a loop?
ShimmerFairy off-hand I can only think of implicit smartmatching happening in when blocks and flipflop ops, and neither of those expose the result of that smartmatch anywhere. I should also point out that ACCEPTS most often does return a boolean, so most often you already know what the value is.
ChoHag: 'last' comes to mind to stop a loop and move on. There are also loop labels, but I'm not very familiar with them. 06:05
Sgeo So, in Haskell with pattern synonyms, this 06:09
pattern Head x <- x:xs
Defines a pattern that will match on a list with at least one element (that's what x:xs is, a list with x as the first element and xs as the rest) 06:10
So you can write a function like foo (Head x) and then in the body x, the head, will be available
jdv79 could you call perl6 and go opposites wrt to complexity at least in terms of modern langs?
or rather, the two extremes 06:11
Sgeo At least Perl 6 doesn't pretend to be statically typed in a useful way. I mean, er
jdv79 Sgeo: have you used go? 06:17
Sgeo I know a bit about it, haven't tried it myself
Let me know when it has generics >.>
jdv79 i just read and watched some conf talks on it... haven't got beyond toy examples yet.
06:24 jjido left
skids Sgeo: there seem to be some not-yet-implemented plans for given $capture { when * -> $sig, $sig2 { ... } } to defer binding tests to the when clause's block, which can be pointy/signatured. 06:24
Thus giving you access to any of the vars you decide to name 06:25
awwaiid hm. well I installed panda a few more times and now it works. 06:28
06:34 xinming_ left, skids left, bin_005 joined, bin_005 left 06:38 bin_005 joined
awwaiid mmmm nevermind. still broken. 06:40
jdv79 maybe try rebooting:) 06:41
awwaiid hehe
06:43 aborazmeh left, FROGGS joined
awwaiid jdv79: "rebooting" worked -- aka rakudobrew nuke 06:48
ChoHag The three R's are just as valid outside Windows. 06:50
Restart, Reboot, Reinstall.
07:04 rurban joined 07:11 xinming joined 07:16 rindolf joined 07:17 bin_005 left 07:18 bin_005 joined 07:19 darutoko joined 07:39 elohmrow left
jdv79 nice 07:44
i've resorted to rm -rf'ing and recloning all the time:( 07:45
well, i recently just keep pristine repos around and cp -r them instead since cloning can take a lot of time 07:46
well, not exactly but something like that 07:50
07:54 espadrine joined
ShimmerFairy I usually use git clean -dfx to clear a repo of any untracked files, and git reset --hard to undo any changes I've yet to commit 07:58
08:00 TEttinger left
timotimo in the (not so?) far future, impressionable newbies will be told to "run git -xdf in your / directory" rather than "rm -rf --no-preserve-root /" 08:05
jdv79 hmm, i didn't noticed the x option. i think that's why i sidelined git clean. nice. i'll try that next time. 08:08
08:20 brrt joined
dalek ast: 319fcbf | lizmat++ | S04-statements/with.t:
Remove with/without modifier tests

They are tested elsewhere		 08:23
08:27 llfourn left
brrt good morning 08:27
08:29 mr-foobar left
jdv79 good aft ^H 08:29
Sgeo (-> Any $x { @e.push: $x }).($j); 08:30
08:30 mr-foobar joined
Sgeo What is the . between the block and arglist fr? 08:30
for
brrt invocation, i think :-) 08:36
the block is the object, circumfix<()> the method
jdv79 iirc in special situations the . can be left out like {} [] () 08:39
lizmat m: say (Int,Str,"a").map: { next unless $_; $_ } # works
camelia rakudo-moar a32c14: OUTPUT«a␤»
lizmat m: say (Int,Str,"a").map: { next without $_; $_ } # does not work (yet)
camelia rakudo-moar a32c14: OUTPUT«5===SORRY!5=== Error while compiling /tmp/pIRikiEiyH␤Undeclared routine:␤ without used at line 1␤␤»
lizmat and I wonder whether it shouldn't be possible to reduce the latter to just "next without" :-) 08:40
08:40 haroldwu left
timotimo "try again, this time without $_" 08:41
lizmat also: maybe this should be fixed by just adding a sub "without" ?
m: sub without(\a) { !a.defined }; say (Int,Str,"a").map: { next without $_; $_ }
camelia rakudo-moar a32c14: OUTPUT«Cannot call next(Bool); none of these signatures match:␤ ()␤ (Label:D \x)␤ in block <unit> at /tmp/Lt25e46xHX:1␤␤»
08:42 haroldwu joined
lizmat hmmmm.... 08:42
perhaps not :-)
08:43 bin_005 left 08:45 bin_005 joined 08:48 BenGoldberg left
Sgeo m: (1..10).pick(*) 08:51
camelia ( no output )
Sgeo m: say (1..10).pick(*)
camelia rakudo-moar a32c14: OUTPUT«1 10 8 6 5 4 3 9 2 7␤»
Sgeo m: say (1..10).pick(*-1) 08:52
camelia rakudo-moar a32c14: OUTPUT«Cannot call pick(Int: Int, Int, Int, Int, Int, Int, Int, Int, Int, WhateverCode); none of these signatures match:␤ (Range $: *%_)␤ (Range $: Whatever, *%_)␤ (Range $: Cool $n, *%_)␤ in block <unit> at /tmp/43RR90lY07:1␤␤»
Sgeo Why do all of them have slurpy params?
m: say (1..10).pick("3") 08:53
camelia rakudo-moar a32c14: OUTPUT«9 8 10␤»
Sgeo Why is this ... why?
Because Perl 5 did it that way?
08:53 jkva joined
timotimo pick takes a Cool and calls .Int on it 08:54
coercive parameter
08:55 mr-foobar left
ShimmerFairy Sgeo: those *%_ are known as "implicit *%_", they appear on all methods so that unrecognized named params can be passed to other multi candidates upon use of nextsame/callsame. Because it somehow makes things more resilient? (I am not a fan of it.) 08:57
timotimo oh, that's what sgeo referred to
08:57 mr-foobar joined
Sgeo timotimo, the first question referred to what ShimmerFairy said, the second/third referred to what you answered 08:57
ShimmerFairy timotimo: Incidentally I just realized the halfway solution would be to introduce implicit *%_ on only multi methods, and exclude them from 'only' methods, however difficult that may be to implement :) (unless my understanding of *same is off and they can call the same name from inherited classes, which would be surprising to me) 08:59
09:02 gfldex joined, isBEKaml joined
timotimo thing is, you can introduce a method as a multi method in a derived class, but still want to "nextsame" 09:04
ShimmerFairy timotimo: on the subject of that nqp::while bug: when I copied the deepmap function from to play around with it, I found that adding a simple ($rpa[$i]), line after each loop's nqp::bindpos call makes it work. So apparently the loop needs to reference the list position it just bound, now? 09:08
timotimo ... wtf? :(
anyway, i'll be on a drive for about 7 hours in about an hour
oh god, i'm running a bit late
still need to finish packing
see ya!
ShimmerFairy timotimo: see ya! This is definitely a perplexing bug, hopefully I can figure more examples showing it off. 09:10
timotimo have you looked at the patch itself at all? maybe there's something very bogus i put in there and i was just too blind (or too up-close) to see it?
09:11 jkva left
ShimmerFairy I looked at git log -p for quite a while. I couldn't figure out any simple changes that affected anything. (The only simple thing I didn't change was putting back the thing that returns a value, because I get the feeling I'd need to undo the rest of the changes to make that work :/ ) 09:12
timotimo: my best guess is that the registers played with on MoarVM by nqp::while were relied on by some other parts of the VM-based machinery. Which is a very broad and unhelpful guess. 09:13
timotimo: anyway, have a good drive regardless! :) 09:14
09:15 jkva joined
timotimo i probably just ought to look at the spesh dump and the obvious problem will smack me in the face 09:17
ShimmerFairy if spesh is an optimizer thing (been a while since I heard it), I could easily see the lack of return value allowing the optimizer to accidentally remove something. 09:18
ShimmerFairy is briefly reminded of the 'volatile' keyword in C++ :)
timotimo wellllll, if the code-gen knows something wants a/the return value, it'll generate an object register and set it to null (i think?) 09:19
otherwise it'll set the resulting piece of code to have a VOID return value
you should have a look at spesh, it's cool. set MVM_SPESH_LOG=foo.txt and you'll get the before and after bytecode of frames in the spesh-generated SSA form 09:20
ShimmerFairy timotimo: you got rid of push_op(@loop_il, 'set', $res_reg, @comp_ops[0].result_reg); (which seems like it could be the source of issues, but I'm not sure if it works without paying attention to the requested object type like the before code), and some IMM_ARG stuff I assume is irrelevant now. 09:21
I'll be sure to look at spesh output, then :)
timotimo just in general :)
there's a tool in moarvm/tools called "graph_spesh" that'll generate a graphviz graph for a single frame you've copy-pasted in 09:22
ShimmerFairy timotimo: for the record, I could've just reverted the commit (and in a more production-driven setting, probably would have), but I'm just too interested in figuring out _what_ the problem is :P
timotimo i'm glad to hear you're not working on it because i've been begging for help, but because you're enjoying it, too :) 09:23
ShimmerFairy The fact that something so innocuous as » was broken is interesting, and after looking a bit into it, it's _really_ perplexing :) 09:25
timotimo but be careful: going too deep into the backendy things may lead you to become a moarvm dev :P
nine One does not simply walk into MoarVM. 09:26
ShimmerFairy m: ((97,), (102,))».chrs.say # welp, could be related to strings too then?
camelia rakudo-moar a32c14: OUTPUT«a f␤»
timotimo what if we used a catapult to bring the bytecode right into the underlying hardware?
m: ("a", "b")>>.ords.say 09:27
camelia rakudo-moar a32c14: OUTPUT« ␤»
timotimo m: ("abc", "bcd")>>.ords.say 09:28
camelia rakudo-moar a32c14: OUTPUT« ␤»
timotimo well, to be fair
if the thing ords returns is a bit different, it'll work out
have you seen my golfs yesterday? (or was that the day before?)
where i tried to emulate different kinds of .ords?
ShimmerFairy I'm having trouble getting it to find MAST::Ops :/
timotimo i had to specifically put in a (^5).map or something
ShimmerFairy I'm not sure I did.
timotimo you may have to run update_ops.p6? 09:29
the MAST::Ops.pm6 may be missing from the git repository or something
*mumble mumble*
09:29 rurban left
ShimmerFairy maybe, there _is_ an Ops.p6 file in there, which I guess doesn't count as a module file. 09:29
timotimo disappears for a bit 09:30
09:30 RabidGravy joined
Sgeo m: say Array.WHENCE 09:30
camelia rakudo-moar a32c14: OUTPUT«Method 'WHENCE' not found for invocant of class 'Array'␤ in block <unit> at /tmp/eph7ktf7FA:1␤␤»
09:31 rurban_ joined, rurban joined 09:36 isBEKaml left 09:38 cognominal joined 09:45 diana_olhovik_ joined 09:53 brrt left 10:00 jjido joined 10:01 jjido left 10:11 araujo left 10:14 araujo joined 10:23 BenGoldberg joined, llfourn joined 10:24 leont joined 10:25 araujo left 10:27 llfourn left 10:30 FROGGS left 10:32 araujo joined 10:34 telex left 10:35 llfourn joined 10:36 telex joined 10:37 rurban left 10:43 isBEKaml joined
timotimo WHENCE is something a scalar can have 10:44
... waiting for the transporter to arrive so we can put our stuff in (and i can drive it a part of the way) 10:45
jnthn afternoon, #perl6 o/
nwc10 good UGT heresy, jnthn
timotimo ohai jnthn 10:46
how are you on this fine sunday?
jnthn Not too bad
At least it's not as insanely hot today as it was yesterday... 10:47
10:48 bin_005_d joined 10:49 bin_005 left
nine It's not? 10:53
jnthn nine: Not here :) 10:54
Yesterday here hit 40C. Today it's not meant to get too far over 30C
isBEKaml jnthn: I always wonder why people say "It's not meant to be this hot" or "It's not meant to go over 30C today" esp. when talking about the capricious weather ;-) 10:56
nine 40C?! "insane" is a good word for that :) Where is this insane "here" if I may ask? 10:58
jnthn nine: Prague
nine: I...don't think this is particularly normal weather for here.
nine That's just 200km north east of here. 10:59
11:01 smls joined
nine Ok, having compiled a perl and installed hundreds of modules for the third time, I can conclude, that my refcounting error simply does not occur on a perl built with debugging options enabled 11:04
jnthn Does not occur, or is hidden by? 11:05
jnthn has a had a lot of incidences of "bug vanishes under the debugger", simply 'cus memory ends up organized differently
nine Probably the latter. Though ironically those debug options are there to actually make those bugs more visible. 11:06
jnthn Yeah, sadly that doesn't always work out as well as hoped. I've had SEGVs re-appear by disabling the debug malloc library that adds in some extra checks to catch bugs... 11:07
nine The good news is: the 17000 tests of our production codebase pass, even when I extend a central DBIX::Class resultset's class using Inline::Perl6 :) 11:08
With the debug options enabled, the somewhat reproducable "Attempt to free unreferenced scalar: SV 0x1d6051f0 during global destruction" messages seem to get replaced by extremely rare segfaults that vanish when run under gdb 11:09
smls jnthn: Post-GLR, does .[] still flatten its arguments? Do X and Z still flatten their arguments? 11:10
In pmichaud's design, those were outstanding questions I believe.
jnthn nine: The OSX debug malloc has a malloc_history command where you can feed it a memory address and it'll tell you the "history" of its allocations/frees. 11:11
smls Also, (1, 2, 4, 8 ... *)[10] will have to be written as (1, 2, 4, 8 ... *).list[10], right? (Just making sure I understand correctly.)
jnthn nine: Dunno if there's a similar feature on whatever platform you're devloping on (if it's not OSX... :)) 11:12
smls: The sequence one is easy to answer: yes, you'll need the .list there
nine jnthn: I'm bearded and long haired. I have to use Linux :)
jnthn smls: I didn't do .[] and X/Z in the work so far. I don't think I can see a reason for X and Z to .flat their arguments. 11:15
smls ok 11:16
jnthn smls: As for .[...] I'd be inclined to take the same route as with [...] and array assignment
smls right
jnthn .[@a] # slice based on things in @a, like today
.[1, 2, 3] # also as today
nine New debugging strategy: adding a warn $self in every place where I create an object that is going to be passed to Perl 6 code and use my terminal's Find to match the unreferenced scalar's address to constructed object's ones. Now I finally know _which_ object's refcount is wrong :) 11:17
smls though from a practical POV, it's less useful to befine a nested list as .[] arg only to have sublists interpreted as their lengths
jnthn .[$@a] # also as today
smls .[1, 2, 5..8, 10] # in this case, flattening would be more DWIMMy
jnthn It would, but we'll pay for it on every single slice. Do we want that? 11:18
smls Does flattening *have* to be expensive? 11:19
Isn't it supposed to be lazy?
jnthn How would laziness help here when we need all the values?
It's not that it's expensive per se, but it's not free either.
smls well, maybe lazy is not the right word her
I meant, doesn't have to really affect lists that are already flat 11:20
jnthn Well, .flat on an Array returns identity
uh
No, it doesn't
It returns a Seq with a normal iterator for the array
But it knows it doesn't have to do any work.
Aside from iterate through the things. 11:21
Anyway, if you can show me globs of real world code that do .[1, 2, 5..8, 10] I'll probably be more intrested.
I don't care much for "it'll DWIM on this hypothetical example"
smls ignoring the actual inner workings and only looking at the *effective* behavior, for @elems { when Positional { .length } ... } and for @elems { when Positional { #.[iterate sublist] } ... } would be expected to have the same performance on already-flat lists, no? 11:22
jnthn You're confusing Positional with Iterable 11:23
But sure, they would, except you actually have to do the loop
And the check
smls wouldn't you have to do it regardless? 11:26
In any case, I suppose having to write .[flat 1, 2, 5..8, 10] or .[1, 2, |(5..8), 10] wouldn't be so bad either.
jnthn Yes, that's what I'd prefer.
I just don't recall seeing things like this often enough in real world code to think it's worth pessimizing everything for. 11:27
jnthn notes we're fast converging on a design where the only "magical" flattening happens through slip, and everywhere else you ask for it (with taking a slurpy *@foo counts as asking for it) 11:34
smls On that note, what built-ins will "magically" use Slip? 11:36
Will .map({ .foo, .bar }) automatically return a Slip to make the result list flat? 11:37
dalek line-Perl5: db25ccf | (Stefan Seifert)++ | p5helper.c:
Fix one source of "Attempt to free unreferenced scalar during global destruction"

This was just left over code from before the inheritance re-design. No need to create a new reference to bless the object. We can just bless the reference we were passed.
The "unreferenced scalar" error came, because we used newRV_noinc to create the reference and then never increased or decreased the refcount afterwards. Without an SvREFCNT_dec, perl never gets a chance to free the SV and rightfully complains about that when it finds it during shutdown.
11:37 fxer joined
jnthn smls: Nothing magically makes a slip 11:37
smls ok
jnthn smls: They just magically interpolate it
smls: So map is aware that it should check values coming back for it 11:38
smls: Since it's type based, we'll also (later) be able to elide the check for map blocks that declare a return type. 11:39
.map(-> $x --> Int { ... }) # Slip impossible
smls neat
dalek kudo/glr: fc933f7 | jnthn++ | src/Perl6/Metamodel/BOOTSTRAP.nqp:
Also stub Slip in the BOOTSTRAP.
 11:45
kudo/glr: b8355fe | jnthn++ | src/core/core_prologue.pm:
Stub Iterable, set up Empty constant.
smls jnthn: Will .grep on a Seq return a new Seq, or a List that remembers all values?
11:47 bin_005_d left
smls "map/grep/etc., gather/take, and ... (and likely more) all return a Seq." ok, never mind then :P 11:47
jnthn smls: A new Seq
And if you .grep a HyperSeq you get back a HyperSeq 11:48
smls So that's a break from the current semantics of having map/grep always return the same type (currently Parcel)?
Will you extend that to things like native arrays so that .map/.grep will return a new native array? 11:49
11:52 Aurelie joined
jnthn .map and .grep will always return a Seq 11:53
smls btw, yay for "for loop over gather/take ==> was 5.86534547805786s, now 0.880507230758667s".
jnthn *unless* you already did .hyper() and opted into parallel processing, in which case you'll get a HyperSeq 11:54
smls ok
jnthn Basically, most operations (map, grep, etc.) take you into a "pipeline mode". .hyper and .race take you into a "parallel pipeline mode". 11:55
And you get out of it with an array assignment, coercion, or an operation that is some form of reduction. 11:56
.eager is the only thing I'm aware of that keeps you in pipeline mode, but also introduces intermediate memory of values 11:57
(And it only does that if it has to)
smls So .eager does not immediately iterate its invocant, but only marks it to be iterated eagerly once iterating starts? 11:58
jnthn Yeah 11:59
smls ok
jnthn I thought about whether that was right
The thing that kinda forced it was that I realized that making it that way made it possible to make "my @a = eager @b.map(...);" cost pretty much what it would without the eager if the eager was equivalent to just evaluating up to the first lazy thing. 12:00
That is, in that case I can avoid the intermediate buffer
smls right 12:01
Will there be a way to get the nth value of a sequence, without having to store all n-1 prior values in memory? 12:03
Thinking about things like constant @primes = grep, *.is-prime, 1..*; say "The hundreth prime is @primes[99]; 12:04
12:05 [Sno] joined, ][Sno][ left
smls If it has to be written as @primes.list[99] it will store everything, right? 12:05
12:05 isBEKaml left
jnthn Actually that will be a compile-time error, 'cus Seq isn't Positional 12:05
So the bind to the @primes symbol will fail
Unless we decide that such a constant declaration will be a place we call .list for you 12:06
Which we may need to do given constant %h = a => 1, b => 2; can only work if we call .hash for you.
smls Ah. So no more my @fib := 1, 1, *+* ... *; either.
jnthn Right, that'll have to be ::= or, you know, just = :) 12:07
12:07 brrt joined
smls What would ::= turn it into, a List? 12:11
jnthn yes 12:12
dalek kudo/glr: 6e7f804 | jnthn++ | src/vm/moar/ops/perl6_ops.c:
Fix silly thinkos in tweak to p6recontrv op.
 12:15
jnthn lunch; bbiab 12:21
12:23 kaare__ left 12:27 nys joined, fxer left 12:29 jkva left 12:43 azawawi joined
azawawi hi 12:43
12:44 brrt left
azawawi what's perl 6 syntax for MyModule->my_sub? 12:44
12:44 xinming left, xinming joined
RabidGravy you mean "module Foo { our sub bar() { } }; Foo::bar();" ? 12:45
azawawi cool thx
RabidGravy the "our" is important there 12:46
12:50 Aurelie left 13:02 fxer joined 13:07 leont left
azawawi m: module Foo { our sub foo() { say "Foo called"; } }; say Foo::<foo>.defined; 13:11
camelia rakudo-moar a32c14: OUTPUT«False␤»
13:12 rmgk_ joined, rmgk left, rmgk_ is now known as rmgk
smls m: module Foo { our sub foo() { say "Foo called"; } }; say Foo::<&foo>.defined 13:12
camelia rakudo-moar a32c14: OUTPUT«True␤»
smls need a & to refer to subroutines as nouns
According to the design docs, &Foo::<foo> should also work I think, but doesn't in Rakudo. 13:13
azawawi it is confusing from an OOP Java background :) 13:14
why could we simply say Foo.meta<foo>.defined...
smls what would be the benefit of that? 13:15
Perl 6 has sigil while languages like Java don't; that's just something one has to get used to.
*sigils 13:16
azawawi so & is an operator or a sigil for module/sub reference? 13:17
smls it is a sigil for referring to the function object of a subroutine
because a bare subroutine name *calls* it. 13:18
m: say &lc;
camelia rakudo-moar a32c14: OUTPUT«sub lc (Cool $s) { #`(Sub|54401328) ... }␤»
13:19 skids joined
smls m: say rand; say &rand; #clearer example 13:19
camelia rakudo-moar a32c14: OUTPUT«0.746277711735317␤sub rand ( --> Num) { #`(Sub|50532104) ... }␤»
azawawi smls: thanks for the explanation 13:23
13:25 jkva joined 13:29 jkva left
jnthn wonders how two non-OO languages features can be confusing from any OOP background... :-) 13:30
The closet Perl 6 equivalent to a Java static method is a Perl 6 method that happens to not access any attributes. 13:32
(Except Perl 6 ones are virtual)
nine azawawi: the equivalent to MyModule->my_sub is MyModule.my_sub. Or rather MyModule.my_method, since a sub that's called with -> really is (used as) a method in Perl 5. 13:38
13:44 rilly42 left 13:49 lolisa joined
azawawi jnthn: true but remember JavaClass.staticMethod is way simpler to remember than MyP6Module::our_sub :) 13:50
jnthn: s/remember/remember that/ 13:51
13:52 aborazmeh joined, aborazmeh left, aborazmeh joined
azawawi File::HomeDir->my_home is a bit more elegant than File::HomeDir::my_home. Please note the extra :: namespace noise at the end 13:52
RabidGravy azawawi, if the my_home doesn't need to use any attributes of an instance then it can be a method 13:54
so e.g 13:56
m: class Foo { method bar() { say "bar" } }; Foo.bar 13:57
camelia rakudo-moar a32c14: OUTPUT«bar␤»
azawawi RabidGravy: i switch between production java, javascript and experimental perl 6 script every day... so yeah confusion does happen :)
13:57 kaare__ joined
RabidGravy :) 13:57
as soon as you add an attribute it goes wrong 13:58
azawawi RabidGravy: my main worry is that future p6 programmers will use classes instead of modules because of the syntax 13:59
RabidGravy m: class Foo { has $!thing = "bar"; method bar() { say $!thing; } }; Foo.bar
camelia rakudo-moar a32c14: OUTPUT«Cannot look up attributes in a type object␤ in method bar at /tmp/4Yx9fUWC0t:1␤ in block <unit> at /tmp/4Yx9fUWC0t:1␤␤»
azawawi m: class Foo { my $thing = "bar"; method bar() { say $thing; } }; Foo.bar 14:00
camelia rakudo-moar a32c14: OUTPUT«bar␤»
azawawi so my is afterall a static shared modifier :)
ShimmerFairy I don't think Foo::bar is as bad as you're making it out to be, honestly. 14:01
nine azawawi: there's still EXPORT for modules providing subs
azawawi nine: i know :)
dalek ast: 2b1c2dc | skids++ | S04-phasers/begin.t:
Add (fudged) test for RT#123777
 14:06
14:13 rurban_ left 14:14 aborazmeh left
dalek ast: afea75d | skids++ | S04-phasers/begin.t:
Add (fudged) test for RT#123776
 14:15
kudo/glr: 9f00b92 | jnthn++ | / (6 files):
Toss various *Iter classes.

They're being replaced as part of GLR, and are no longer top-level visible types.		 14:17
kudo/glr: c4ad536 | jnthn++ | / (4 files):
Toss Parcel and LoL (go away in GLR).

We still need to resolve what type a semilist produces, but it may not end up being called LoL, and certainly won't end up the same on the inside.
smls
.oO( MultiList )
 14:21
14:22 colomon left
jnthn Given it doesn't have to be lazy on the first level and doesn't have to support assignment, and it's really an implementation type, I think a little bit about whether it wants to be one of those low-level types (based on VMArray) 14:22
So far I think that **@foo will just be a List like *@foo 14:23
The latter just binds an iterator that flattens the incoming things
dalek ast: a7118a8 | skids++ | S14-roles/basic.t:
Add test for RT#120646 which is closeable now
 14:30
ShimmerFairy jnthn: the only thing I can think of where a multi-dim array would be useful would be with shaped multi-dim variables, e.g. my @foo[3;2]; I'm not sure how that would look if it was just a List of Lists, for example. (Put another way, I don't know how similar List of Lists would be to std::vector<std::vector<...>> in terms of being not-so-structured :P)
14:30 azawawi left
jnthn sivoais: To be clear, I'm discussing what the type of the "3;2" is 14:31
oops
ShimmerFairy: ^^
14:33 araujo left
ShimmerFairy ah 14:33
14:34 rurban joined
ShimmerFairy For 3;2 specifically, offhand I only see those happening in [] subscripts and as a convenient way of writing out multi-dim lists (that is, @list = (1,2;3,4;5,6) ), so maybe a low-level type for the cases where it's not just convenience syntax would be appropriate. 14:36
14:36 maettu joined
skids Well, we'll know after we implement Keccak I guess. :) 14:37
ShimmerFairy the only place such a low-level type would crop up would then be in custom definitions of [] subscripting, I think.
jnthn To be clear what I mean by "low level", I don't mean it behaves extremely weirdly 14:38
Just that you can't mix in to it, and if you inherit from it you can't add attributes.
ShimmerFairy something akin to the native types then, if I understand correctly?
jnthn Pretty much 14:39
native arrays and bufs behave the same way, fwiw
ShimmerFairy I can't think of enough cases for a multidim array that can't be handled by putting Positionals inside Positionals to justify the "prominence" of a higher-level type :) 14:41
dalek kudo/glr: 378fcdd | jnthn++ | / (3 files):
Add IterationBuffer to CORE.setting.
kudo/glr: e7293c5 | jnthn++ | / (9 files):
Add GLR iterator/iterable bits to CORE.setting.

This replaces Iterable and Iterator (swapping around their order in the build) with the versions from the intial GLR exploration. It also brings in a number of related types that are part of the iteration API.
jnthn ShimmerFairy: It's primarily a dispatch question as to why ; can't just produce List, fwiw. 14:42
postcircumfix:<[ ]> needs to distinguish @a[1,2,3
postcircumfix:<[ ]> needs to distinguish @a[1,2,3] from @a[1;2;3]
smls ShimmerFairy: Having LoL as a separate type allows things like postcircumfix [] and prefix | to dispatch based on type.
Maybe user-defined things will want dto do the same 14:43
ShimmerFairy smls: yeah, I acknowledged that already :)
14:44 khw joined
smls it might not necessarily be custom implementations on [] though 14:44
ShimmerFairy jnthn: heh, I can't think of another way to handle that aside from a distinct type. (maybe with a really ugly 'where' clause. _maybe_.)
smls people miught find other places where a similar distinction is needed 14:45
in custom APIs
jnthn It pretty much has to be a type distinction
ShimmerFairy jnthn: just ooc, would this multidim type be n-dimensional, or is it just, say, 2-dimensional?
jnthn Most of our interesting optimizations hang off type distinctions.
ShimmerFairy: It's not anything dimensional really 14:46
ShimmerFairy: It's the thing you (primarily) use to talk about accessing something that *is* multi-dimensional
ShimmerFairy: And its number of elements will typically match the number of dimensions the target you're indixing into has
my @a[3;3]; @a[0;0] = 42; # @a is 2-dimensional, so 0;0 has two elements 14:47
Well, "so" is a bit misleading, but... )
ShimmerFairy So it's more like a wrapper, then? If so, then I understand :) 14:48
jnthn It's a thingy that holds the indexes you want to apply to the different dimensions.
ShimmerFairy Yeah, it's not a new array implementation, just a description of an array that's more interesting than the usual 1-dimensional varieties, as I understand it. :) 14:49
jnthn Yeah, data strucutre wise it's terribly boring. We only care for the type distinction. 14:51
14:52 lucasb joined, diana_olhovik_ left
smls jnthn: pre-GLR, LoL is also being used for other things that describing indices though. 14:52
For example, .tree returns a LoL 14:53
ShimmerFairy
.oO( sub postcircumfix:<[ ]>(MultiDimIndices $a) { ... } )
lucasb omg, GLR is Getting LoL Removed?
smls jnthn: And .tree is needed pre-GLR to work around excessive flattening when dealing with things like metaops, e.g. @a X (@b Z @c).tree.list 14:54
timotimo NO LOL
ShimmerFairy smls: I think think those uses of LoL would just be Lists of Lists now :P
jnthn You can all still work on project euler though :P 14:55
smls ShimmerFairy: yeah, looks like it.
jnthn smls: Right, I don't think we'll need .tree anything like so often
smls: And maybe it'll want to be more about itemizing 14:56
ShimmerFairy I wasn't aware of .tree , much less its use to prevent excessive flattening. On the one hand I wonder where I could've used it, but on the other I'm glad that soon I won't have to find out :) 14:57
14:57 isBEKaml joined
smls jnthn: yeah 14:57
jnthn: Though it .tree is just a shortcut for .map(*.item) maybe it's not needed at all... :P 14:58
jnthn smls: There's also that, yes. :)
14:59 Zoffix_ is now known as Zoffix, llfourn left
dalek kudo/glr: 9236d43 | jnthn++ | / (6 files):
Bring in Seq and HyperSeq from initial GLR work.
 15:01
maettu 2.5 weeks until P6 hackathon @perl-workshop.ch :-)
jnthn \o/ 15:02
dalek ast: b3831d1 | skids++ | S32-str/comb.t:
Add (fudged) tests for RT#123760 which could perhaps be tagged LHF
 15:04
ChoHag Oh is GLR working its way into core now? 15:05
15:06 maettu2 joined, maettu2 left
skids ChoHag: jnthn++ is working it in in a branch 15:07
ChoHag Nifty.
15:07 maettu left, colomon joined 15:08 rurban left 15:09 maettu joined 15:17 pmurias joined
dalek ast: 1e0f7ad | skids++ | S12-coercion/coercion-types.t:
Mention both RT#123770 and RT#124839. These tickets should be merged.
 15:24
jnthn Any idea why minmax is taking **@args rather than *@args like min and max? 15:26
jnthn just tosses the ** 15:28
The tests'll tell me
RabidGravy that's the spirit 15:31
arnsholt I do that all the time. Refactor some thing, and then just give the tests a whack to see what falls out =) 15:35
nine That's what tests are for ;) 15:36
15:39 colomon left
dalek kudo/glr: 1e0eb3c | jnthn++ | src/core/Any.pm:
Toss reference to MapIter.
 15:42
kudo/glr: 6f887c7 | jnthn++ | src/core/Mu.pm:
Add GLR note about Mu.infinite.
kudo/glr: c01a6ba | jnthn++ | src/core/Any.pm:
Clean up list/hash coercion-y methods.
ast: df84392 | skids++ | S32-scalar/perl.t:
Add test for RT#123741 which can likely be closed
15:43 llfourn joined
dalek kudo/glr: e8423e0 | jnthn++ | src/core/Any.pm:
() will make a List, so ().list is just () now.
 15:47
kudo/glr: b98bc5f | jnthn++ | src/core/Any.pm:
Fix double-coercion thinko.
kudo/glr: 23eb297 | jnthn++ | src/core/Any.pm:
Toss .lol and comment out .tree for now.

And leave GLR note on tree as something needing resolving.
nine YES YES YES! I got it :) 15:48
dalek line-Perl5: f018d4b | (Stefan Seifert)++ | p5helper.c:
Fix "Attempt to free unreferenced scalar during global destruction"

Arguments to Perl 6 methods or callables are passed as array reference. We created the array, and created a reference to this array to pass it on to Perl 6 code. After the call we decreased the refcount of the array instead of this reference.
The reference then got leaked and would be collected by perl during global deconstruction. The reference's dtor would then try to decrease the refcount of the array it was referencing which already was destroyed.
This is what Perl was detecting and the source of the error. Fixed by decreasing the refcount of the reference instead of the array.
jnthn nine++ 15:49
nine I'm so glad we're getting rid of refcounting in Perl 6. This is horrible :)
dalek kudo/glr: 751f6bf | jnthn++ | src/core/Any.pm:
Fix minmax inconsistency with everything else.

Also toss commented out code left with no reason.		 15:50
15:52 diana_olhovik_ joined 16:01 Begi joined
Begi my @a = 0..*; my @a = 0...*; Is there a difference ? 16:02
mst nine: I like timely destruction :) 16:07
leedo Begi: one causes a segfault? :P gist.github.com/leedo/8abb6d227e8d843b259f 16:08
jnthn Odd, I'd have expected it to run out of memory rather than SEGV. 16:09
leedo jnthn: i think it is an issue with the repl, does'nt happen with perl6 -e
jnthn But yeah, .. makes a Range object, and 0...* makes an infinite sequence, and asking for the minimum thing in an infinitely long thing isn't going to go too well...
nine mst: yes, it certainly has its uses. But no way it can be worth the pain. Every time I come back to the refcount topic, I need a day to start making progress, because one has to keep all the intricacy of the API in ones head at the same time to get it right. And I find my way around perl's source code quite well because the docs only tell you have of what you need to know. 16:10
dalek line-Perl5: 265bcdb | (Stefan Seifert)++ | / (4 files):
Fix isa(P5::subclass) of a P5::subclass of a P6 extended P5 class

Take a Perl 5 class. Extend it using v6-inline and v6::extend in the constructor. Create a Perl 5 subclass of this construct.
With this patch, objects created by the constructor pass an
  ->isa(MySubclass) check.		 16:12
jnthn Time for a break; maybe more GLR later, if not then got all day on it tomorrow.
RabidGravy with one of callwith/nextwith/callsame/nextsame in a method can you supply a different invocant somehow? 16:14
16:25 fxer left, isBEKaml left 16:35 rurban joined 16:43 Begi left 16:46 firefish5000 joined 16:47 pat_js joined
skids RabidGravy: you want to use the current multi/inheritence dispatch candidate state from one object on another? That sounds dangerous. 17:08
RabidGravy it's in a wrap 17:09
the hack I am doing *is* dangerous
:) 17:10
skids Are they at least of the same class? I don't have a solution, just intrigued how the heck you got that deep in the thicket. 17:11
RabidGravy yeah
skids Sort of like when you encounter a rusted car wreck in the middle of the woods and wonder how the heck it got in there in the first place :-) 17:12
RabidGravy specifically I'd like to arrange a class such that if the method gets called on the type object then the wrapper arranges that it gets called on a singleton instance 17:13
17:16 lolisa left
RabidGravy obviously one could design a class such that is was only class methods and keep state in my variables or so, but I'm curious to see if a class could be made that works both ways 17:18
just a doodle in my internal pattern book ;-) 17:21
17:22 brrt joined 17:23 domidumont joined
jnthn RabidGravy: I'd probably solve that just with a custom meta-class that overrides add_method to wrap the thing up and compose to automatically add some Singleton role like role Singleton { method get-instance() { state $instance = self.new } } or so 17:25
so add_method would be like callwith($name, -> \obj, |c { obj.DEFINITE ?? $meth(obj, |c) !! $meth($obj.get-instance, |c) } or so 17:26
17:26 firefish5000 left
RabidGravy ahah! 17:29
brrt good * 17:33
RabidGravy jnthn, hadn't quite grokked callwith before that moment 17:34
skids m: my $a = "1"; my $b = "2"; $b [R~]= $a; $b.say # yay I finally got to use R 17:35
camelia rakudo-moar a32c14: OUTPUT«12␤»
17:38 raiph joined, rangerprice joined
RabidGravy jnthn, where I was going was something more like: 17:44
m: gist.github.com/jonathanstowe/13f5...ce74776f48
camelia rakudo-moar a32c14: OUTPUT«this one␤»
RabidGravy which appears to work 17:45
17:46 llfourn left
jnthn RabidGravy: ah, that also works nicely, at the cost of having the label all the single methods 17:47
17:52 pmurias left
jnthn (Which isn't a problem if only some should do it.) 17:52
17:54 rurban left
nwc10 jnthn: you expect origin/glr to fail (no ASAN explosion, just a regular fail) in the setting build? 18:06
Stage parse : Error while constructing error object:Could not locate compile-time value for symbol X::Parameter::InvalidType 18:07
Error while compiling, type X::Parameter::InvalidType
suggestions: (unstringifiable object) typename: LoL at line 3733, near " \\keys, *%"
at gen/moar/stage2/NQPHLL.nqp:516 (/home/nicholas/Sandpit/moar-san/share/nqp
oops ... /lib/NQPHLL.moarvm:panic:105)
etc
RabidGravy BOOM! BOOM! FILL THE ROOM! 18:10
rangerprice Hello i'm a cheezburger
eat me if you can
RabidGravy nah you're fine I already ate 18:14
jnthn nwc10: Yeah, it's going to be explodey for a while 18:22
nwc10: There's no good way to do it in non-break-the-build steps (unless you take one huge step :)) 18:23
18:23 jjido joined 18:26 zakharyas joined 18:29 yqt joined
dalek kudo-star-daily: dae7da0 | coke++ | log/MoarVM- (5 files):
today (automated commit)
 18:33
rl6-roast-data: 2408b3e | coke++ | / (9 files):
today (automated commit)
rl6-roast-data: 34411fa | coke++ | / (9 files):
today (automated commit)
[Coke] (swiss perl workshop hackathon) - I'm only a maybe at this point. 18:40
japhb [Coke]: Sadly, I'm at *best* a 20% maybe right now. :-( 18:43
18:45 rindolf left, jjido left 18:46 jjido joined
brrt wow, many cool :-o developer.nvidia.com/devbox 18:49
18:54 shinobi-cl joined
shinobi-cl r: class C1 { has @!odds; has @!data; submethod BUILD(:@!data*) { @!odds = gather for (@data) -> $v { take $v if ($v div 2 == 0) } }; method new (@new-data) { return new(data=>@new-data); } }; 18:57
camelia rakudo-{moar,jvm} a32c14: OUTPUT«5===SORRY!5=== Error while compiling /tmp/tmpfile␤Missing block␤at /tmp/tmpfile:1␤------> 3dds; has @!data; submethod BUILD(:@!data7⏏5*) { @!odds = gather for (@data) -> $v {␤»
shinobi-cl r: class C1 { has @!odds; has @!data; submethod BUILD(:@!data) { @!odds = gather for (@data) -> $v { take $v if ($v div 2 == 0) } }; method new (@new-data) { return new(data=>@new-data); } }; 18:59
camelia rakudo-{moar,jvm} a32c14: OUTPUT«5===SORRY!5=== Error while compiling /tmp/tmpfile␤Variable '@data' is not declared. Did you mean '@!data'?␤at /tmp/tmpfile:1␤------> 3d BUILD(:@!data) { @!odds = gather for (7⏏5@data) -> $v { take $v if ($v div 2 == ␤»
shinobi-cl r: class C1 { has @!odds; has @!data; submethod BUILD(:@!data) { @!odds = gather for (@!data) -> $v { take $v if ($v div 2 == 0) } }; method new (@new-data) { return new(data=>@new-data); } };
camelia rakudo-{moar,jvm} a32c14: OUTPUT«5===SORRY!5=== Error while compiling /tmp/tmpfile␤Undeclared routine:␤ new used at line 1 (in Perl 6 please use method call syntax instead)␤␤»
shinobi-cl r: class C1 { has @!odds; has @!data; submethod BUILD(:@!data) { @!odds = gather for (@!data) -> $v { take $v if ($v div 2 == 0) } } method new (@new-data) { return C1.new(data=>@new-data) } }; 19:00
camelia rakudo-{moar,jvm} a32c14: OUTPUT«5===SORRY!5=== Error while compiling /tmp/tmpfile␤Strange text after block (missing semicolon or comma?)␤at /tmp/tmpfile:1␤------> 3 -> $v { take $v if ($v div 2 == 0) } }7⏏5 method new (@new-data) { return C1.new(␤ expecting an…»
shinobi-cl r: class C1 { has @!odds; has @!data; submethod BUILD(:@!data) { @!odds = gather for (@!data) -> $v { take $v if ($v div 2 == 0) } }; method new (@new-data) { return C1.new(data=>@new-data) } };
dalek ast: 22ee35b | skids++ | / (2 files):
Teach Test::Compile to chain compunits; add fudged tests for RT#123679
camelia ( no output )
19:01 [Coke] left, [Coke] joined
smls jnthn: In src/core/Seq.pm of the glr branch, you have a role called PositionalBindFailover but refer to it as MemoizeAsPositional in the preceeding comment (I think). 19:11
Also, if this is gonna be public API, shouldn't that role have a cuter name such as Listy? :P 19:13
19:16 BenGoldberg left
dalek ast: 4df4e7b | skids++ | S32-scalar/perl.t:
Add a 'use lib' needed when running make spectest
 19:25
19:27 lucasb left
jnthn I suspect MemoizeAsList is closer, but its role in binding is hinted better by the other name... 19:38
No, it shouldn't have a cute name at all
Generally, all the things with short cute names are the ones we expect people to use in normal programs.
If you are using this role you've probably implemented something like a library that runs list operations off on the GPU and want the pipeline to behave like Seq and HyperSeq do. 19:39
Or something similarly unusual.
timotimo hopefully that will not be too unusual 19:40
though I don't quite see how you could get at the code for things in the pipeline to translate to GPU instructions
19:42 llfourn joined 19:46 llfourn left 19:47 jkva joined 19:48 espadrine left 19:51 ChoHag left 19:52 telex left 19:54 telex joined
jnthn Back tomorrow :) o/ 19:57
20:00 cognominal left 20:01 gnosti joined 20:04 BenGoldberg joined
gnosti Hi, I uploaded rakudobrew, then run 'rakudobrew moar', then 'rakudobrew panda'. The latter fails with pasted.co/1a6f9a5f Any suggestions? I'd like to experiment with NativeCall. 20:08
20:08 nys left
gnosti s/uploaded/updated/ 20:08
20:14 rurban joined 20:16 BenGoldberg left 20:22 zakharyas left, cognominal joined 20:23 darutoko left
RabidGravy ghosti, that's odd. it looks like it's using a way old version of rakudo 20:24
20:24 brrt left
RabidGravy but it's "rakudobrew build-panda" 20:24
smls Don't you have to write it "rakudobrew build-panda"? Rather than "rakudobrew panda"
oh, RabidGravy beat me to it 20:26
:P
RabidGravy :)
to be fair I'm probably just typing faster 20:28
smls that's quite possible.
I'm not a fast typer, and my keyboard could use a replacement... 20:29
RabidGravy I've been doing it so long that I wave my hands in front of the keyboard and words come out, sometimes they even make sense 20:31
gnosti Right, the second command was 'rakudobrew build-panda', and I can't understand why 'rakudobrew moar' would end up with a rakudo from a year ago. Several git 'up to date' messages appeared. I suppose I can blast everything away and try again.
RabidGravy I also have a lighty up keyboard which clearly helps
gnosti, it's "rakudobrew build moar && rakudobrew build-panda" for the avoidance of any doubt 20:33
if you have a particularly old "rakudobrew" it may also be a "rakudobrew self-upgrade" first in the unlikely possibility it's sufficientlly old it's getting the wrong branch 20:34
20:36 diana_olhovik_ left
gnosti RabidGravy, "rakudobrew self-upgrade" says already up-to-date. 20:38
RabidGravy got me beat then, right in the middle of something I want to finish tonight otherwise I would look in more depth 20:41
20:42 llfourn joined
timotimo worst case, a rakudobrew nuke might help 20:43
gnosti sees a problem that moar-nom was left in "detached head" state with last commit in 2014. Checking out 'nom' and repeating "rakudobrew moar" 20:45
20:46 cognominal left, llfourn left 20:48 TEttinger joined
gnosti Thanks timotimo, I'll nuke, rinse and repeat. 20:49
RabidGravy but you'll love the NativeCall stuff when you're done ;-)
20:54 shinobi-cl left 20:59 domidumont left 21:02 rurban left
dalek ast: 66a1ea8 | skids++ | S06-signature/caller-param.t:
Reap merged ticket numbers, and mention RT#123660
 21:02
21:03 jkva left 21:06 nwc10 left
gnosti Thanks all, the build process went fine this time :) 21:06
RabidGravy yay! 21:07
timotimo cool 21:08
enjoy the fruit of our collective labours!
RabidGravy remember ther's stll 25,600 modules to be finished before christmas ;-) 21:09
lizmat m: $:: 21:14
camelia rakudo-moar a32c14: OUTPUT«===SORRY!===␤MVMArray: Index out of bounds␤»
bartolin_ lizmat: AFAIU there are a few tickets related to that error: RT #125659, RT #113042, RT #115326, RT #119999, RT #120252 21:17
lizmat yeah, I think I've found a way around this.... 21:18
bartolin_ lizmat++
lizmat I mean, it should just die on a malformed longname, no?
6 '$::'
===SORRY!===
No named components found in longname
something like this, no?
bartolin_ I was looking at it and wasn't sure about that.
std: $:: 21:19
camelia std 28329a7: OUTPUT«ok 00:00 134m␤»
bartolin_ lizmat: std parses that the same way as a single $
21:19 ShimmerFairy left
dalek ast: ed79556 | skids++ | S09-typed-arrays/arrays.t:
Add test for RT#122440/RT#124079
 21:25
bartolin_ m: my $foo = 42; say $::("foo")
camelia rakudo-moar a32c14: OUTPUT«No such symbol '$::foo'␤ in block <unit> at /tmp/MakaEiqUtF:1␤␤Actually thrown at:␤ in block <unit> at /tmp/MakaEiqUtF:1␤␤»
bartolin_ that one is differen locally: 21:26
m 'my $foo = 42; say $::("foo")'
42
m: my $foo = 42; say $::<foo> # shouldn't that be the same as $::("foo") 21:27
camelia rakudo-moar a32c14: OUTPUT«===SORRY!===␤MVMArray: Index out of bounds␤»
21:27 pat_js left
lizmat $ 6 '$::' 21:29
===SORRY!=== Error while compiling -e
Variable '$' is not declared
is that a more acceptable error?
smls bartolin_: RT #123903 too 21:30
RabidGravy phew got in the commit, for 76 days ;-)
bartolin_ smls++ (I missed that one) 21:31
21:32 ShimmerFairy joined
bartolin_ lizmat: what do you get with your patch for '$::<foo>'? 21:34
lizmat $ 6 '$::<foo>' 21:35
===SORRY!=== Error while compiling -e
Variable '$' is not declared
so I guess, even though the spectest is clean, this is not the right way
bartolin_ looking at STG
oops 21:36
looking at STD.pm6 it seems to be more sophisticated than Grammar.nqp with regard to 'token variable': github.com/perl6/std/blob/master/STD.pm6#L2444
RabidGravy I'd just like to thank perl 6 for making this kind of refactpr so easy: github.com/jonathanstowe/Audio-Con...a65bb1bea8
21:37 fxer joined
lizmat m: $::<foo> 21:40
camelia rakudo-moar a32c14: OUTPUT«===SORRY!===␤MVMArray: Index out of bounds␤»
dalek kudo/nom: 68f20f4 | lizmat++ | src/Perl6/World.nqp:
Attempt at fixing #125772 and friends

This merely replaces one type of LTA error by another hopefully less LTA error. This feels like a bandaid and is probably not the right way to do this. But I hope it will put people on track for this one.		 21:41
21:41 FatalNIX left, FatalNIX joined
lizmat I'll look at this again tomorrow when I'm not as tired 21:42
bartolin_ lizmat++
RabidGravy shouldn't "less LTA" be "slightly more awesome"? 21:43
dalek ast: db772bd | skids++ | S03-smartmatch/array-array.t:
Add (fudged) tests for RT#123144
 21:44
RabidGravy or possibly "slightly nearer to awesome" 21:45
:)
gnosti timotimo, it all looks very tasty ;) 21:46
bartolin_ btw $:: and friends seemed to blow up in 'method variable_components' in src/Perl6/World.nqp, called from src/Perl6/Actions.nqp, line 1866
oh, that's where lizmat++ changed the code. nevermind 21:48
bartolin_ should sleep now 21:49
RabidGravy over-rated 21:53
21:59 jkva joined 22:04 jkva left 22:05 mjgardner joined 22:31 smls left, smls joined, smls left 22:34 jjido left
dalek ast: 0d72d2b | skids++ | S04-statements/if.t:
Add tests for RT#105872 (the while case needs one after GLR)
 22:40
22:42 mjgardner left 22:43 llfourn joined 22:44 mjgardner joined 22:48 llfourn left, perturbation joined 23:02 yqt left 23:05 cognominal joined 23:06 tinyblak joined, perturbation left
dalek ast: f97295c | skids++ | S04-statements/for.t:
mention RT #118705
 23:19
RabidGravy m: say (^2**8).pick(10).map({ ($_/2**8)*(2**32)}) 23:26
camelia rakudo-moar a32c14: OUTPUT«1493172224 201326592 2734686208 1157627904 1056964608 1375731712 536870912 2046820352 1711276032 1996488704␤»
RabidGravy given a bunch of values describing a physical dimension expressed as 8 bit values is that the best way to comvert the scale to 32 bit values 23:30
23:30 Peter_R left
dalek kudo-star-daily: d8e3e32 | coke++ | log/MoarVM- (5 files):
today (automated commit)
 23:33
skids RabidGravy: depends. You might want dithering. :-) 23:35
RabidGravy don't admit to understanding those things around me ;-) 23:36
you might get roped in
but for dumb that's right yeah? 23:37
skids But buf32.new[@values X+< 24] might work.
(You're really leaning on strength reduction by the compiler with the multiplications and divisions) 23:41
23:44 dayangkun joined
RabidGravy my actual interest in this is emulating the Oto Machines Biscuit in software 23:45
skids Don't bother going down to 8 bit for that, just mask off the lower bits at the depth you want to work with. 23:51
What will be hard s replicating the input and output analogue filters.
RabidGravy does that work for emulating an 8 bit dac from some arbitrary bitwidth input ? 23:59

Powered by Cro and the Raku® Programming Language.

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

Thank you!