»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.
caymanboy Alvaro says hi 01:03
from the Cayman Islands
TimToady howdy 01:04
caymanboy waiting on Perl6 1.0
For Xmas
TimToady fondly remembers a miniature golf course there...
caymanboy by Bamboo/Deckets on Seven Mile beach? 01:05
TimToady don't remember the name, but it was kind of up on a little hill
probably blew away in the hurricane 01:06
caymanboy yes, I pass it almost everyday
nope still there
TimToady cool, my wife will be glad to hear that
gotta get back there someday, but not till after Xmas :)
caymanboy bring her back. its grown much since the hurricane
TimToady I'll look forward to it 01:07
caymanboy Bring her for the "Perl fans in Cayman" tour 01:08
caymanboy Brink the wife to www.xqs.ky, rey have a perl6 burger 01:11
^they
caymanboy So could someone splain me is perl6 idenpendent of a vm? I'm somwhat confused with the parrot vm and other vms 01:20
TimToady yes, perl6 is defined by the test suite, so anything that passes the test suite is official Perl 6 01:22
so, for instance, the rakudo compiler runs on either MoarVM or on JVM
(doesn't run at the moment on parrot, because it was being too hard ot maintain, and we won't get done by Xmas if we try to use too many backends) 01:24
caymanboy I thought at somepoint there was a rakudo implementation on the parrot vm. for a while thought parrot and perl6 were somewhat symbiotic 01:28
TimToady the goals of the two projects diverged 01:31
caymanboy Have not been keeping up as much as I should have. been battling my python friends.
is the xmas date still on the radar? 01:34
TimToady yes, we're going to have some thing by this Christmas 01:35
caymanboy as in 1.0? 01:38
TimToady well, a first official language test suite, whether it's called 1.0 or somethingelse
and at least one implemention of it, on MoarVM (and maybe JVM) 01:39
language versions being separate from compiler versions, y'see
caymanboy yes, sort of 01:41
TimToady dinner & 01:53
caymanboy rum 01:54
AlexDani` what's the difference between <somerule> and <.somerule> in grammars? 02:17
ugexe <.somerule> doesnt get captured
still matches 02:18
AlexDani` ugexe: yea, thanks! 03:08
is there any way to skip some of the results from the match object but keep their insides? For example: regex cont { <in1> <in2> }, I want to have <in1> and <in2> in the match object but I don't really care about <cont> itself. Using <.cont> will not help because <in1> and <in2> will not be captured. 03:12
raiph AlexDani`: irclog.perlgeek.de/perl6/2015-04-29#i_10522960 03:31
raiph AlexDani`: ^^ prolly worth a read but maybe not what you want 03:33
ugexe i've been using Action classes to clean that type of thing up into a final data structure instead of trying to pass around a huge Match object 03:34
ugexe i know your pain though. i have stuff like: $http.<HTTP-message>.<header-field>.[1].<value>.<accept-value>.[0].<media-range>.<type> 03:37
AlexDani` raiph: nice stuff anyway 03:40
dalek kudo/nom: c17db6c | (Nick Logan)++ | src/core/Proc (2 files):
Proc.shell/run/spawn and QX :$cwd param

Allow setting :$cwd for Proc and Proc::Async. Lowercase 'cwd' per suggestion.		 03:47
kudo/nom: ec3605f | lizmat++ | src/core/Proc (2 files):
Merge pull request #460 from ugexe/proc-cwd

  Proc.shell/run/spawn and QX :$cwd param
AlexDani` if I have a regex like: regex stuff { <start> <hardstuff> <end> } is there any way to give a hint to the parser to first find <end> and only then attempt to match <hardstuff> in between <start> and <end>? Basically I want to skip expensive parts by throwing away paths that are obviously wrong. 03:50
ugexe <before hardstuff> or <after start> might work like that 03:55
AlexDani` ugexe: I've thought about before and after, but where do I write these to achieve that effect? 03:57
ugexe m: my $x = "abc123"; say $x ~~ /<alpha> <before <digit>>/ 03:58
camelia rakudo-moar 0c8852: OUTPUT«｢c｣␤ alpha => ｢c｣␤ before => ｢｣␤»
AlexDani` hm, and what's the difference between <before and <?before ? 04:00
AlexDani` ugexe: I'll try that, thanks 04:02
ugexe never actually thought about that. i suppose its if you dont care to save the fact its an 'after' or 'before' type match 04:03
there is also <!before and <!after
ugexe so now we have: Proc.out, Proc::Async.stdout, and IO::Socket::Async.[chars|bytes]_supply :o 04:21
dalek ast: f713932 | skids++ | / (2 files):
Add Test::Idempotency package

Fudge a few of said tests, RT needed (rakudo misorders subsig/default) Add a fudged test for :(:a(:b($a))).perl, RT needed		 04:46
dalek ast: 9677e70 | skids++ | S06-signature/introspection.t:
Remove spurious semicolon from last commit
 05:10
kanl m: if '[FOO:@]' ~~ /'[' (\w+) ':'? (<-[\]]>*) ']'/ { $0.say; $1.say; my $s = "foo.$0{ $1 ?? "-{ $1 eq '@' ?? 'bar' !! 'baz' }" !! '' }"; $s.say } 05:41
camelia rakudo-moar ec3605: OUTPUT«｢FOO｣␤｢@｣␤Use of Nil in string context in block at /tmp/DVqVf9JNoE:1␤foo.␤»
kanl m: if '[FOO:@]' ~~ /'[' (\w+) ':'? (<-[\]]>*) ']'/ { $0.say; $1.say; my $s = "foo.$0-{ $1 ?? "-{ $1 eq '@' ?? 'bar' !! 'baz' }" !! '' }"; $s.say } 05:47
camelia rakudo-moar ec3605: OUTPUT«｢FOO｣␤｢@｣␤foo.FOO--bar␤»
kanl ok
kanl gurus, if i were to do a global match and replace, and the replacement string needs to be determined by some computation on the match/capture. do i have to wrap .subst inside a loop of match, compute, replace? i suppose it'd make sense to have multi method subst(Str:D: $matcher, $pointy-sub, *%opts) 06:03
moritz kanl: you can just call .subst(:g, $regex, $sub) 06:05
kanl oh? nice! the docs did't say, so i didn't know :p thanks!! 06:06
wasn't sure about the interaction of $sub with $/ 06:07
but now that i think about it, it makes sense. 06:08
m: my $s = 'poop'; $s.subst-mutate: /p/, -> $m { 'b' }, :g; $s.say 06:11
camelia rakudo-moar ec3605: OUTPUT«boob␤»
ugexe does this looks like it could fail to emit captured data if line 85 ($emitting = 0) is not reached before a separate thread with the final chunk of buffer reaches line 71 github.com/rakudo/rakudo/blob/nom/...pm#L65-L86 ? its a sort of hand-off between locks
@buffer doesnt get cleared rather 06:15
RabidGravy morning! 07:34
nwc10 morning 07:35
dalek ast: 9d9b36a | lizmat++ | S32-hash/adverbs.t:
Comprehensive testing of hash slicing with adverbs
 07:41
kanl my @a = 1, 2, 3; say ( @a.all > 3 ).Bool; my @b; say ( @b.all > 3 ).Bool 07:43
m: my @a = 1, 2, 3; say ( @a.all > 3 ).Bool; my @b; say ( @b.all > 3 ).Bool
camelia rakudo-moar ec3605: OUTPUT«False␤True␤»
moritz m: say so all() 07:44
camelia rakudo-moar ec3605: OUTPUT«True␤»
cdc_ Hello #perl6! For information the "Make a donation" link from news.perlfoundation.org/2015/04/gra...lease.html is broken 07:45
kanl moritz: why would "all of @b > 3" be true, when @b is empty?
dalek kudo/nom: 37b5793 | lizmat++ | t/spectest.data:
Make sure we run hash slicing adverbs tests
 07:46
cdc_ (but the one from www.perlfoundation.org/perl_6_core_...pment_fund is OK)
moritz kanl: for the same reason that all() is True
the empty list/set/junction is the neutral element of junction composition 07:47
and so all(empty list) must be the neutral element of the 'all' / 'and' operation, which is True
or, phrased more simply: if you can't find a counter example, the result is True. You can't find a counter example in an empty list. 07:48
lizmat cdc_: thanks for the report, I've forwarded it to someone at TPF... 07:49
kanl i wanna say that it all makes perfect sense, but it still doesn't sit well with me, so i'd have to digest it a bit :)
thanks, moritz++ 07:50
dalek ast: 0d1ab4c | lizmat++ | S32-hash/adverbs.t:
More correct test of bare zen slice
 07:52
cdc_ lizmat: another issue when creating an account from secure.donor.com/pf012/give -> [email@hidden.address] is invalid"
I do confirm this is a valid address :) 07:53
lizmat forwarded as well... 07:55
afk for at least a few hours&
cdc_ lizmat: thanks!
kanl <riddle> if @b.all > 3 && @b.all < 3 && @b.all == 3, what's in @b? <answer> nothing! 07:56
kanl though i still get the unsettling feeling as though it will spring out of some dark corner and bite me.. 07:59
kanl not sure if it's analogous, but it feels like the issue masak++ blogged, that if $n >= 0 is not true, $n is not necessarily negative :( 08:02
kanl it makes natural sense with .none, but with .all, really not a lovable thing. 08:06
masak morning, #perl6 09:27
kanl: is your complaint that any condition on the elements of an empty list is vacuously true?
masak .oO( "all elements of this list have been to the moon, and gotten a refund!" ) 09:28
masak .oO( "all elements of this list are your personal Secret Santa, and they were also instrumental in planning 9/11" ) 09:29
masak .oO( "all elements of this list will one day return in psychotic wrath to destroy the universe, plus-minus an apple" ) 09:30
well, 09:31
(1) you're in good company. there are lots of dead Medieval logicians who would agree with you in a heartbeat. (they also had serious foundational problems because of this belief.)
(2) if either of the statements above is false, you should be able to come up with a counterexample, namely an element that *hasn't* been to the moon etc. I invite you to do this. 09:32
nine masak: the more you argue, the more I _want_ to find a flaw in your logic ;) 09:37
masak nine: I invite you to do this :)
masak .oO( keep your friends close, keep your debating partners in infinite loops )
nine Problem is, I'm firmly on your side ;) 09:38
masak srsly though, sometimes "common sense" has to stand back for the progress of knowledge. 09:39
ShimmerFairy "All elements of this list are not present in all other lists" 09:40
masak backlogs and notices that moritz already did the "counterexample" schtick
moritz++
nine Even common sense has to admit that if a list is built by putting all elements that satisfy a given condition into it, then all elements in this list must satisfy this condition.
Now if the condition is impossible to satisfy, this still does not change the previous sentence.
masak common sense feels confused! 09:41
ShimmerFairy masak: It reminded me of a -> b , how if a is false then a -> b is always true. At first it makes no sense, but of course there's a logical explanation :P
masak the logical explanation is just that, if 'a' is false, then as far as 'a -> b' is concerned, it's already done and doesn't need to care more. 09:42
the same kind of "short-circuiting" thinking as with && and || in Perl 6.
or, put differently, 'a -> b' isn't a statement about the trueness of 'a', it's a statement about the trueness of 'b' *in a world where 'a' is true* 09:44
if we're not in that world, then we're done because the statement was clearly about some other world :)
ShimmerFairy m: say False orelse True 09:46
camelia rakudo-moar 37b579: OUTPUT«False␤»
ShimmerFairy I'm sure that's not right, unless I misunderstand orelse 09:47
masak you misunderstand orelse 09:49
orelse distinguishes success from failure. False is a falsy value, but it's not failure
you also have precedence wrong 09:50
m: say (Mu orelse 42)
camelia rakudo-moar 37b579: OUTPUT«42␤»
masak m: say (False orelse 42)
camelia rakudo-moar 37b579: OUTPUT«False␤»
masak ShimmerFairy: you can read more about `orelse` in S03
especially the thunkish semantics, which seems to me the big reason to use it 09:51
hm, s/success from failure/defined from undefined/ 09:52
masak which makes me realize that orelse would be pretty well suited for error handling of .index 09:52
m: my $s = "the quick brown fox"; for <brown red fox> -> $word { my $ix = $s.index($word) orelse say "Didn't find the word '$word'" and next; say "'$word' is at $ix" } 09:54
camelia rakudo-moar 37b579: OUTPUT«'brown' is at 10␤Didn't find the word 'red'␤'fox' is at 16␤»
masak m: my $s = "the quick brown fox"; for <brown red fox> -> $word { my $ix = $s.index($word) andthen say "'$word' is at $ix" } 09:55
camelia rakudo-moar 37b579: OUTPUT«'brown' is at 10␤'fox' is at 16␤»
masak m: my $s = "the quick brown fox"; for <brown red fox> -> $word { $s.index($word) andthen say "'$word' is at $_" }
camelia rakudo-moar 37b579: OUTPUT«Use of uninitialized value $_ of type Any in string context in code at /tmp/16_jHckQl2:1␤'brown' is at ␤Use of uninitialized value $_ of type Any in string context in code at /tmp/16_jHckQl2:1␤'fox' is at ␤»
masak apparently we don't bind $_ properly yet.
masak ponders submitting a rakudobug
ShimmerFairy m: sub infix:<→>($a, $b) { $a ?? ?$b !! True }; say $_[0], "→", $_[1], "=", $_[0]→$_[1] for ((True, False) X (True, False)) # I wonder how useful this would actually be in code 09:57
camelia rakudo-moar 37b579: OUTPUT«True→True=True␤True→False=False␤False→True=True␤False→False=True␤»
masak ShimmerFairy: I used it once, in strangelyconsistent.org/blog/boxes-and-pebbles 10:00
sub infix:«⇒»($premise, $conclusion) { !$premise || $conclusion } 10:02
we implement it in different ways, but those two ways have the same truth table :)
ShimmerFairy masak: Yeah, I saw your version described somewhere, but I prefer the version that has a slightly easier to follow truth table :P 10:03
masak is reminded of how much he likes QuickCheck
ShimmerFairy: matter of taste, I guess.
"either the premise is false or the conclusion is true" reads very well to me. and that's implication.
ShimmerFairy masak: There's something about !! True that looks funny to me :P (esp. when in context: "if true then stuff else true") 10:05
masak ShimmerFairy: any time you put True in a ternary ?? !!, you could have used || instead. 10:11
ditto False, &&
ShimmerFairy I can see that.
I wrote down some thoughts I had on S23 (security), since it seems to me that's the one missing synopsis that's actually quite important: gist.github.com/lue/bdb5be0ab2ba78654a01 10:15
masak in fact, that's a cute way to intoduce && and || semantics in a course; based on ?? !!
ShimmerFairy I'm not well-versed in matters of security, so I expect most of what I wrote to appear nonsensical to someone who does :)
masak ShimmerFairy: two comments: 10:18
(a) the document starts out in a way that makes it seem that secutiry is not that important, and a thing that some people switch on in some cases. I think that's a dangerous attitude. 10:19
(b) there seems to be a widespread confusion in the whole document between "security" and "tainting".
ShimmerFairy With (b) , I considered an object's "security" and "taintedness" to be the same thing. 10:20
And as to (a) , I did get a "use strict; use warnings;" feeling from how I organized it, but I did that because I don't want the security in Perl 6 to be intrusive, esp. this late into it. 10:21
masak: I'm fully willing to admit that I know nothing about how to make something secure. I basically just read perlsec and went from there, so that there was some kind of starting point for S23. 10:22
masak ok. 10:23
well, ShimmerFairy++ for writing something. it's easier to iterate on things that exist :)
security is a very wide topic. tainting is one aspect and one solution to it. 10:24
ShimmerFairy It's certainly interesting trying to figure out how to add what is likely considered a fundamental part of the language when we're so close to a release, and there's so much existing code completely unaware of what you're trying to add :)
masak it's also a topic where it doesn't pay to be flippant in the least. want your software to be secure? well, then you have to get *every* step right. get one of them wrong and you get hacked. 10:25
tainting protects against some kinds of injection attacks. it's a pretty good tool for detecting the possibility of an injection attack. 10:26
but it doesn't address all the other possible attack vectors at all.
masak and a document purporting to talk about "security" can't talk only about tainting. 10:26
ShimmerFairy For tainting, I briefly considered having it be a ternary thing (secure, insecure, and indeterminate), but didn't go with that. However, I ended up having to a bit of contorting to fit within just two states. 10:27
masak: Yes. There's a very good reason I didn't try to sell this as a draft S23 in the slightest :)
masak :) 10:29
ditching "indeterminate" sounds good.
"tainted" tends to mean either "insecure" or "indeterminate" 10:30
that is, the onus is on the programmer to guarantee safety
r: class Conf { has @.boxes where { all(@$_) >= 0 } }; my $c = Conf.new(:boxes[ 2, 2, -1 ]); say $c.perl
camelia rakudo-{moar,jvm} 37b579: OUTPUT«Type check failed in assignment to '@!boxes'; expected '<anon>' but got 'Int'␤ in block <unit> at /tmp/tmpfile:1␤␤»
masak r: class Conf { has @.boxes where { all(@$_) >= 0 } }; my $c = Conf.new(:boxes[ 2, 2, 0 ]); $c.boxes[2] = -1; say $c.perl 10:31
camelia rakudo-{moar,jvm} 37b579: OUTPUT«Type check failed in assignment to '@!boxes'; expected '<anon>' but got 'Int'␤ in block <unit> at /tmp/tmpfile:1␤␤»
masak wow. I am impressed.
ShimmerFairy Yeah, perlsec made it very clear that it was the programmer's fault if the tainting got messed up. (Also, note how I didn't keep around that weird "hash keys are always safe" thing from P5)
masak except by the error message :)
masak well, tainting can indicate the presence of an injection, but not prove its absence 10:32
ShimmerFairy I think the error might be better written as "'-1' doesn't pass where constraint", or something to that effect
I can see now that some of my problems with writing my notes probably came from misunderstanding tainting as secure/insecure , instead of the insecure/not yet insecure it actually is :) 10:35
masak how do I answer the question "if I called a method on $obj (with these args), what class would it dispatch to?" in Perl 6? without actually making the call. 10:36
ShimmerFairy: yes, and maybe even print the where constraint. 10:37
m: subset Foo of Any where { $_ > 0 }; my Foo $n = -7
camelia rakudo-moar 37b579: OUTPUT«Type check failed in assignment to '$n'; expected 'Foo' but got 'Int'␤ in block <unit> at /tmp/46wGj6yVeD:1␤␤»
masak huh, I remember some error message actually printing the where constraint. apparently not that one. 10:38
ShimmerFairy (I definitely envision RESTRICTED as a second level of security, basically "I see all your nifty security features, Perl 6, but I need more. It's reeeally important.") 10:39
masak I think the document should be written from the point of view of "ok, you know that you might be hacked and that that would be really not good. so here's what we can do to help you avoid that." 10:42
I mean, reference leakage in objects could be considered to be part of this. 10:43
so could various types of information leakage, timing attacks, Denial of Service, etc
ShimmerFairy Sure. And I bet there are potential attacks on the underlying system that might need to be considered, if appropriate.
masak if you construct a regex from input, even if you do it safely without EVAL somehow, a crafty attacker could perhaps suppy you with an input that would create a regex that wouldn't finish before the heat death of the universe. 10:44
ShimmerFairy (And for some reason, I'm really interested in seeing something done about /<?>+/ and similar)
perhaps regexes need less-interpolating versions, like how strings have Q and :q 10:45
masak m: say ("0" x 3 ~ "1").substr(2)
camelia rakudo-moar 37b579: OUTPUT«00␤»
masak :/
masak should look into that one 10:46
rt.perl.org/Ticket/Display.html?id=123602, fwiw 10:47
ShimmerFairy m: say ("a" x 3 ~ "b").substr(2) # just to confirm it's not _possibly_ something weird about "0" left over 10:53
camelia rakudo-moar 37b579: OUTPUT«aa␤»
dalek ast: 00ea8c8 | (Stefan Seifert)++ | S06-signature/definite-return.t:
Fix typo in fudged test s/X::Adhoc/X::AdHoc/
 10:58
Woodi hallo #perl6 :) 11:11
pairs "constructors" like: a => 1 or :a(1) are very loudy when just: a: 1 can be used... or { a:1, b:2 } form 11:14
and such "enumerated" pairs could be put into perfect hashes (imutable) 11:15
nine Woodi: the colon is already being used for something else 11:24
masak yes, that's a label. 11:55
Woodi it's not ^':' here... 12:12
ShimmerFairy m: say (([~] "0" xx 3) ~ 1).substr(2) 12:24
camelia rakudo-moar 37b579: OUTPUT«00␤»
ShimmerFairy masak: ^ it affects list repetition too, it seems
(at least for strings) 12:26
m: say (0 x 3 ~ 1).substr(2); say (([~] 0 xx 3) ~ 1).substr(2); # also potentially helpful 12:27
camelia rakudo-moar 37b579: OUTPUT«00␤01␤»
lucasb m: my Int @a = 1,2,3; @a[1] = Nil; say @a.perl 13:34
camelia rakudo-moar 37b579: OUTPUT«Array[Int].new(1, Int, 3)␤»
lucasb m: my Int @a = 1,Nil,3; say @a.perl
camelia rakudo-moar 37b579: OUTPUT«Type check failed in assignment to '@a'; expected 'Int' but got 'Any'␤ in block <unit> at /tmp/Y6ejD8dOL8:1␤␤»
lucasb ^^ funny that one way works but not the other
RabidGravy yay! noise! It helps with streaming not to be sending one byte at a time 13:41
RabidGravy so for reference it is actually quicker copying 4096 bytes of a 320k mp3 file to a CArray[uint8] than it is to stream it, which is fortunate ;-) 13:49
awwaiid Whatcya doin, RabidGravy? 14:16
smls m: my @array; $x := @array[0]; $x = 0; @array[0] = 42; say $ 14:18
camelia rakudo-moar 37b579: OUTPUT«5===SORRY!5=== Error while compiling /tmp/vEGXaQOHnK␤Variable '$x' is not declared␤at /tmp/vEGXaQOHnK:1␤------> 3my @array; 7⏏5$x := @array[0]; $x = 0; @array[0] = 42;␤»
smls m: my @array; my $x := @array[0]; $x = 0; @array[0] = 42; say $x
camelia rakudo-moar 37b579: OUTPUT«42␤»
smls m: my @array; my $x := @array[0]; @array[0] = 42; say $x
camelia rakudo-moar 37b579: OUTPUT«(Any)␤»
smls ^^ should that print 42 too?
RabidGravy awwaiid, binding to libshout to stream audio to icecast
timotimo cool :) 14:21
lucasb m: my @a; my $x := @a[10]; $x = 1; say @a.elems 14:23
camelia rakudo-moar 37b579: OUTPUT«11␤»
lucasb ^^ This vivifies the elements in the array without touching it
smls: In your second example, it didn't get vivified, right? 14:24
smls looks like it 14:25
lucasb m: say ('abc' x 2 ~ '1234').substr(3) 14:27
camelia rakudo-moar 37b579: OUTPUT«abc1234␤»
lucasb m: say ('abc' x 2 ~ '1234').substr(4)
camelia rakudo-moar 37b579: OUTPUT«bcabc1␤»
lucasb ^^ That bug is an "off by one"? :)
lucasb I said 'abc' x 2, but it is like 'abc' x 3 14:29
It chops the number 1234 at 1 maybe because of the length of the original string 14:31
j: say ('abc' x 2 ~ '1234').substr(4) 14:33
camelia rakudo-jvm 37b579: OUTPUT«bc1234␤»
lucasb Is it a moarvm thing then? Maybe about the implementation of nqp::x
DrForr .seen Util 14:47
yoleaux I saw Util 17 Jun 2015 19:12Z in #perl6: <Util> smls: That is the GLR bug. As for re-writing as [\*] 1, @primes, ... DOH!
timotimo m: my @test; @test.unshift; say @test 15:16
camelia rakudo-moar 37b579: OUTPUT«␤»
timotimo m: my @test; @test.unshift; say @test.perl
camelia rakudo-moar 37b579: OUTPUT«[]<>␤»
timotimo should this throw an error?
smls Perl 5 allowes it, but prints a "Useless use of unshift with no values" warning 15:19
AlexDaniel is there any way to make the current rule not match? I mean, if I put { fail } it will jump out without matching anything at all, but I just want to say "ok, this regex is not going to match, just jump out of here but keep trying". Is it one of these : :: ::: backtracking controls?
smls timotimo: I Perl 6 I guess it's a consequence of unshift's (\a, *@elems) signature 15:20
timotimo but if it's \a, *@elems should make that work, shouldn't it? 15:21
i mean: give an exception properly
smls no, slurpies params also match the zero argument case 15:21
timotimo yeah, but \a won.t 15:22
oh, are you talking about the sub form?
smls yeah
smls AlexDaniel: <!> 15:25
AlexDaniel smls: what's that? 15:26
smls m: say "aaa" ~~ /aaa <!> | aa/
camelia rakudo-moar 37b579: OUTPUT«｢aa｣␤»
AlexDaniel smls: indeed, this works 15:27
smls: but it would be cool to know how this thing is called 15:29
smls: and whether there are docs for that or not
timotimo: by the way, it seems like it is a whole class of problems when calling stuff without parameters produces no errors 15:30
timotimo: I remember seeing several rts like this one
smls AlexDaniel: design.perl6.org/S05.html#line_2150 15:33
It doesn't seem to be documented on doc.perl6.org yet.
timotimo anything with slurpy-only params
smls timotimo: Actually, Perl 5 only prints the warning when the argument is missing in the source code. 15:35
smls If there's an array argument that just happens to be empty at run-time, there's no warning. 15:35
timotimo ah, mhm 15:36
smls I believe lizmat++ has previously put some thought into how we could handle such cases (argument required, but allowed to be empty) in Perl 6
timotimo well, we can't really do that unless we already know at compile time that it's the right class etc etc
smls not sure if anything came of it though
right, for methods I guess it wouldn't be feasible 15:37
but for subs it could be made to work, no? 15:38
timotimo we can just put a check into these methods, but ... :(
spesh could eliminate these checks
smls I do think @a.push(@b) should work even if @b is empty 15:39
timotimo but maybe it'd be better to have it in the signature directly somehow
smls to avoid always having to check that case in user code
timotimo you mean @a.push(|@b)
@a.push(@b) wouldn't trigger the check for "no arguments passed"
smls ah, you mean the check would kick in before flattening into the slurpy? 15:40
AlexDaniel smls: rt.perl.org/Public/Bug/Display.html?id=125257
there is a comment by liz
timotimo i think you're misunderstanding how *@a in a signature works
m: sub test(*@a) { say @a.elems }; my @foo = 1, 2, 3, 4; test(@foo) 15:41
camelia rakudo-moar 37b579: OUTPUT«4␤»
AlexDaniel and a fix, actually
timotimo .. huh?
m: sub test(*@a) { say @a.elems }; my @foo = 1, 2, 3, 4; test(@foo, @foo)
camelia rakudo-moar 37b579: OUTPUT«8␤»
timotimo i didn't know that flattens
smls slurpies and list assignment are the last remaining bastions of auto-flattening in Perl 6 :P 15:45
and maybe the list assignment one will fall too, in the course of the GLR... 15:46
m: sub test(**@a) { dd @a }; my @foo = 1, 2, 3, 4; test @foo, @foo; 15:47
camelia rakudo-moar 37b579: OUTPUT«(1, 2, 3, 4; 1, 2, 3, 4)␤»
smls ^^ though there's also ** slurpies
skids slurpies are also special in that they can receive feeds (once feeds get up to snuff)
timotimo i think that stumped me already 15:49
the thing about *@a flattening @-passed things 15:50
AlexDaniel smls: oh noes! <!> does not really work in my case, sorry for saying that it worked 15:52
wrong input file
smls: <!> just always fails to match, I, on the other hand, want to stop matching this rule completely 15:53
ugexe <!.> ? 15:54
AlexDaniel I have something like this: regex rule { <s>.*? [ < }
ooops
nvm that line 15:55
this: regex rule { <aaa> [ <bbb> || <stopTrying> <!> ] } 15:56
in other words, if you stumble upon <stopTrying> just eject from this regex
ugexe <aaa> [<bbb> || { die }] hehe
nine { return } ? 15:57
AlexDaniel ugexe: I've already mentioned { fail } but if this part of the grammar did not match it does not really mean that the whole thing cannot match 15:58
skids Should just have to return whatever a failure to match returns.
timotimo you'd return a Cursor i believe 15:59
smls AlexDaniel: what do you mean by "just eject from this regex"? 16:00
timotimo make the current regex { ... } fail
with <!> you can just make a branch fail
skids Nil
timotimo if you have a | or ||, the rest still does backtracking
skids: i don't think that's compatible in this place 16:01
ugexe eh, does it still backtrack if <this> || <that> this matches before that?
timotimo in that case it'll try the next piece of the branch. if there's a scan around (actually, before) that, it'll try the whole alteration again
gotta go get groceries and stuff
skids 'The method is expected to return a lazy list of new match state objects, or Nil if the match fails entirely.' That is assuming there isn't GLR churn.
smls regex rule { <aaa> <!before <stopTrying> > <bbb> } 16:02
AlexDaniel: ^^ maybe refactor the regex like this?
timotimo which method are you talking about?
skids Search for that in S05
timotimo does the "return" really travel up to the match method? 16:03
or parse or whatever?
oh
i'm apparently mistaken :)
okay :)
AlexDaniel smls: no-no, because "some text here STOPTRYING some text here STOPTRYING", there is no guarantee that it will stop after the first occurance
AlexDaniel or hm... 16:04
I'm confused actually
ugexe i dont see why it wouldnt stop
smls I may not understand correctly what you want :P
ugexe unless you are using a greedy operator somewhere 16:05
AlexDaniel yeah, well, maybe I should come up with an example
skids m: grammar foo { regex TOP { <foo> d || bd }; regex foo { a || b { return Nil } || c }; }; foo.parse("bd").say; # yeah return Nil doesn't work. 16:09
camelia rakudo-moar 37b579: OUTPUT«Nil␤»
skids Hmm.
skids m: grammar foo { regex TOP { <foo> d || bd }; regex foo { a || b { return Empty } || c }; }; foo.parse("bd").say; # It is returnin from the whole parse. 16:09
camelia rakudo-moar 37b579: OUTPUT«Empty␤»
ugexe figured. 16:10
skids I don't think it should.
ugexe ive emitted values out of a parse when i needed to get values out 16:11
AlexDaniel skids: while I'm coming up with an example, your code totally makes sense to me
skids: that's exactly what I want, if only it worked 16:12
ugexe replace return Empty with an impossible to match token and it should do the same thing, no?
skids m: grammar foo { regex TOP { <foo> d || bd }; regex foo { a || b { &?ROUTINE.leave } || c }; }; foo.parse("bd").say # Hrmf 16:13
camelia rakudo-moar 37b579: OUTPUT«Regex.leave() not yet implemented. Sorry. ␤ in regex foo at /tmp/oS2Ph07dTO:1␤ in regex TOP at /tmp/oS2Ph07dTO:1␤ in block <unit> at /tmp/oS2Ph07dTO:1␤␤»
skids ugexe: no. 16:14
nine What's the ? in self.?message 16:16
skids nine: "if the method exists" IIRC. 16:17
smls I believe it means self.^can("message") ?? self.message !! self
TimToady we usually use <!> for the "never matches" assertion
masak m: class C {}; C.new.?message; say "alive"
camelia rakudo-moar 37b579: OUTPUT«alive␤»
TimToady but methods in a grammar must always return cursors 16:18
ugexe ?method() is like method() // Nil
masak no, not really.
ugexe so if its a false value it also returns false
smls yay, TimToady is online, I can step back from giving (inferior) grammar advice now :P
skids m: grammar foo { regex TOP { <foo> d || bd }; regex foo { a || b { return foo } || c }; }; foo.parse("bd").say
camelia rakudo-moar 37b579: OUTPUT«(foo)␤»
TimToady well, yer assumin' I'm awake
skids m: grammar foo { regex TOP { <foo> d || bd }; regex foo { a || b { return self } || c }; }; foo.parse("bd").say 16:19
camelia rakudo-moar 37b579: OUTPUT«foo.new␤»
masak ugexe: .method() // Nil would still die if there's no .method()
nwc10 and caffeinated
ugexe ah, true 16:19
masak m: class C { method message(C:D:) { say "only an instance method" } }; C.?message; say "alive"
camelia rakudo-moar 37b579: OUTPUT«Invocant requires a 'C' instance, but a type object was passed. Did you forget a .new?␤ in method message at /tmp/MuAO2imCSt:1␤ in block <unit> at /tmp/MuAO2imCSt:1␤␤»
masak m: class C { multi method message(C:D:) { say "only an instance method" } }; C.?message; say "alive" 16:20
camelia rakudo-moar 37b579: OUTPUT«Cannot call message(C: ); none of these signatures match:␤ (C:D $: *%_)␤ in block <unit> at /tmp/Blgeb2qBAD:1␤␤»
masak ...why does that last one fail?
isn't the point of .?method to avoid dispatch failures? even multi dispatch failures?
TimToady currently it's only "missing method"
masak sheesh, I'm back from a nice afternoon errand for 3 minutes, and I find a rakudobug :P 16:21
TimToady this was discussed last week or so
masak TimToady: ok, so the above is actually correct behavior?
TimToady it doesn't catch bind failures as it is currently implemented 16:21
masak right, I see that.
...but do we want it to?
TimToady that's more in the bailiwick of try currently
masak stands down bug alert 16:22
nine Or the real question: why does calling a P5 function during stringification of the P5 $error object in a P6 die($error) screw up $error in a "Can't use anonymous symbol table for method lookup." kind of way?
skids m: grammar foo { regex TOP { <foo> d || bcd }; regex foo { a || bc <!> || b+c }; }; foo.parse("bcd").say # better test case ugexe++ 16:23
camelia rakudo-moar 37b579: OUTPUT«｢bcd｣␤ foo => ｢bc｣␤»
TimToady we haven't decided if we want .? to do more yet; there's arguments on both sides, of which I am not awake enough to reproduce
smls nine: Sounds like a P5 error message to me, not one from P6 16:24
nine smls: it is
nine somehow the P5 object is just broken after the stringification 16:24
But only if this stringification is done in the course of throwing the object with die() in P6 16:25
This is the cause of S01-perl-5-integration/exception_handling.t failing btw. 16:26
AlexDaniel So, here is the illustration of my problem: gist.github.com/AlexDaniel/5fde506394d862f23626 16:27
<!> does not help
in other words, 'special' should match ** anything ** unless XX is inside 16:28
AlexDaniel in this case 'special' should just fail, but the parsing should continue 16:28
smls '**' (.*?) '**' <?{ $0 !~~ /XX/ }> 16:31
'**' [ <!before XX> . ]*? '**'
TimToady that would need the NYI backtracking control to work, ::: <!>
AlexDaniel TimToady: that's what I said initially
nine AlexDaniel: now you have a good reason to implement backtracking control :) 16:32
TimToady '**' ~ '**' [ <!before XX> . ]*?
AlexDaniel nine: I don't think that I'm ready for this 16:33
ugexe no one is going to try and sneak a XX+ % '**' in there?
TimToady having looked at the current backtracking mechanism, I'm not sure I'm ready for it :) 16:34
nine AlexDaniel: one truth I've learned so far: you're never ready. You just jump in and do it anyway :)
TimToady implemented ::: in STD, but STD uses a completely different backtracking mechanism 16:35
AlexDaniel nine: yea, well. But that requires time
ugexe kill them all and let the implementors sort them out
AlexDaniel smls: your first line actually makes sense, but as far as I can see it runs in exponential time 16:38
correct me if I'm wrong 16:39
hoelzro morning, #perl6!
AlexDaniel smls: ooooooooooooooooh 16:42
oooooooh
hoelzro m: multi foo(int $n) { say 'here' } ; foo(3) 16:44
camelia rakudo-moar 37b579: OUTPUT«Cannot call foo(Int); none of these signatures match:␤ (int $n)␤ in block <unit> at /tmp/nHJCaCYxb2:1␤␤»
AlexDaniel I've tried this yesterday: regex special { '**' [ . <!before XX> ]*? '**' }
hoelzro should an int parameter never be considered for an Int literal when finding candidates?
AlexDaniel but then figured out that it will fail on this string: 'test **XX test** test' 16:45
and somehow stopped exploring that
now, smls, you are a genius!
just put before before . and there you go
as far as I can see, this should work 16:46
ugexe i was gonna say... this doesnt do what you think [ . <!before XX> ]*?
TimToady wonders if it would be useful to have ! and - prefix shortcuts, so you could just say -'XX' to mean <!before XX> . 16:50
smls TimToady: But maybe people would expect that to gobble 2 character 16:51
TimToady and <-[x]> could be -x
arnsholt Morpheus: I plugged my Occulus in again now that I have a new graphics card, and I have to say that after a little while, even the generous 120 degree viewing angle feels kind of tight and uncomfortable. I have no idea how Morpheus is going to get by with just 90.
Calling of Various Duties: Really not my thing. No comment.
Whoops.
Pasto >.<
AlexDaniel just tested that in my original code, it works! Wooohooo 16:52
smls m: say "** abc XX ** def **" ~~ /^ "**" (.*?) "**" <?{ say $0; $0 !~~ /XX/ }>/ 16:54
camelia rakudo-moar 37b579: OUTPUT«｢ abc XX ｣␤｢ abc XX ** def ｣␤Nil␤»
ugexe and we almost had him ready to implement backtracking features... darn
TimToady hoelzro: seems like that could be a bit dwimmier, but we'd have to interrogate jnthn++ as to why it works the way it does
smls AlexDaniel: I don't think the one with the explicit assertion ^^ would be exponential time, because it only checks the assertion after matching ** 16:55
however, it *would* needlessly try to match beyond the first **
hoelzro TimToady: alright, I'll await his response then
smls that's the problem with *?
(if you're not careful with it)
so yeah, forget that suggestion.
Glad to hear the one with <!before> works. 16:56
AlexDaniel smls: oh, indeed, that's not exponential time 16:58
TimToady one might be able to speed up by matching <-[X*]>+ instead of . 17:04
well, you'd still have to traverse a single X somehow too
but the <!before> solution is not exponential, it just has a larger constant on the O(n) 17:05
AlexDaniel TimToady: sure, I just thought that <?{ }> check will fire up every character, but that's not the case 17:06
smls TimToady: regarding speed, do you think .trans will beat .match here after both are optimized? github.com/dnmfarrell/URI-Encode/issues/4
Or am I correct in my assumption that something <[\x00..\xff]> in a regex becomes and efficient 0 < ord($current) < 255 check which an .trans won't be able to compete with? 17:09
*something like
TimToady hard to say, since the trans is depending on mutable arrays, and we don't really have a good mechanism for telling when one of those has changed
TimToady it also kinda depends on the frequency of the weird characters 17:14
smls It does? 17:15
TimToady especially since you're doing an expansion
the subst is gonna be really good at skipping the N characters that don't change 17:16
smls m: say "N = ", ("a".."z", "A".."Z", 0..9, "_", ".", "~", "-").flat.elems 17:17
camelia rakudo-moar 37b579: OUTPUT«N = 66␤»
TimToady whereas trans tends to look up every character, even the ones that don't change, and gluing together every char separately
I mean N in a row
smls ah
hoelzro could someone have a look at github.com/perl6-community-modules...ts/pull/2? it adds a warnings category for modules, so people can see which distributions are currently emitting warnings, so we can fix them before those warnings become errors 17:18
TimToady as a degenerate case, the subst just does one string copy if no escaping happens
smls So is there any case where the .trans version might be faster?
TimToady caching issues aside, it really depends on how smart .trans gets about skipping unchanging characters 17:22
if it's trying to do LTM on a set of chars that is all length 1, that's kind of a waste of energy
smls what could it do that would be smarter than what the regex version does? 17:23
TimToady if it's doing one char at a time concatenation, that's gonna be way slow too
well, precomputed results are a lookup rather than a .fmt, which could conceivably be faster 17:24
smls one could replace the RHS part of the .subst with a hash lookup then :P
TimToady indeed 17:25
knowing the max replacement length lets you put an upper bound on how much result string you have to allocate, so you never have to copy partial results to a larger location 17:26
if you're willing to trade wasted memory for speed
(a trade P5 makes with some frequency)
smls good point
TimToady that's assuming contiguous strings, of course; since we have a partial ropes implementation, that may not hold 17:27
smls do you think future Rakudo will cheat and use mutatable strings internally wherever it can get away with it? 17:30
TimToady smls: doesn't help much when you're expanding in place; you might as well copy 17:32
smls true
TimToady for old-style tr/a..z/A..Z/, sure, we could cheat, given some static analysis 17:33
hoelzro .seen colomon
yoleaux I saw colomon 3 Jul 2015 14:43Z in #perl6: <colomon> life’s too short to change working code unnecessarily.
TimToady of course, that should just be a tcuc() or so, to be unicodically correct 17:33
hoelzro .tell colomon could you have a look at github.com/perl6-community-modules...ts/pull/2? 17:34
yoleaux hoelzro: I'll pass your message to colomon.
hoelzro are actions classes (as in what one provides to Grammar.parse) spec'd anywhere? I only see a brief mention of them in design.perl6.org/S05.html#line_4383 17:43
smls hoelzro: design.perl6.org/S32/Rules.html#Grammar also mentions them 17:45
hoelzro smls: true, but I don't know if what actions classes are/how they work are spec'd 17:46
TimToady well, hasn't been much need for docs since they're drop-dead simple; it just calls the same method name that just got reduced 17:47
smls "reduced"?
smls oh, you're talking parser theory lingo 17:48
TimToady successfully matched
hoelzro I was thinking of the confusion of a new user reading the specs to learn, trying to figure out what an action class is without having to ask #perl6
maybe it's my failure in assuming people will be learning Perl 6 from the design docs
smls they should read doc.perl6.org/language/grammars#Action_Objects then :) 17:49
TimToady one can put { make } directly into the rule, but then you can have only one set of actions for a given grammar
this gives the ability to bind different sets of actions to the same grammar, such as syntax highlighting
or linting 17:50
smls in theory
TimToady smls: darn it, you're spoilig our rant :P 17:51
smls ;)
TimToady one of the optimizations we'd like to attempt is to avoid doing the 2nd method dispatch every time 17:53
one could imagine binding a particular set of actions to a grammar, such that there is just one method call that basically mutates the original rule into [ <originalrule> ] { make stuff } 17:55
well, that wasn't good English, but something like that
smls might as well pass a hash instead of an object instance to .parse then :P 17:57
hoelzro should I not bothering spec'ing actions then? 17:59
TimToady I don't see any problem with speccing it; the optimization would be relatively transparent, mehopes 18:00
my researches last night indicate that it's trivially easy to make [] not itemize, it's relatively difficult to get list assignment not to flatten, and nearly impossible to keep [] from flattening itself before list assignment even gets ahold of it (given current nqp list semantics)
hoelzro TimToady: action methods still fire regardless of backtracking and lookahead, right? 18:01
the reason I'm thinking about this is that I ran into some (to me) odd behavior regarding action methods and a lookahead 18:02
TimToady they fire whenever the rule in question returns; it's up to the action methods to attach info only to the cursor using 'make' to keep the bookkeeping straight
any side effects outside of the current cursor can go haywire
hoelzro ah ha 18:03
make sense() # to me 18:04
dalek line-Perl5: c8e0bc5 | (Stefan Seifert)++ | p5helper.c:
Fix method calls on P5 exception objects

When we started supporting stringification of Perl 5 objects, S01-perl-5-integration/exception_handling.t started failing.
The reason is that the ERRSV pointer never changes. Instead, the SV it points to gets overwritten with the exception object. As a result, we cannot just wrap the ERRSV pointer in a Perl5Object, since this only works until Perl 5 resets $@. From that point on all method calls will fail.
Fixed by creating a copy of the SV before converting it to a Perl 6 scalar.		 18:14
nine S01-perl-5-integration/exception_handling.t fixed :)
And it was only a day's worth of work for a one line fix ;)
TimToady \o/ /o\ 18:15
TimToady nine++ anyway 18:15
that's for the bug fix, and another nine++ for the perserverance
smls_ Is the return value of action methods used for anything? 18:20
TimToady no
just as the { make stuff } doesn't influence the return cursor of the rule
smls_ right 18:21
masak m: sub foo(int $i) { say "OH HAI" }; foo(3) 18:30
camelia rakudo-moar 37b579: OUTPUT«OH HAI␤»
masak m: multi foo(int $i) { say "OH HAI" }; foo(3)
camelia rakudo-moar 37b579: OUTPUT«Cannot call foo(Int); none of these signatures match:␤ (int $i)␤ in block <unit> at /tmp/QDBQhyo6N9:1␤␤»
masak oh. hadn't realized that there's this difference before.
I'm sure there's a good reason... :)
timotimo huh 18:31
TimToady m: multi foo(int(Int) $i) { say "OH HAI" }; foo(3)
camelia rakudo-moar 37b579: OUTPUT«Method 'int' not found for invocant of class 'Int'␤ in sub foo at /tmp/9SkuqD30Vr:1␤ in block <unit> at /tmp/9SkuqD30Vr:1␤␤»
TimToady heh
masak something fishy there, 'tseems
ooh, and int(Int) ought to work with both int and Int inputs... 18:32
vendethiel waves
masak m: sub foo(int(Int) $i) { say $i }; my Int $x = 42; foo($x); my int $y = 43; foo($y)
camelia rakudo-moar 37b579: OUTPUT«Method 'int' not found for invocant of class 'Int'␤ in sub foo at /tmp/xTvgusbUqw:1␤ in block <unit> at /tmp/xTvgusbUqw:1␤␤»
masak hurh 18:33
smls_ masak: As I've mentioned before, coercion subs/methods are a kind of a mess right now in Rakudo, and probably need more design work too
masak vendethiel: kon'nichiwa~~~
vendethiel masak: hehehehe :) 18:34
masak smls_: ok.
smls_ I just hope that lizmat/TimToady/etc will find the time to think about that stuff before 6.0
masak the most important thing is not to miss chances for improvement that are very hard or impossible after 6.0.0 18:36
not everything has to be perfect by 6.0.0, but it should preferably be improvable :)
smls_ +1 to that
BenGoldberg m: my int $x = 3; 18:37
camelia ( no output )
BenGoldberg m: my int $x = 3.int; 18:38
camelia rakudo-moar 37b579: OUTPUT«Method 'int' not found for invocant of class 'Int'␤ in block <unit> at /tmp/neOo7nA6Xk:1␤␤»
BenGoldberg m: my $three = 3; my int $x = $three;
camelia ( no output )
BenGoldberg m: my $three = 3; my int $x = $three; say $x;
camelia rakudo-moar 37b579: OUTPUT«3␤»
BenGoldberg m: my $three = 3; my int $x = $three; say $x.WHAT;
camelia rakudo-moar 37b579: OUTPUT«(Int)␤»
lucasb m: multi f(int) { 'int' }; multi f(Int) { 'Int' }; say f(42) 18:43
camelia rakudo-moar 37b579: OUTPUT«int␤»
lucasb m: multi f(int) { say 'int' }; multi f(Int) { say 'Int' }; f(42)
camelia rakudo-moar 37b579: OUTPUT«Int␤»
lucasb My conclusion: don't "say" anything :) 18:44
dalek ecs: beb870d | hoelzro++ | S05-regex.pod:
Spec action objects/methods
 18:48
ecs: e4f7917 | hoelzro++ | S99-glossary.pod:
Link to action objects section in glossary
CIAvash this eats up CPU and memory: 19:12
class A {}; class B {}; A.^find_method('new').wrap(B.^find_method('new')); A.new;
ugexe does wrapping methods really work? 19:24
TimToady wrapping in general only works if your wrapper invokes nextsame or callsame 19:27
CIAvash m: class A {}; class B {}; A.^find_method('new').wrap(method {say 'hi'}); A.new;
camelia rakudo-moar 37b579: OUTPUT«hi␤»
TimToady you can't just wrap an arbitrary routine around another without it knowing when it should call the wrapped routine
I'd also point out that A and B actually have the same new method 19:29
m: class A {}; class B {}; A.^find_method('new').wrap(method {say 'hi'}); B.new; 19:30
camelia rakudo-moar 37b579: OUTPUT«hi␤»
TimToady see the B.new htere
*th
TimToady you're actually wrapping the generic new method that comes from Mu, probably 19:30
CIAvash my main issue was that the command was not stopped, when it was eating memory 19:31
TimToady and since you're not calling into the actual new method, you've screwed up all object creation
in this case, you've wrapped the new routine in itself 19:32
vendethiel m: loop (;;) {}; #not stopped :P 19:33
camelia rakudo-moar 37b579: OUTPUT«(timeout)»
CIAvash :) yeah, didn't know both new methods are the same
TimToady m: class A {}; class B {}; A.^find_method('new') === B.^find_method('new') 19:34
camelia ( no output )
TimToady m: class A {}; class B {}; say A.^find_method('new') === B.^find_method('new')
camelia rakudo-moar 37b579: OUTPUT«True␤»
dalek ecs: 30aa605 | hoelzro++ | S99-glossary.pod:
Document (camel|kebab|snake) case naming conventions
TimToady anyway, I think this comes out to a DIHWIDT 19:35
CIAvash yes, I was just playing with wrap when this happend 19:38
TimToady
.oO(Power tools, power tools, it's my favorite thing to use...Hello, emergency?)
 19:39
hoelzro what's the idiom for making some sort of cleanup happen in a deterministic fashion? ex. try(...) in Java, using(...) in C#, with(...) in Python. is it my $var will leave *.cleanup?
vendethiel hoelzro: LEAVE? 19:40
or yes, will leave
TimToady 'will leave' is a variant of LEAVE
hoelzro vendethiel: alright, that's what I thought
TimToady or if you only need to cleanup on failure, 'will undo' 19:41
hoelzro .tell jnthn could you weigh in on irclog.perlgeek.de/perl6/2015-07-04#i_10848766 if you have some time? 19:45
yoleaux hoelzro: I'll pass your message to jnthn.
hoelzro I'd like to know if I should rakudobug that
lucasb When a plain literal appears in the text like 42 or "abc", they are boxed (Int,Str) or native (int,str)? 19:55
TimToady we'd like to get to the point where literals can be allomorphic wrt boxing 19:58
well, some of us would, anyway :)
lucasb In 'multi f(int) {...}; f(42)', if the block contains just a simple expression, then it works. If it has statements, then it acts as if the 'int' has been upgraded to an 'Int', and then the dispatch fails. 20:00
TimToady but if you're using the distinction for multi-method dispatch, you can't simultaneously not use the distinction
smls_ if .WHAT boxes int -> Int, how can we introspect native types? 20:38
timotimo nqp::objprimspec :)
may have to decont, though
smls_ heh
timotimo hmm 20:47
how to best switch two @-sigiled vars around?
oh, i got something that works; the problem i was seeing came from somewhere else entirely 20:48
smls_ m: my @a = 1, 3; my @b = 2, 4; :(@a, @b) := @b, @a; dd @a, @b 21:02
camelia rakudo-moar 37b579: OUTPUT«@b = [2, 4]<>␤@a = [1, 3]<>␤»
smls_ nice
though dd still prints the old names :P
timotimo :) 21:13
RabidGravy tum ti tum 21:14
timotimo still no clue what in my script is leaking so much memory :( 21:20
dalek ast: cf2e631 | lizmat++ | S32-hash/adverbs.t:
More comprehensive testing of hash slicing adverbs
 21:28
masak realizes that he can define a mocked thing as a role with a parameter, so he can share state across that parameter 21:41
masak really sweet way to write tests :> 21:41
and one of the few cases of intentional/beneficial reference leakage
timotimo hah, kinda cute idea 21:42
smls_ doesn't get it 21:43
.oO( Maybe if masak were to write a blog about it... )
 21:44
;)
masak ooh, now I came up with a cuter way to do it :>
masak smls_: let me tell you real quickly, even though I now have a better way to do it. 21:44
smls: so, in the implementation there is a class called Performer. it receives events by having the .perform(Event) method called on it.
smls: in the tests I want to check that I got the right events. 21:45
so I created a `role Mock::Performer[@events] { method perform($e) { @events.push($e) } }`
and then in the tests I could just do `my $performer = Mock::Performer[my @events];` and then inject $performer into the system under test. 21:46
smls why not class Mock::Performer { has @events; ... }
timotimo not fond of "has $foo" 21:47
masak that was my better idea :)
timotimo without the dot or exclamation mark, that is
masak: but then suddenly your Performer mock has a different API! :) 21:48
masak I just need to put the test method on the mocked performer, too. then I can completely encapsulate @events
smls timotimo: ah, right, I meant with dot
timotimo mhm
masak timotimo: that's not a problem, as long as it fulfills the contract of a Performer
timotimo i know that you can do it without the ., but i forgot what it does in that case
masak Liskov cuts but one way
I also mock time in these test, which I always love doing. :> 21:49
timotimo liskov's razor is commonly used for organ transplants
masak $time.wait(5, minutes);
japhb is busy trying to put an r/w attribute proxy API in front of an event-sourced CQS engine, just to see what that looks like 21:57
masak sounds... weird. :) 21:58
one of the big points of CQS is that commands and queries carry more intent than just modifications to attributes.
japhb Basically, get *almost* back to the original mutable OO API, while in fact underneath it's all event sourcing and domain processing and such.
masak so a command can be .fix_address or .change_address and they would both modify an $!address attribute somewhere, but the commands would have very different intents. 22:00
timotimo $time.wat(5, minutes)
japhb masak: As I'm playing around with it, I'm seeing a couple things: 1) It works well for scalar-valued attributes; not so much for anything more complex. 2) It seems useful for the proxy to simply *add* the r/w scalar attribute API, not replace the underlying CQS methods. 22:03
masak japhb: hm. makes sense, I think. 22:04
japhb Anyway, mostly a matter of figuring out what a combined API even looks like, and if it ends up being useful in real code, as opposed to merely in theory.
masak *nod*
atweiden deep copying Hash[Array[Objs]] is proving very difficult: ix.io/jtA 22:40
masak 'night, #perl6 22:41
atweiden is there any similar code to look at? docs are a bit sparse 22:43
smls did you try .deepmap(*.clone) ? 22:47
atweiden smls: solved 22:59
with deepmap!
tyty
RabidGravy japhb, I made AccessorFacade for the simpple case on a native api 23:00

Powered by Cro and the Raku® Programming Language.

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

Thank you!