»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.
Zoffix Sounds scary :) 00:00
timotimo lucasb: maybe the combinations method should flat-out throw an exception if it gets an Index instance? :P 00:02
TimToady m: say <a b c>.combinations("0").perl 00:03
camelia rakudo-moar a38b59: OUTPUT«Cannot call combinations(Str: Str, Str, Str); none of these signatures match:␤ (List $: Int $of, *%_)␤ (List $: Range $ofrange = { ... }, *%_)␤ in block <unit> at /tmp/JxchiHTiyJ:1␤␤»
timotimo as it makes no sense to get as many items in the combinations as the position of something in some string is ...
lucasb In the specific case of .combinations, the simple fix would be a line like "return ((),) if $k == 0"
timotimo: I know my example was contrived :) 00:04
but maybe there are other places that accepts a Int and tests it with 'if $n' or 'unless $n' ... 00:05
Sgeo Maybe 0 should just be truthy
TimToady I was leaning toward ifdef/whiledef until I realized whiledef doesn't help much, since whiledef "abc".index('a',$pos) -> $pos {...} doesn't actually work
Sgeo Or maybe there should be an if-defined function that passes to a pointy block the result
timotimo Sgeo: not in general, no
if-defined doesn't help, i don't think 00:06
er, wait, yeah it does
time for me to get some rest
b2gills If combinations coerced to Int it would work as is
TimToady ifdef index "abc",index('a') -> $pos {...} is okay
er
ifdef "abc".index('a') -> $pos {...} is okay 00:07
Sgeo Maybe if could be flexible in some way that allows both transforming into something for a truthy test and keeping the original
TimToady since you're not trying to feed $pos back into the index
b2gills m: say "abc".index('a').Int.^name
camelia rakudo-moar a38b59: OUTPUT«Int␤»
timotimo schwartzian conditional 00:08
Sgeo Actually sounds a bit like when if it wasn't too overly elaborate to do the given or $_ assign
b2gills m: my $i = "abc".index('a'); say so $i; say so $i.Int
camelia rakudo-moar a38b59: OUTPUT«True␤False␤»
TimToady really, we should just have a version of index that returns all the entries as a list, and not try to use explicit pos for that 00:09
then it's just: for "abc".indices('a') {...} 00:10
dha So. If I were to change the -M section in 5to6.pod, should I do it directly to the main repo, or fork and put in a pull request for someone else to review and change?
TimToady forgiveness > permission 00:11
00:11 laouji joined
dha All right, then. :-) 00:12
TimToady the only downside is you have to be prepared to be forgiven :)
timotimo good night everybody! 00:14
dha night timotimo
00:16 mohij left
tony-o how do we get blog posts up in p6 weekly/? 00:16
flussence step 1: get noticed, step 2: ??? 00:19
tony-o cool 00:20
super cool
flussence that's about it, I've been mentioned there without any interaction 00:21
dha step 3: Profit!
ShimmerFairy TimToady: I think I like the sound of indices, and testing its size to see if anything's found (and maaaaybe get rid of index in favor of specifying from which direction with first-/last-index ?)
b2gills There is a question about "How to get start with Perl6?" on StackOverflow if someone wants to add anything stackoverflow.com/q/31791362/1337 No account necessary, unless you want those sweet rep points
00:27 BenGoldberg joined
TimToady ShimmerFairy: I run with t/spec linked into roast 00:28
works fine 00:29
ShimmerFairy TimToady: I set that up, I just have to be careful about git clean -dfx now :)
TimToady yeah, I never do that 00:30
TimToady leaves too much tmp stuff sitting around for that
ShimmerFairy btw, it seems first-index is broken: 00:33
m: say "abc".first-index("a"); say "abc".index("a")
camelia rakudo-moar a38b59: OUTPUT«Nil␤0␤»
dha I have a note here that you can still "no" a pragma in Perl 6, but not a module. Is that, in fact, correct?
TimToady yes, we generally rely on lexical scoping for limiting the scope now 00:34
since imports are lexical
dha *nod* thanks.
dalek c: 8c725bf | (David H. Adler)++ | lib/Language/5to6.pod:
Changed 5to6.pod to reflect the actual status of -M/-m
 00:36
ShimmerFairy I almost want pragmas to have a different syntax from modules, unless pragmas are supposed to be vaguely module-esque (as I understand it, currently pragmas just flip switches in P6's grammar)
00:36 lucasb left
b2gills pragmas in Perl 5 are modules 00:37
skids wonders if a camelia directive could be spun up with jnthn++s GLR code loaded.
ShimmerFairy
.oO( no warnings; ... { yes warnings; ... } )
00:39 laouji left 00:40 laouji joined
ShimmerFairy Unless Perl 6 is still planned to allow pragmas written by external sources in the form of modules, I think it'd be nice if pragmas looked a bit different from modules in usage. 00:40
00:42 laouji left, laouji joined
Sgeo m: if 0 -> $x { say "o.O"; } else { say $x; } 00:51
camelia rakudo-moar a38b59: OUTPUT«5===SORRY!5=== Error while compiling /tmp/GtXcNRyoHh␤Variable '$x' is not declared␤at /tmp/GtXcNRyoHh:1␤------> 3if 0 -> $x { say "o.O"; } else { say 7⏏5$x; }␤»
Sgeo m: if 0 -> $x { say "o.O"; } else -> $x { say $x; }
camelia rakudo-moar a38b59: OUTPUT«0␤»
Sgeo :)
ShimmerFairy m: if (42 but False) -> $x { say "o.O"; } else -> $x { say $x; }
camelia rakudo-moar a38b59: OUTPUT«42␤»
ShimmerFairy huh, didn't expect that to work :)
m: if (42 but False) -> $x { say "o.O"; } elsif (21 but False) { say "still o.O" } else -> $x { say $x; } 00:52
camelia rakudo-moar a38b59: OUTPUT«21␤»
ShimmerFairy looks like 'else' gets passed the last branch's test. Not sure if that's a bug or not though :) 00:53
00:55 llfourn joined 01:01 llfourn left
Sgeo Maybe the first way is best, if if is going to stay magic 01:02
So you could get each test's value in later tests 01:03
But then you lose the idea of pointy blocks for if
TimToady we will never scope a variable to outside its block like that 01:04
the only exception is signature declarations like that
but P5's mysterious extra scopes are not gonna happen in P6
ShimmerFairy so else getting a previous test value is a bug, I take it?
Sgeo else getting a previous test value can be useful though, e.g. if you're testing for valuable or failure 01:05
But getting only the previous one seems limiting
TimToady the main use case does not care about elsif 01:09
ShimmerFairy: not a bug, since you pass it explicitly
01:09 yeahnoob joined
Sgeo m: if Mu but True { say "Mu but True"; } else { say "Mu but True is falsey?"; } 01:11
camelia rakudo-moar a38b59: OUTPUT«Mu but True␤»
Sgeo Does that make a new Mu or a shim around Mu or something? 01:12
ShimmerFairy My random guess as to what's happening is that, since the if/elsif/else construction is taken as one expression, said construction keeps around just one 'condtional' value across the branches, writing it with whatever the latest result is. 01:13
geekosaur design.perl6.org/S12.html#Anonymous...ut_or_does 01:14
[Coke] m: (Mu but True).perl.say
camelia rakudo-moar a38b59: OUTPUT«Mu+{<anon>}␤»
geekosaur should probably use the Snn:ll syntax but looked it up in the html, not the text
anyway it's an anonymous mixin
01:18 dha left, dayangkun joined 01:19 dha joined 01:24 rmgk_ joined, rmgk is now known as Guest93505, Guest93505 left, rmgk_ is now known as rmgk 01:27 dha left
tony-o m: say "hello".^can('something'); 01:34
camelia rakudo-moar a38b59: OUTPUT«␤»
tony-o m: say "hello".^can('something'); say 'hello'.^can('something').DEFINED;
camelia rakudo-moar a38b59: OUTPUT«␤Method 'DEFINED' not found for invocant of class 'Parcel'␤ in block <unit> at /tmp/oH9eGctttL:1␤␤»
tony-o m: say "hello".^can('something'); say 'hello'.^can('something').DEFINITE;
camelia rakudo-moar a38b59: OUTPUT«␤True␤»
japhb m: say "hello".^can('something').WHAT; 01:35
yoleaux 2 Aug 2015 07:47Z <azawawi> japhb: re your discussion with timotimo++ about Electron. It is basically chrome + nodejs + API which we are accessing via JSON::RPC. We're setting up a nodejs jsonrpc server and using it via JSON::RPC and wrapping all into lovely p6 classes :)
camelia rakudo-moar a38b59: OUTPUT«(Parcel)␤»
yoleaux 2 Aug 2015 07:48Z <azawawi> japhb: examples are found here github.com/azawawi/perl6-atom-elec...r/examples
japhb tony-o: ^^ That's why, .^can is returning all matches, so always a list-like thing (a Parcel, pre-GLR) 01:36
tony-o i'm more trying to figure out what .DEFINITE does
japhb tony-o: If it's an instance, not a type object. 01:37
tony-o i saw it somewhere that i thought was getting false positives
japhb This is faster than .defined, and also not overridden by adding a 'method defined'
.tell azawawi Ah, thank you for the Electron info! 01:38
yoleaux japhb: I'll pass your message to azawawi.
tony-o makes sense why with a parcel it gets True where I'd not suspect it 01:39
then, thanks japhb
japhb yw!
01:46 colomon left 01:47 colomon joined
Sgeo Can .DEFINITE be overridden? 01:49
"X::ControlFlow 01:56
Thrown when a control flow construct (such as next or redo) is called outside the dynamic scope of an enclosing construct that is supposed to catch them."
next is dynamically scoped?
TimToady no, .DEFINITE is quite primitive
Sgeo m: sub foo() { next; }; for 1, 2 { foo(); say "Did foo()"; } 01:57
camelia ( no output )
Sgeo m: sub foo() { }; for 1, 2 { foo(); say "Did foo()"; }
camelia rakudo-moar a38b59: OUTPUT«Did foo()␤Did foo()␤»
TimToady default next is dynamic, with lable, lexotic 01:58
*label
(a bit distracted) 01:59
01:59 yqt left 02:02 roguelazer left, roguelazer joined
Sgeo "No compiler compiler available for language '$.lang'"; 02:03
02:06 EINVAL joined 02:14 EINVAL left
Sgeo m: say &lol 02:16
camelia rakudo-moar a38b59: OUTPUT«sub lol (*@l) { #`(Sub|60094288) ... }␤»
labster m: lol lol lol lol &lol 02:23
camelia ( no output )
02:30 gfldex joined 02:38 BenGoldberg left 02:40 llfourn joined 02:45 noganex_ joined 02:47 noganex left 03:28 nys left 03:42 khw joined 03:44 kaare_ joined
rangerprice so 03:50
skids sooo....? 03:51
03:52 gfldex left 03:59 cognominal left
labster m: say so .WHAT 03:59
camelia rakudo-moar a38b59: OUTPUT«False␤»
rangerprice so i love Perl 04:17
Sgeo So, when does @foo expand in an argument list and when does it not? Does it only expand when it's where a slurpy is? 04:19
04:22 gfldex joined 04:25 khw left
Sgeo m: {$_}.WHAT 04:27
camelia ( no output )
Sgeo m: say {$_}.WHAT
camelia rakudo-moar a38b59: OUTPUT«(Block)␤»
04:31 davido__ left 04:33 davido__ joined
Sgeo Hrm, there are a lot of identity-like functions :/ 04:33
04:34 davido__ left 04:36 davido__ joined 04:41 ChoHag joined 04:49 beastd left
skids Sgeo: post-GLR flattening behavior was first described here: pmthium.com/2014/10/apw2014/ then modified slightly here: github.com/perl6/specs/blob/master...-draft.pod and jnthn++s gist today is showing the first bits of an implementation prototype: gist.github.com/jnthn/aa370f8b32ef98e4e7c9 04:58
Sgeo ty 04:59
rangerprice Bye ! 05:05
Sgeo Seq is not Positional, thus cannot be bound to array variables. 05:07
05:07 rangerprice left
Sgeo That's different from current behavior with := lazy stuff, right? 05:07
TimToady yes, but we have to get better control of which types cache values 05:13
Sgeo Will slips have syntax sugar? 05:14
TimToady we might be able to unify them with |
Sgeo I think I can imagine if slips were the only things that got flattened
or... flattened's the wrong word there, but if we have slips do we need flattening?
TimToady slip is one level, flattening is recursive 05:15
05:18 skids left 05:51 nightfrog left 05:53 llfourn left 06:13 llfourn joined
ChoHag Perl#s running out of symbols. 06:25
⌤ 06:28
06:36 nightfrog joined
moritz jnthn++ # gist.github.com/jnthn/aa370f8b32ef98e4e7c9 06:37
eeks 06:42
moritz@pat:~$ dig www.bitcard.org +short
moritz@pat:~$
so, no bitcard login to rt.perl.org :(
06:43 FROGGS joined
moritz wanted to reject rt.perl.org/Public/Bug/Display.html?id=125745 06:46
FROGGS "You shall implement an ACCEPTS method" 06:47
ChoHag Well now you'll have to fix it than. 06:48
06:49 azawawi joined
moritz ok, answered by email; I hope to be able to log in later 06:49
06:49 TEttinger left
moritz (asking the authorotative DNS server gives an answer for www.bitcard.org, so I guess it's only a caching issue right now) 06:50
ok, DNS works again 06:51
ChoHag How can I have class Foo in Foo.pm and class Bar in Bar.pm where they both depend on each other? 06:57
If they were in the same file I could stub one and it'd be fine, but use-ing the files falls over.
FROGGS you cant 06:58
ChoHag That's not the sort of thing I expect to hear in relation to perl. 06:59
FROGGS you can stub Foo in Bar.pm, declare Bar and use Foo, but Foo won't know about Bar
the thing is that we have compilation units 07:00
and when you compile Foo, all the symbols it knows about must be resolved
so, the only way is to put both in the same compilation unit (file)
07:09 rurban_ joined 07:10 rurban joined 07:17 telex left, llfourn left 07:18 telex joined 07:19 darutoko joined 07:22 llfourn joined 07:24 brrt joined
nine .tell dha please merge your 5-to-6 docs into docs.perl6.org. Yesterday I would have had a use for it but couldn't find it. 07:25
yoleaux nine: I'll pass your message to dha.
07:26 RabidGravy joined, zakharyas joined
azawawi hi 07:27
yoleaux 01:38Z <japhb> azawawi: Ah, thank you for the Electron info!
azawawi how do i .can('sub') on a module?
RabidGravy marnin!
07:29 simcop2387 joined
azawawi i.e. test whether subs exist in a module or not. Any example? 07:29
labster moritz: I rejected your rejection. If there aren't any uses of declaring a new infix:<~~> that will ever work, I think we need a warning message. 07:30
RabidGravy azawawi, something like this perhaps? 07:34
m: module B { our sub c() {} }; say &B::c.defined'
camelia rakudo-moar a38b59: OUTPUT«5===SORRY!5=== Error while compiling /tmp/LEWYcVTr2o␤Two terms in a row␤at /tmp/LEWYcVTr2o:1␤------> 3 B { our sub c() {} }; say &B::c.defined7⏏5'␤ expecting any of:␤ infix␤ infix stopper␤ postfix␤ s…»
RabidGravy m: module B { our sub c() {} }; say &B::c.defined
camelia rakudo-moar a38b59: OUTPUT«True␤»
nine m: module B { our sub c() {} }; say &B::d.defined
camelia rakudo-moar a38b59: OUTPUT«False␤»
RabidGravy has to be an "our" tjough 07:35
nine Which is only logical. Why should you be able to test existence of a function you are not supposed to see? 07:36
RabidGravy oh yeah
azawawi RabidGravy: thx 07:41
07:54 azawawi left 08:04 pdcawley joined 08:10 zakharyas left 08:11 llfourn left 08:12 rurban left 08:13 zakharyas joined 08:16 zakharyas left, zakharyas joined 08:18 rurban joined 08:34 mr-foobar left 08:35 ponbiki_ left, oha left, SHODAN left, Ulti left, Sgeo left, [Coke] left, Util left 08:37 ponbiki_ joined, oha joined, Ulti joined, SHODAN joined, Sgeo joined, [Coke] joined, Util joined, agentzh joined, Possum joined, shmibs joined, Celelibi joined, bloonix joined, ilbelkyr joined, garu joined, japhb joined, sjn joined, ruoso joined
jnthn I don't think we should confuse flattening into a caller argument list with slip construction comfortably 08:40
uh, could ... comfortably
08:48 Mhoram left 08:49 Mhoram joined 08:56 vytas`` joined, vytas`` left
ChoHag Can you have a private and public method of the same name & signature, such that calls from within the class call the private and external code the public? 09:01
jnthn Sure 09:09
Note the syntax for calling a private is different 09:10
So there's no ambiguity
09:20 brrt left
Woodi hi #perl6 :) 09:30
do slip needs to be some data structure ? flat isn't... 09:31
09:32 Abc333 joined 09:33 laouji left
Woodi and realy do not know why @a = [1,2], [3,4] # 4 elems is big deal. on how many levels it is bad or something ? 09:33
Woodi checked what Perl5 is doing in = @b, @c case and it slipped into WHAT IT IS DOING ?? ;) 09:34
09:36 |Tux| joined
jnthn Actually I think flat will return a Slip 09:36
I don't think looking at "what does Perl 5 do here" is very useful. We've already very well established through the nearly-double-digit previous number of list implementations that we don't want those kinds of implicit flattening semantics. 09:38
09:40 virtualsue joined
ChoHag To be honest what perl 5 does with lists isn't all that useful in itself. 09:41
I've thought "I've got to disrupt the flow and pass and then unwrap a bloody reference again" far more than "Oh how useful - I never wanted a discrete variable anyway" 09:43
Woodi umm, think I just wrote it wasn't helpful :) Perl5 just consumed @c...
but "To concatenate two lists/arrays use flat" is weird...
ChoHag "Flatten two adjacent lists into one" isn't that weird. 09:44
jdv79 tony-o: what is wrong with the existing grammar tut? 09:45
Woodi and I'm not talking about eg. @b bying [1,[2,3]] but just [1,2,3]... maybe this is problem ?
ChoHag And having a distinction between "append this list onto that list" and "append the contents of this list onto that list" is useful and also non-weird.
09:45 rindolf joined
Woodi also there was a problem with Nil elements in lists in some cases. probably "slip them" was orginal recipe... 09:46
jnthn The new implementation doesn't take any interest in Nil 09:47
The only thing that'll vanish is an empty Slip
Woodi if you saying that's OK then it's OK :)
jnthn Well, once Nil was designed to vanish in lists, but then it was decided this wasn't desirable 09:48
09:49 yeahnoob left
Woodi anyway @a = @b, @c # 2 elems is deadly logical :) 09:51
09:52 mr-foobar joined
ChoHag Funny that logic and programming should coincide... 09:52
DrForr So rare nowadays.
ChoHag (Something about kids and lawns)
huf plant some kids on my lawn! 09:53
DrForr My brother keeps raving about how with JS nobody needs a server anymore, I keep explaining I want an *interesting* language to work with. 09:54
ChoHag Yeah who wants an internet anyway? 09:55
DrForr Apparently it's node.js all the way down. 09:56
As I sit here applying "don't repeat yourself" to codebase after codebase. 09:59
10:01 mr-foobar left 10:03 mr-foobar joined, espadrine joined
itz "You can't adverb that" <=- surely you can't verb an adverb in the error message either? 10:06
jnthn lol 10:07
But that's verbing a noun
Which is well established as productive in English, even is prescriptivists hate it. :) 10:08
ChoHag Winding up language purists is a laudable goal.
10:08 llfourn joined
itz I can't actually think of a better, short replacement anyway :/ 10:08
nwc10 or is this a bug report that the error message is LTA because it doesn't mangae to verb an adverb, and acchieve sentience (or something equally terrible)
virtualsue i'd say it's not explicable to what i assume is your target audience, e.g. people who are able to write programs and understand english reasonably well :-) 10:10
ETOOCUTE
nwc10 I think you're right, and you've expressed it better than me 10:11
in particular, is it possible to replace "that" in the message with the token that triggered the problem?
virtualsue that would probably be helpful 10:12
jnthn Uh
I'm pretty sure right below that message is a pretty red/yellow/green thingy with an eject system showing exactly where we got confused.
*symbol 10:13
10:13 llfourn left
nwc10 sorry, fail on my part. I failed to ask "or is it pointing at the offending thingy"? 10:13
jnthn m: say [1, 2, 3]:hmm
camelia rakudo-moar a38b59: OUTPUT«Unexpected named parameter 'hmm' passed␤ in block <unit> at /tmp/kj_P8RUVWW:1␤␤»
jnthn gah :)
How do I actually trigger the error...
m: class Foo { } :bar
camelia rakudo-moar a38b59: OUTPUT«5===SORRY!5=== Error while compiling /tmp/gm5uh2ic5m␤You can't adverb that␤at /tmp/gm5uh2ic5m:1␤------> 3class Foo { } :bar7⏏5<EOL>␤ expecting any of:␤ pair value␤»
virtualsue well sorry, i saw that message and it was not helpful 10:14
i'm not saying it needs to be fixed ASAP or that it's even a blocker
to anything ever ;-)
jnthn virtualsue: Was the primary problem not knowing what it meant by "adverb", or that you weren't sure what it thought you were trying to adverb, ooc? 10:15
nwc10 I think it would be clearer as "You can't use "bar" as an adverb"
jnthn If the former we could maybe do "You can't use adverb ':bar' here" 10:16
nwc10 (with the usual way of doing quotes, that I don't know how to manage
yes, how you said it
virtualsue i know what adverbs are in english. it wasn't immediately obvious to me what the message meant in the context of the code
jnthn nwc10: :bar is a perfectly fine adverb; it's the place it's being used that's at issue.
I think we could probably include the adverb itself in the error 'cus it's likely short
But if we include the entire term the parser took it as applying to, it could be something rather long... 10:17
nwc10 the adverb is enough, I think
virtualsue i'm barely a dabbler in perl6. i'm just possibly about to start learning the new required lingo 10:18
ChoHag With the :.
How do you get construction logic into a role? 10:25
Multiple roles' submethod BUILDs conflict with each other. 10:26
lizmat good *, #perl6! 10:27
ShimmerFairy: re irclog.perlgeek.de/perl6/2015-08-04#i_10998537 , first-index takes a list
ChoHag Can it even be done without the consuming class taking a hand in the resolution?
lizmat m: "foo".first-index("foo").say
camelia rakudo-moar a38b59: OUTPUT«0␤»
lizmat ShimmerFairy: and "foo" is a 1 elem lust 10:28
*list
and then whatever you specify, is smartmatched with each element
m: "foo".first-index("bar").say
camelia rakudo-moar a38b59: OUTPUT«Nil␤»
itz what about "You can't adverb that - you may need to move your colon" :)
jnthn ChoHag: We've been discussing submethods role-compose such that you end up with all of them being called
lizmat m: say "foo" ~~ "bar"
camelia rakudo-moar a38b59: OUTPUT«False␤»
ChoHag What's come of the discussions so far? 10:30
jnthn ChoHag: No code :( 10:33
ChoHag: It's pretty much at the point where we need to try to implement it 10:34
ChoHag What's 'it' then? Each role may have a BUILD submethod which is called in order or something more exotic?
jnthn ChoHag: I think it's about that. 10:36
ChoHag: Basically, arranging that find_method ends up containing something that invokes the lot
uh, *retruning* something
And installs in the method cache 10:37
Which is how it'll work out most of the time
It'll want some care, so we don't break .^lookup (which should probably just go on returning the direct declaration of BUILD in the class, if any)
Maybe submethod_table will need breaking up into submethod_table and submethod_invokee_table or so 10:38
ChoHag ... implementation details I grok only by inference. 10:39
RabidGravy in the meantime there is probably some kind of hack you can perform but it's not immediately clear to me 10:40
jnthn ChoHag: These are the names of things that actually show up in the code, which I mention in case anyone wants to grep for them, get to know the code in question, and have a go at it :)
ChoHag I'd like to have a look (really!) but it's hard enough to do my own thing - which I understand - when I can only code in 5 minute snatches. 10:42
I need a lock on my door...
RabidGravy my intuition is that it would probably involve making a specific trait_mod:<does> for the role itself which creates the method before applying the role 10:43
jnthn RabidGravy: Noh, it'd want handling entirely in the MOP code
RabidGravy: In RoleToRoleApplier/RoleToMethodApplier, and then tweaking find_method and publish_method_cache 10:44
ChoHag I think he's talking about the hacky workaround I'm considering whether to avoid writing.
jnthn Oh... :)
Anyway, I can look into it, but first I need to get the GLR stuff sorted-er
RabidGravy yeah, by way of answering the original question with what is avaulable now ;-) 10:45
jnthn And then want to get the new supply stuff in shape
10:46 sergot_ is now known as sergot 10:49 colomon_ joined 10:50 colomon left, colomon_ is now known as colomon
jnthn TimToady: Interestingly it turns out very naturally that flat won't go overly deep on [...] after all. [[1, 2], [3, 4]] actually works out as [$[1, 2], $[3, 4]] because an Array is a set of Scalar containers. So my @a = [1, 2], [3, 4]; my @b = [4, 5], [6, 7]; my @c = flat @a, @b; will end up with @c having 4 elems. Which seems good. :) 10:50
TimToady: Note that since [[1, 2], [3, 4]] is equivalent to [$[1, 2], $[3, 4]] we can of course .perl to the pretty thing without $$$ everywhere. :) 10:51
ChoHag Can you ensure that role A is never composed alongside role B? 10:53
10:53 colomon left, colomon joined 10:58 colomon left
jnthn Hm, I guess... 10:58
m: role A { }; role B { die 'no!' if any(::?CLASS.^roles_to_compose) ~~ A }; class C does B { }
camelia ( no output )
jnthn m: role A { }; role B { die 'no!' if any(::?CLASS.^roles_to_compose) ~~ A }; class C does A does B { }
camelia rakudo-moar a38b59: OUTPUT«===SORRY!===␤Cannot find method 'role_typecheck_list'␤»
jnthn Somehow the exception is swallowed and it gets upset
But it successfully ensures it :P
ChoHag heh 10:59
jdv79 jnthn++ # concrete glr stuffs 11:00
11:00 zakharyas left
jdv79 though the benchs for 1/2 of those were the same or slightly worse here 11:00
*benches?
jnthn benchmarks :)
jdv79 yes, those 11:01
jnthn I see one that is worse
And a couple of others are a small win
And a couple are big wins
What they hide is that some of them also involve a notable memory usage decrease.
On the one that is worse, I fed it to the profiler. It's suffering mostly out of some sucky code-gen on a hot path. 11:02
jdv79 no doubt. but memory is still cheaper than time:) probably always will be.
ChoHag Could/should memore usage not be benchmarked?
jdv79 oh nice 11:03
jnthn ChoHag: Probably, but I'm not going yak shaving...
jdv79: I'm not going optimizing particularly hard yet, though. I want to optimize the correct semantics. If we're winning all but one benchmark before optimization, and some of them by a factor of 5-10 or so, then we're not doing bad. :) 11:06
timotimo \o/ 11:07
is the code-gen you've seen Perl6::Optimizer-level stuff?
jnthn timotimo: More QAST -> MAST level 11:08
11:08 llfourn joined
jnthn timotimo: And not seeing spesh doing the inlining I'd hope 11:08
nwc10 jnthn: are you planning to merge after the August release? 11:09
jnthn nwc10: Not sure yet.
nwc10: We need to understand how painful the disruption will be.
|Tux| test 50000 42.285 42.195
test-t 50000 41.911 41.821
jnthn, at least NO slowdown! (even a bit better, but within noise) 11:10
jnthn |Tux|: Nice. Though I don't think I committed anything since you last checked, given I'm working out of repo on the GLR stuff for the moment :)
ChoHag Oh while I remember, can I test changes to files in src/core or src/Perl6 other than by a full recompile? 11:11
jnthn ChoHag: "make" will compile as little as is actually needed.
ChoHag: Which is still an amount
ChoHag: For changes to src/core stuff I often find a way to try them out of there first 11:12
11:12 rurban left
ChoHag What way is that? 11:12
jnthn ChoHag: If it's a new sub/method I just try it in a normal code file, or augment the class I would add it to, for example. 11:13
ChoHag Note that I haven't even tried looking yet, so apologies if it's obvious.
11:13 llfourn left 11:16 Psyche^ joined
sergot hi #perl6 \o 11:16
11:20 Psyche^_ left
lizmat sergot o/ 11:23
long time no see ?
timotimo jnthn: the QAST -> MAST part isn't what i've been working on recently, right? but maybe it's something you could roughly instruct me to look at?
11:31 rurban joined
sergot lizmat: \o yeah... I'm not ok with this. 11:31
11:31 pdcawley left
sergot lizmat: I want to come back to you people :) 11:31
jnthn timotimo: Well, looking at the code geneated for pull-one in the map iterator implementation would be a good starting point 11:32
timotimo i'll have a look later
jnthn The profiler output is telling too 11:33
lizmat sergot: welcome again :-)
jnthn You may also play with inlining thresholds
timotimo right now i'm a bit weirded out by a grammar apparently matching, but then returning (Any) :|
jnthn Of note, trying to get the size of code post-specialization to be checked against the threshold too
11:34 pdcawley joined
sergot lizmat: thanks :) 11:34
jnthn 'cus spesh tosses so many checks in a method prelude...
11:35 pdcawley left, dayangkun left 11:36 rurban left
timotimo oooh 11:37
it must be because my custom <ws> wants a \s at the end of the document ... but it's EOF instead
yes, that was it
jnthn Darn. If flat returns a Slip then it's going to have to remember stuff as it flattens. 11:42
Maybe we don't need it to return a Slip.
jnthn tries it with a Seq, and guesses we'll see if that hurts at some point 11:43
11:45 espadrine left 11:52 pdcawley joined 11:53 ttt_ joined, ttt_ left 11:57 espadrine joined
jnthn So far so good. Seems I've itemization and flattening working out. :) 11:58
Lunch time...bbiab 11:59
12:06 yqt joined 12:07 xnrand left 12:08 xnrand joined
tony-o jdv79: nothing is wrong with the existing grammar tutorial, itd be nice to have more than one and something that expands on it and maybe touches on actions..even something trivial like building a hash 12:11
12:12 brrt joined
DrForr Sounds interesting... lunch, then I might settle in for some reading :) 12:12
12:14 pdcawley left 12:22 lucasb joined 12:27 xinming_ joined
lucasb .combinations returns the concatenation of all the 0,1,2..N combinations together. What is your opinion about moving this behavior to .combinations(*) with a Whatever argument? This way, the no-arg error would give a hint to the user that the method accepts a argument, since more commonly he would want a combination of a specific number of 2, 3, etc. elements at a time. 12:28
12:28 rurban joined
brrt seems... perl6-ish 12:30
m: my @a = 1,2,3; say @a.combinations;
camelia rakudo-moar a38b59: OUTPUT« 1 2 3 1 2 1 3 2 3 1 2 3␤»
12:30 xinming left
timotimo m: my @a = 1,2,3; say @a.combinations.perl 12:30
camelia rakudo-moar a38b59: OUTPUT«((), (1,), (2,), (3,), (1, 2), (1, 3), (2, 3), (1, 2, 3))␤»
brrt yeah, that's better :-)
DrForr That's getting dangerously close to a powerset :) 12:31
nine ChoHag: regarding your Role initializer question. jnthn++ suggested using attribute default setters. And maybe this helps you: github.com/niner/Inline-Perl5/blob...5.pm6#L875
brrt m: my @a = qw(a b c); say @a.combinations(2).perl 12:32
camelia rakudo-moar a38b59: OUTPUT«5===SORRY!5=== Error while compiling /tmp/jhc0AmsMlw␤Undeclared routines:␤ a used at line 1␤ b used at line 1␤ c used at line 1␤ qw used at line 1␤␤»
brrt how to do qw in perl6?
DrForr say <a b c>
m: say <a b c>
camelia rakudo-moar a38b59: OUTPUT«a b c␤»
lucasb qw/.../ or qw (...) (with a space between)
brrt ah 12:33
m: my @a = <a b c>; say @a.combinations(2).perl;
camelia rakudo-moar a38b59: OUTPUT«(("a", "b"), ("a", "c"), ("b", "c"))␤»
brrt cool
DrForr Oh, yeah, qw() doesn't work without the whitespace, the interpreter gets confused with a nonexistent qw() function :)
timotimo mst: simsalabim bamba saladu saladim! 12:34
DrForr o/' Para bailar la bamba, se necesita una poca de gracia o/' 12:35
12:40 cschwenz joined 12:42 leont left 12:51 pdcawley joined, llfourn joined 13:01 llfourn left 13:02 dsm left 13:08 leont joined 13:16 skids joined
timotimo Grammar::Tracer says: match ALL the whitespace! 13:18
13:23 pdcawley left 13:25 llfourn joined, zakharyas joined
lizmat .tell TimToady if you're serious about ifdef, we should probably revert the existence of Index before the next release 13:26
yoleaux lizmat: I'll pass your message to TimToady.
lizmat .tell TimToady or do you think they both have a reason of existence? (as in TIMTOWTDI)
yoleaux lizmat: I'll pass your message to TimToady.
13:33 rurban_ left 13:47 rindolf left
dalek c: 0190858 | smls++ | lib/Language/subscripts.pod:
subscripts: minor expansion and tweaks to the "Basics" section
 13:52
FROGGS hmpf
FROGGS hates pcre in php sooo much
13:55 zakharyas left
timotimo i'm too dumb to parse some C code right now :| 13:56
13:56 zakharyas joined
FROGGS O.o 13:57
timotimo: too hot?
timotimo it's not hot in here at least 13:58
i'm just not seeing what's broken
13:59 smls joined
timotimo would you like to take a quick look? 13:59
FROGGS sure
smls I was wondering: Why are from-the-end indices numbered *-1, *-2, ... instead of *-0, *-1, ...
In Perl 5, they *have* to start with -1, because -0 cannot be differentiated from 0 14:00
lizmat smls: because * stands for .elems in that case
smls but with the whatevercode notation, that wouldn't have been a problem right?
lizmat: but it *could
* have been made to stand for .end instead
timotimo FROGGS: find it in MoarVM's tools/parse_jitgraph.c, you'll likely need to run tools/update_ops.p6, too to get MAST::Ops in p6
and it'll want you to have Grammar::Tracer available 14:01
it seems like the problem is best explained by the ws rule causing MATCH "\n /* atpos_i w(int64)"
after /*, it's supposed to match everything up to and including */ 14:02
lizmat smls: it could, but it feels incompatible with other * uses
smls like what?
PerlJam smls: you'd lose symmetry-around-0 -- ..., *-3, *-2, *-1, 0, 1, 2, 3, ... 14:04
lizmat also:
14:04 ponbiki_ is now known as ponbiki
lizmat m: my @a = ^10; say @a[*] 14:04
camelia rakudo-moar a38b59: OUTPUT«0 1 2 3 4 5 6 7 8 9␤»
smls PerlJam: Why *should* there be symmetry around the first element?
timotimo or * div 2 to find the right element for the median 14:05
dalek kudo/nom: 045c148 | lizmat++ | src/core/Cool.pm:
Naively implement Cool.indices

As discussed at irclog.perlgeek.de/perl6/2015-08-04#i_10998495
14:05 spider-mario joined
smls PerlJam: If "wrapping" indices are deemed desirable, why have the * notation at all instead of simply allowing negative indices like other languages? 14:06
timotimo: Ok, I see that's a practical benefit.
14:06 dsm joined
smls though having to write [(* / 2).ceiling] would not be the end of the world either 14:07
or [(* + 1) div 2]
timotimo what about ^*? 14:08
smls what is that useful for in this context?
PerlJam smls: a) aesthetics :) b) because "whatever" is a more general mechanism and allowing negative indices would allow for accidental indexing in some cases
timotimo nothing, it just fits 14:09
PerlJam smls: and your "find the middle" versions require more cognitive load than * div 2
smls I dunno, I think "mirror" symmetry would have been more approriate than "wrapping-around" symmetry. 14:10
FROGGS timotimo: eww, I've got problems here to get modules installed... 14:12
timotimo oh
i think i may just need a little break 14:13
lizmat will refrain from writing tests for Cool.indices until we have consensus on implementation 14:14
"fooo".indices("oo") # should that be (1,2) or just (1,)
14:15 jonathan_ joined, zakharyas left, jonathan_ is now known as YvargDibar 14:16 Abc333 left
timotimo oh, oof, overlapping, too? 14:16
good question
14:16 RabidGravy left, YvargDibar is now known as RabidGravy 14:17 mr-foobar left
smls lizmat: Shouldn't that be index("oo", :g) or index("oo", :all) like .match? 14:17
PerlJam Why not just use .match at that poitn? 14:18
lizmat I refer you to: irclog.perlgeek.de/perl6/2015-08-04#i_10998495
smls I meant that it would be controlled by a modifier like .match is, not that it would do what match does. 14:19
PerlJam FWIW, I tend to favor the default being overlapping matches. (I don't have a real reason other than I think that one would be more useful) 14:20
i.e. "fooo".indices("oo") returns (1,2) 14:21
lizmat smls: afaik, indices was to return *all* matches 14:22
smls PerlJam: re. from-the-right indexing: There are good reasons why all modern langauges start indices with 0 rather than 1, aren't there? Why should those reasons suddenly no longer apply just because we're indexing from the right rather than the left?
lizmat PerlJam: my thinking exactly and that's how I implemented it
smls lizmat: So like :ex/:exhaustive on regexes?
lizmat: The lack of clarity about which of :glibal/:exhaustive it should do, is imo another indication that it should be explicitly controlled by adverbs (rather than just picking one behavior). And at that point one could as well reuse the existing .index name for it. 14:24
I wonder if TimToady considered that option 14:25
14:26 rurban left
lizmat well, "indices" is plural of "index", so I guess it's implicit in the name 14:26
afk for a bit
smls alternatively, have .indices with an :ov/:overlap adverb. 14:27
TimToady I think people will expect non-overlap by default 14:28
yoleaux 13:26Z <lizmat> TimToady: if you're serious about ifdef, we should probably revert the existence of Index before the next release
13:26Z <lizmat> TimToady: or do you think they both have a reason of existence? (as in TIMTOWTDI)
14:28 census joined
TimToady I suspect ifdef is a nice general feature, while Index is kind of a specific feature that may doesn't pull it's weight 14:30
jdv79 tony-o: the existing tut does have actions 14:33
skids Well, beyond array indices, it does offer a general thing for people who want 0 but True cleanly integrated into abuiltin I guess.
hoelzro o/ #perl6 14:35
TimToady hard to know whether people would prefer clean Ints and a forced ifdef, or a weird Ints that work with if counterintuitively
skids I guess it could ease transcriptions from languages that have 0 but True... which are...? 14:36
TimToady I think ifdef/Int is a lighter-weight solution 14:37
I agree we can use adverbs on indices to indicate overlap 14:39
but I still think index is good for when only one is wanted, since you don't have to set up a lazy pipeline for that
nine I guess it depends on if we find another construct besides if $foo.index('x') -> $pos { ... } where the 0 but True solution would help. 14:40
PerlJam re ifdef: I dunno ... ifdef/Int is a lighter-weight solution, but it adds to the overall cognitive load of the language ("wait, there's *another* form of if statement?") whereas Index only adds cognitive load in one specific spot (or few spots) 14:41
skids Well, there are bound to be some NativeCall libraries that have to map a "-1 if failure, 0+ is a handle" 14:43
14:43 llfourn left, nys joined
lucasb Testing only the booleanness of some value and then assuming it is equal or not equal zero can turn out to be wrong-ish because you can't antecipate the crazy values other users will feed to your code. If the programmer wants the code to be strict, he should explicit test things with ($n == 0), ($s eq ''), @a.elems == 0, shouldn't he? If he justs tests booleanness, the he must be aware of the existence of ob 14:44
jects that overrides .Bool/.defined and things like '0 but True'.
skids lucasb: or constrain his types.
m: say Index ~~ Int
camelia rakudo-moar 045c14: OUTPUT«True␤»
skids Hrm. 14:45
skids would not want to see sub a (::T Int $i where { not T.new(0) }) everywhere. 14:47
But I guess that's beyond the blame of just Index since anyone can make one. 14:48
lucasb sorry, what I said was only tangentially related to what were being discussed :) 14:49
but it is the specific case of .combinations 14:50
m: say (@).combinations(0).perl
camelia rakudo-moar 045c14: OUTPUT«((),)␤»
dalek kudo/nom: d11c710 | lizmat++ | src/core/Cool.pm:
Implement Cool.indices(:overlap)
lucasb m: say (@).combinations('a'.index('a')).perl
camelia rakudo-moar 045c14: OUTPUT«()␤»
lizmat commute to Amsterdam.pm meeting &
14:50 lizmat left 14:51 brrt left
lucasb and again, I know this .combinations/.index example is contrived :) 14:51
skids m: sub a ( Int(Int) $i) { say so $i }; a( "a".index("a"));
camelia rakudo-moar 045c14: OUTPUT«True␤»
skids wonders if that coerce should strip the "0 but True" 14:52
TimToady that's the problem with 0 but True, it has indeterminate scoping issues 14:53
lucasb m: sub f(Int() $x) { say $x.WHAT }; f('a'.index('a')) 14:54
camelia rakudo-moar 045c14: OUTPUT«(Index)␤»
TimToady I really think ifdef is cleaner, and we'll end up encouraging people to think more about defined/undefined (I almost typed underfed)
lucasb Interesting that Int() doesn't coerce to Int, because Index already is a Int
so, it is different that $index.Int
*than
skids The Int(Int) question stands with or without Index, though. 14:56
nwc10 when I am dictator of the Universe (TM) all paper in all of Google's offices will be A4.
TimToady you should have thought about that before you let us invent Google before you 14:57
nwc10 oh, well, once I have a time machine I'll also pop round to Sun and fix their printers too. 14:59
nebuchadnezzar nwc10: you should travel far away and invent the metric system before any others ;-) 15:03
TimToady actually, if you go far enough away, the metric system might be in base 8, if the majority number of fingers in the galaxy is 4 instead of 5 15:04
15:08 lizmat joined 15:09 rindolf joined
moritz [^citation needed] 15:10
PerlJam speculations need no citations
moritz [^citation needed] :-) 15:11
PerlJam :)
TimToady ET cites moritz for driving with 5 fingers instead of 4. 15:12
PerlJam Does the JVM backend have some problem with gather/take? I notice in Baggy.pm there's some #?if jvm (use map) #?if !jvm (use gather) 15:13
15:13 raiph joined
jnthn PerlJam: Yes, and it involves continuations and control exceptions, and guess how mnay people are up for debugging those things... 15:14
lizmat m: sub a(Int(Index) $a) { say $a.WHAT }; a 42.Index # would consider this to be a bug
camelia rakudo-moar 045c14: OUTPUT«(Index)␤»
15:15 khw joined
jnthn lizmat: Missing warning about useless use of coercion type? 15:15
lizmat: If you expect anything else then you're not getting it.
lizmat no, the fact that it doesn't coerce to an Int
jnthn It IS AN INT.
That's how types work!
lizmat well, fwiw, I always interpreted the coercion sugar Foo(Bar) $a as, call $a.Foo if $a.WHAT === Bar 15:17
moritz that's not what it does
jnthn I think we even docuemnt .WHAT === as being an anti-pattern.
moritz it does $a.Foo unless $a ~~ Foo
lizmat I understand that now
jnthn Yeah but
masak .WHAT is only for debugging
jnthn I'd be surprised if .Int on Index didn't return identity too if it's a true coercion method. 15:18
masak I don't even use it for that anymore -- I prefer .^name
(hi #perl6)
lizmat m: say 42.Index.Int.WHAT
camelia rakudo-moar 045c14: OUTPUT«(Int)␤»
lizmat I implemented it that way specifically so that you could remove the "0 but True" magic from an Index 15:19
tony-o jdv79: one tutorial doesn't mean more tutorials or more in depth tutorials isn't a good idea
lizmat
.oO( I guess that's just one more nail in the Index's coffin )
although I do think this points to a more general problem wrt to subclassing and coercion 15:20
what would be against removing the check in the coercion logic altogether? 15:21
15:21 muraiki joined
lizmat sub a(Foo(Bar) $a) { # always call .Foo on what is being passed before putting it in $a 15:21
jnthn The contract of a coercion is that it's identity if the thing already matches the type to coerce to 15:22
masak lizmat: even if you did that, wouldn't the fact that Index ~~ Int mean that you either did no cast, or the .Int method breaks the contract of casting? 15:23
TimToady yes, coercion is only "make sure this can act like one"
jnthn I think the problem is in your understanding of subclasses, not in Rakudo.
masak m: say 42.Index.Int.WHAT 15:24
camelia rakudo-moar d11c71: OUTPUT«(Int)␤»
masak well, and in Rakudo ;)
jnthn Darn it :P
moritz there's nothing wrong with .Int returning an Int
lizmat masak jnthn : that's just how I implemented it 15:25
jnthn gist.github.com/jnthn/aa370f8b32ef98e4e7c9 # latest GLR work, now with itemization and flat
masak m: say [].List.^name
camelia rakudo-moar d11c71: OUTPUT«Array␤»
masak moritz: at the very least, it's inconsistent.
moritz aye
if anybody on hack is consuming big amounts of memory, please kill your processes now 15:26
lizmat TimToady: so if I have an Index, what would be the best way to make it an Int again ?
moritz otherwise the OOM killer will do so, randomly
lizmat m: say +(42.Index).WHAT
camelia rakudo-moar d11c71: OUTPUT«Use of uninitialized value of type Index in numeric context in block <unit> at /tmp/qc0324zEOG:1␤0␤»
lizmat m: say (+(42.Index)).WHAT
camelia rakudo-moar d11c71: OUTPUT«(Index)␤»
jnthn TimToady: I currently have my @a = @b, @c; coming out as 2 elements, matching [@a, @b] also doing so. Unfortunately, there's a bit of a discontinuity... 15:27
TimToady: my @a = @b; is fine; my @a = flat @b, @c; is also fine (flat applies to both of course, just a normal listop). my @a = 1..10; is fine as Range is Iterable. But [1..10] is a single element... 15:28
masak lizmat: I've been telling people asking that on the channel ("how can I get `$thing` back if I did `$thing but $other`?") that they ought just hold on to `$thing` in the first place and not let it go out of scope or be replaced.
TimToady why?
tony-o jdv79: a single tutorial doesn't lend itself to the timtowtdi way of thinking..
jnthn TimToady: I can of coures make [1..10] be [1, 2, 3, ...] provided [@a] means [@a[0], @a[1], ...] 15:29
]
oops
But given [...] doesn't itemize any more that means it's basically a copying no-op
TimToady why wouldn't it just follow the same single-arg rule as everywhere else?
masak lizmat: in the case of the `but` happening inside of an API call, that's not really an option though. which is why I keep hoping that doesn't happen to .index
jnthn TimToady: I...hm. It probably can :)
TimToady: OK, let's do that. Thanks. :) 15:30
TimToady that's how I was prototyping it in glrish
lizmat masak: the "but" is only conceptual, Index is a subclass of Int with "method Bool { self.defined }" mixed in
jnthn Yeah, it makes sense. Something bugged me about it, I can't remember what.
timotimo "how to get the opposite to work"?
jnthn I'll do that and add it to my collection of tests.
TimToady: BTW, things are fitting together quite pleasantly so far. 15:31
TimToady I see that
jnthn I've missed out a bunch of the iterator fast-path overrides for now.
masak lizmat: then I'd argue it's already breaking Liskov substitutability. Int has a 0 value which boolifies to False.
jnthn It occurs to me that if we can get a VM-level memcpy style thing we'll be able to go even faster. 15:32
TimToady as long as you know where they go eventually
jnthn *nod*
Yeah, I've been feeling out the Iterator API
masak lizmat: if you have a 0 value in your type which does not boolify to False, you're not an Int.
TimToady will be interesting to see how fast the sad path is
jnthn Yeah 15:33
lizmat masak: so you're saying Index should not inherit from Int, but it should have all of the functionality of an Int (except boolification?)
masak lizmat: no, I am not arguing in favor of Index at all.
lizmat: also, I'm completely not backlogged. last time I was active (last week), this discussion was about Pos. 15:34
jnthn TimToady: For Array, .flat basically is a no-op *if* it really is a bunch of Scalar containers. Trouble is, somebody could in theory bind an Iterable directly to one of the slots.
diakopter *fatally unbackloggable
jnthn TimToady: Should we (a) consider that erroneous, speed FTW!, or (b) actually care about that possibility? 15:35
(or (c) refuse to let you bind Iterable directly :P)
TimToady I vote for a, I think
jnthn OK, wfm 15:36
nine (1 GLR, 2 GLR, 3).GLRmap({ GLRslip($_, 3 * $_) }) # I'm no longer sure if the GLR is such a good thing. It seems to be so much more typing work. 15:38
nine ducks and runs
15:38 captain-adequate joined
jnthn Of course, it really stands for Gets Longer, REALLY! 15:39
timotimo nine: actually, it's "GLR( GLR1 GLR, GLR2 GLR, GLR3 GLR) GLR. GLRmap({ GLRslip( GLRDollarUnderscore, GLR3 GLR* GLRDollarUnderscore) })
"
nine jnthn: do you already have caps lock bound to "GLR"? ;) 15:40
TimToady we're also going to go to all caps everywhere, for greater consistency
timotimo m: say "we're also going to go to all caps everywhere, for greater consistency".uc 15:41
camelia rakudo-moar d11c71: OUTPUT«WE'RE ALSO GOING TO GO TO ALL CAPS EVERYWHERE, FOR GREATER CONSISTENCY␤»
TimToady wow, that's so much nicer in all caps
nine Maybe we could sell keyboards with GLR keys to Perl GLR6 developers and fund GLR development that way.
TimToady the VT05 had one of them allcaps keyboards... 15:42
not to mention the ASR-33
PerlJam glrs at nine
lizmat driving& 15:43
15:43 lizmat left, coffee` joined
jnthn gist.github.com/jnthn/aa370f8b32ef98e4e7c9 updated again with the [...] single item fix 15:45
lucasb You people should have chosen an acronym that was easier to make puns about
jnthn The trouble is, if it goes badly, it'll be forever known as the Great List Refuctor... 15:46
15:47 FROGGS left
ugexe i think it should forever be known as that anyway 15:47
smls I just realized that splice() could be turned into a subscript adverb... 15:48
When someone brought up the idea before I though it wasn't possible for subscripts to select zero-length slices like splice() can, but that could be made to work using end-exclusive ranges:
my @a = 0, 10, 40; @a[1^..^2]:splice = 20, 30
@a would be 0, 10, 20, 30, 40 now 15:49
TimToady 1..0 is shorter 15:50
[()] is even shorter than that 15:52
smls but it wouldn't carry information about the position of the zero-length slice
TimToady 2..1 is fine for that
smls which would be needed for :splice to know where to put assigned values
m: dd 2..1
camelia rakudo-moar d11c71: OUTPUT«2..1␤»
TimToady except that the flattening might occur too soon
jnthn Darn, making '@a = @a' work on infinite things needs some careful thinking... 15:53
jnthn takes a few moment's break
& 15:54
JimmyZ any reason that some GLRIterationBuffer.CREATE with BEGIN? some not?
timotimo so that the object created by , can share its empty GLRIB with all others 15:56
no need to have multiple of those, since they aren't modified ever
15:57 telex left 15:58 telex joined
smls One problem with a :splice adverb would be to decide what to with non-consecutive (i.e. list rather than range) slices. 16:01
@[2, 6, 7]:splice; could simply remove those elements, byt what would @[2, 6, 7]:splice = 1, 2, 3, 4; do?
16:01 sftp joined
smls maybe it would need to be restricted to Range subscripts. 16:01
TimToady splices are unusual enough (and inefficient enough) that I'm not much inclined to overload subscripting with it 16:02
smls although there's a subset of what splice does - namely removing individual elements - that is not so unusual 16:03
16:04 dg joined
smls And having to go through splice() for that feels suboptimal. 16:04
Its one of those functions that scare newbies, and requires even experienced programmers to look up what each argument was supposed to mean again. 16:05
16:05 leont left
TimToady the replacement seems just as scary 16:05
smls as a subscript adverb, at least the simple/common cases would be friendly @a[5]:splice;
Woodi lizmat: I think you discovered natural born Monad situation... looks .index want to return 2 things, search status and result like struct Index { Bool found; @pos } 16:06
TimToady maybe "ifdef" should be spelled "maybe" 16:07
maybe "abc".index('a') -> $pos {...} else {...} 16:08
Woodi is -> $pos part needed ? 16:09
TimToady it is if you want to use the position
or I suppose we could bind to $_ in that case
Woodi right, $_ was for loops... 16:10
b2gills If it weren't for the `-> $pos` part it would just be `if defined ...`
timotimo if you don't need the -> $pos at all, you can just as well use "if defined"
yeah
TimToady except it can't default to binding $_
timotimo ifdef could, but if defined can't, right? 16:11
b2gills m: if defined 0 -> $_ { .say }
camelia rakudo-moar d11c71: OUTPUT«True␤»
smls "ifdef" will confuse C people :P
TimToady I suppose it could be viewed also as a form of given that also tests definedness
so, like given, ifdef could be used as a modifier 16:12
b2gills I don't think it should be if it has "if" in the name
smls would a longer name like "ifdefined", "if-def" or "if-defined" be so bad?
TimToady we try to avoid - in keywords
smls after all, the method is also called .defined and not .def 16:13
"ifdefined" then?
b2gills `maybe 0 { .say}` is fine but `ifdef 0 { .say }` isn't
TimToady b2gills: how so?
b2gills it would be a common source of confusion, which I don't think is worth it 16:14
smls 'maybe' doesn't read quite right in this context imo.
TimToady and worse as a modifier
b2gills of course find a better word
Woodi how Haskell _use_ monads after they are returned ? 16:15
geekosaur ?
smls if/for/while are all named so that the keyword refers to the condition
ugexe too bad .so is taken
smls but maybe would refer to "maybe this block", not "maybe this conditional expression" 16:16
timotimo so .WHAT
b2gills
.oO( do we really need `.so` when we have `.Bool`? )
smls "ifdefined" would immediately remind people of "if". 16:17
timotimo how about "defined?" :P
TimToady the keyword wants to have more of a topicalizer flavor, but with "guarantee" as part of the connotation
provided maybe
timotimo that.s not so bad 16:18
TimToady works is modifier too
timotimo assuming is already taken
smls
.oO( onconditionthat ... { ... } )
lucasb Is there a shortcut for things like '$x but Defined', '$x but Undefined' ?
Wouldn't it be easier if a role/enum existed for that? 16:19
TimToady why would we want to make that easier?
'but' is already smelly enough
lucasb well, so I can create the same confusion to the new ifdef :) 16:20
timotimo .o( a smelly but )
b2gills lol
... literally
16:23 rurban_ joined 16:28 leont joined 16:30 bhavikdbavishi joined
bhavikdbavishi p6: say 3; 16:31
camelia rakudo-moar d11c71: OUTPUT«3␤»
16:31 bhavikdbavishi left
timotimo i wonder how good thunderbird's irc client is 16:31
16:32 revdiablo joined
TimToady actaully, maybe we could finally use "with", which is just about the right length, and has both topicalizer and conditional connotations 16:32
and works either before or after
diakopter withdef 16:33
hoelzro submethods aren't necessarily tied to object lifecycle, are they? BUILD and DESTROY just happen to be submethods 16:34
timotimo may i suggest "whif" please?
because "with" doesn't seem conditional to me at all
maybe i haven't been conditioned to see the connotation yet
TimToady with permission, he proceeded to the gate 16:35
timotimo hm
TimToady with a ticket, you can get in
b2gills "with" in other languages are similar to "given" in Perl 6 16:36
timotimo i'm with b2gills on this one
[Coke] b2gills++
TimToady well, all topicalizers tend to have conditional overtones in natural languages 16:37
and I just have a gut-level feel that, since we designed Maybe into our type system fundamentally, distinguishing maybeness might well warrant one of the rare topicalizers 16:38
b2gills
.oO( Perhaps there is word that is conditional on being alive )
 16:39
diakopter whiff
timotimo: for if and only if
timotimo %)
diakopter TimToady: waxually, iff? 16:40
or
really
forrealz 16:41
TimToady not a topicalizer, really
16:43 mr-foobar joined, yqt left
b2gills Since it's main purpose is to allow a topicalizer, I agree it should probably be one. I just think it needs to be a fairly unique word among other programming languages because it is fairly unique feature. 16:44
16:46 rurban joined
hoelzro I ask about sumethods because it seems that $.variable is disallowed in *all* submethods, but wouldn't it make sense to allow $.variable in non BUILD|DESTROY submethods? 16:47
m: for lines() -> { .say } 16:49
camelia rakudo-moar d11c71: OUTPUT«Too many positionals passed; expected 0 arguments but got 1␤ in block <unit> at /tmp/t8lE4FpTmH:1␤␤»
hoelzro shouldn't $_ be the implicit argument for the block there? 16:50
b2gills You explicitly stated that the block takes no arguments
TimToady if you leave out the ->
you gave it an explicit sig of ()
b2gills I'm looking at synonyms of "consider" 16:51
hoelzro ah duh 16:55
thanks b2gills
and TimToady! 16:56
TimToady should learn to microbacklog
smls which built-in types which throw an "index out of bounds" error when subscripting a nonexistent index? 16:57
m: 4[1]; 16:58
camelia rakudo-moar d11c71: OUTPUT«Index out of range. Is: 1, should be in 0..0␤ in block <unit> at /tmp/pUbRAb1zXh:1␤␤Actually thrown at:␤ in block <unit> at /tmp/pUbRAb1zXh:1␤␤»
smls ^^ other than that
do we have fix-sized arrays yet?
or other collection types which do that?
TimToady grep for the error
b2gills TimToady: closest word I have is "granted" 16:59
17:03 brisby joined, lizmat joined 17:06 yqt joined
smls Can anyone think of an example where it makes sense to pass a proper code block (rather than whatever-currying) to a subscript? 17:07
I already have these example for non-standard code expressions: 17:08
@array[* div 2]; # select the middlemost element
@array[$i % *]; # wrap around indices ("modular arithmetic")
Would like to show an example of the form @array[{...}] or @array[-> $n {...}] as well to drive how the point of how generic it is 17:09
but can;t think of an example that would actually make sense in practice.
(this is for p6doc btw)
17:10 brisby left
TimToady a rotor subscript based on sqrt $n, maybe? 17:10
timotimo sampling points based on a chevvychev distribution
smls I'm afraid I don't understand either of those suggestions :P 17:11
(but will google)
TimToady how 'bout calculating an inversion slice from the size of a matrix 17:15
timotimo numpy has this cool thing where you can some_matrix[some_matrix == 5] = 6 17:16
b2gills @array[ {␤$_ / 3,␤$_ / 2,␤$_ * 2 / 3␤} ] 17:17
timotimo indeed 17:18
but that can be @array[ * / 3, * / 2, * * 2 / 3 ], too
b2gills
.oO( If "given" was spelled "with" then we could have used "given" instead of "ifdef" )
 17:21
lizmat m: my @a = ^100; say @a[ { (^$^a).grep: *.is-prime } ] # smls contrived, but yeah
camelia rakudo-moar d11c71: OUTPUT«2 3 5 7 11 13 17 19 23 29 31 37 41 43 47 53 59 61 67 71 73 79 83 89 97␤»
smls thanks, I think I'll go with that for now :P 17:22
I didn't know the code was allowed to return multiple indices 17:23
lizmat ah, yeah :-)
smls m: say (0, 10, 20, 30)[{1, 2}]
camelia rakudo-moar d11c71: OUTPUT«10 20␤»
smls m: say (0, 10, 20, 30)[0..{1, 2}]
camelia rakudo-moar d11c71: OUTPUT«Cannot call Real(Block: ); none of these signatures match:␤ (Mu:U \v: *%_)␤ in block <unit> at /tmp/3hytDrrAhD:1␤␤»
smls m: say (0, 10, 20, 30)[0..*-1]
camelia rakudo-moar d11c71: OUTPUT«0 10 20 30␤»
smls ah so in the range case, only WhateverCode is allowed? 17:24
lizmat smls: lemme check 17:27
m: "abc".indices("").perl.say # meanwhile, what should this say? 17:28
camelia rakudo-moar d11c71: OUTPUT«(0, 1, 2, 3)␤»
lizmat I would argue: either (0,1,2) or ()
moritz m: say 'abc'.rindex('')
camelia rakudo-moar d11c71: OUTPUT«3␤»
moritz m: say 'abc'.index('')
camelia rakudo-moar d11c71: OUTPUT«0␤»
moritz I'd argue (0, 1, 2, 3) is correct.
lizmat moritz: ok 17:29
17:30 raiph left
b2gills m: say (0, 10, 20, 30)[0..Whatever] 17:36
camelia rakudo-moar d11c71: OUTPUT«0 10 20 30␤»
17:49 FROGGS joined
dalek kudo/nom: 45f2924 | lizmat++ | src/core/Cool.pm:
Add indices() as a sub
 17:50
kudo/nom: 47654fa | lizmat++ | t/spectest.data:
Add tests for indices()/.indices
ast: cdbf63e | lizmat++ | S32-str/indices.t:
Add tests for indices
 17:51
jnthn JimmyZ: As timotimo said, basically because in some cases we can share a single empty bit of storage due to immutability, but in others we can't
lizmat dinner&
17:51 lizmat left 17:58 [particle]1 left 17:59 kurahaupo joined 18:02 rurban left 18:03 cognominal joined, yqt left
[Coke] so what's the current proposed change to replace Index? 18:05
moritz dunno, Int? 18:06
collect.p6c.org/cgi-bin/collection....format=SVG something seems to hog memory, and then dies, and tries again 18:07
if anybody has an idea what process that is, please tell me (and make sure to memory-limit it) 18:08
jdv79 [Coke]: what do you propose?
timotimo moritz: maybe look through user's mail boxes for failure reports from cron? 18:10
18:11 colomon joined
[Coke] jdv79: I don't. 18:11
but it seemed like there was discussion earlier about a generic mechanism with a name to be bikeshedded. 18:12
timotimo moritz: what coke means is not the class that'll replace Index, it's more about the name of ifdef/given/with/...
[Coke] moritz: so we have 1) Keep Index, 2) Kill Index, 3) Kill index & also provide some new thing.
18:13 cognominal left
jdv79 where else would this new thing be applicable 18:15
RabidGravy next time someone's in there wanna merge this one character PR github.com/rakudo/rakudo/pull/486 ;-) 18:16
[Coke] jdv79: that's why I'm asking if there's a proposal or something. 18:17
dalek kudo/nom: ddee95d | RabidGravy++ | lib/Test.pm:
skip-rest has the wrong name in the error message
kudo/nom: 172b87c | (Will Coleda)++ | lib/Test.pm:
Merge pull request #486 from jonathanstowe/test-errors

skip-rest has the wrong name in the error message
18:18 census left
[Coke] smacks dalek and or github for screwing up his karma. 18:18
flussence
.oO( if:defined $str.index('foo') -> $pos )
 18:20
moritz my $idx = $str.index('foo'); if defined $idx { ... }; # come on, there are cases where two statements are OK! 18:21
RabidGravy [Coke]++ # fix the karma 18:22
flussence I wasn't all that serious there, it's a pretty ugly suggestion :) 18:23
[Coke] IWBNI if someone turned blogs.perl.org/users/aaron_baugher/...tos-6.html into a ticket
moritz now asked serverfault.com/questions/710847/ho...-cron-jobs 18:25
flussence my first thought would be "cgroups" 18:26
I've never actually done such a thing myself though so I've no idea how much of a PITA it'd be
moritz flussence: the problem is that cron runs in its own cgroup, not in the users' cgroup 18:27
flussence hm, that makes it a bit awkward
moritz flussence: if there is no better solution, I'll limit global cron's memory usage, but I'd *really* like it per user
18:27 mohij joined
jdv79 is there a Maybe type? 18:29
moritz all of our types are Maybe types, because you can usually pass the type object too 18:30
m: sub f(Int $x) { say $x }; f 3; f Int
camelia rakudo-moar 47654f: OUTPUT«3␤(Int)␤»
moritz m: sub f(Int:D $x) { say $x }; f 3; f Int
camelia rakudo-moar 47654f: OUTPUT«3␤Parameter '$x' requires an instance of type Int, but a type object was passed. Did you forget a .new?␤ in sub f at /tmp/Dae7Kc455s:1␤ in block <unit> at /tmp/Dae7Kc455s:1␤␤»
18:31 espadrine left
dalek c: efdfaee | smls++ | lib/Language/subscripts.pod:
subscripts#Basics: some rewording to improve clarity
 18:32
c: f837812 | smls++ | lib/Language/subscripts.pod:
subscripts: Fill in section "From the end"
jdv79 ah 18:33
flussence moritz: there's also the option of tweaking the overcommit sysctls (making runaway processes OOM early), but I know from experience the JVM really doesn't like it when you do that 18:34
moritz flussence: and the OOM killer has the not-so-nice property that it sometimes kills innocent bystander processes
flussence that too :( 18:35
(iirc the setting that screws up the JVM is the one where it returns -ENOMEM correctly...)
18:36 rurban joined, rurban left
RabidGravy in Test.pm there is a tiny bit of pod, but only has the one function throws-like documented, would it be add the rest of the public interface to the pod or make it a separate document? 18:38
I'm up for adding it to the pod so that "p6doc Test" works nicely 18:39
18:39 rurban joined
moritz RabidGravy: no; it's documented already in p6/doc 18:40
and we don't have the bandwidth to maintain documentation twice
RabidGravy right, so it's take the pod there is in there out and check the other one is up to date 18:41
18:41 lizmat joined
jnthn [Coke]-- # bot violence ;) 18:42
[Coke]++ for applying the PR though
Not to mention [Coke]++ for all the other things...
[Coke] jnthn: heh; haven't been doing much lately.
jnthn Shhh, I might -- you again for that :P 18:43
jdv79 whats wrong with if:def?
moritz 1) it's ugly 2) it'd be the only keyword that contains a non-alpha 18:44
jdv79 not muuch uglier than ifdef and its clearer that it's a modified if 18:45
but i can see the distaste 18:46
jnthn ifdef may not be pretty, but sticking a colon in makes it butt-ugly... :P
moritz note that it should also work in the 'elsif' case 18:47
jdv79 elsifdef elsif:def
both meh
moritz and since there doesn't seem to be a good solution, I'd just leave it as is 18:48
jdv79 if:D :)
yeah, i also don't get all the attention this is getting 18:49
how valuable is this feature in totality
jnthn Yeah, but commenting on this is FAR easier than reading my GLR gist, understanding it, and discussing that :P 18:50
moritz jnthn: I did the first part, at least :-) 18:51
jnthn moritz++
TimToady it's not about index, it's about testing for success/failure 18:52
nine jnthn: well I've read your gist far enough to notice all those GLR this GLR, GLR that ;)
moritz jnthn: the only thing I don't like about it is that the question of whether we flatten something is still (partially) determined by types, not solely by operators
jdv79 i know
TimToady as I see it, we'll have with/without as the control forms of "andthen"
jdv79 where else is that applicable is what i meant 18:53
TimToady to go with the if/and
jnthn moritz: How would you solve the "how to write .kv" problem? 18:54
18:54 EINVAL joined 18:55 nys left
jnthn moritz: 'cus Slip is basically for taking on the "map/loop/whatever that wants to return some number of items other than one" 18:55
TimToady slip is an implementation type, so not in the same league as the autoslipping that Range and Array currently do
it's abstracting all the evil out into one place
jnthn I don't expect many folks to write slip(...) in actual code, but to explicitly do flat.
18:56 nys joined
TimToady jnthn: I suspect we can probably unify | with slip at some point, depending on context 18:56
jnthn TimToady: I don't really see it; return |($key, $value) is clearly an argument to return
That should flatten into the argument list 18:57
TimToady probably the return type wants to be Slip in that case
jnthn Which is different from return slip($key, $value);
TimToady and possibly we make return notice when there's a single | there 18:58
jnthn Also, if Slip is an implementation type and we don't expect to see it often, why do we need any kind of sugar for it?
TimToady that flipside of the one-arg rule, as it were
18:58 g4 left
smls simply rename the current prefix | (argument interpolation) to prefix ||, then you can use | for flattening or slipping :P 18:59
TimToady my @foo = |@foo, |@bar
jnthn That's a 2 char saving over my @foo = flat @foo, @bar;
smls the specs already mention || as a special form or argument interpolation anyway
TimToady if you don't want to flat everything for some reason
smls *form of
TimToady basically | would be taking on the role that * does in Ruby 19:00
moritz and a separate list concatenation operator is still out of favor?
TimToady jnthn: do you have any objection other than the "return is trying to pretend to be a real function" argument?
jnthn TimToady: Oh, I just picked return as an easy example of a function :) 19:01
TimToady: My broader point was that I don't want to confuse two different things
TimToady my broader point is they might be the same thing
moritz I agree with jnthn here
TimToady: why do you think they might be the same thing? 19:02
TimToady because they both remove the outer objectness of a list, it's just one goes into a capture rather than a normal list 19:03
moritz I see a difference between interpreting a list of elements as a signature (|) and passing a flat list to a single parameter (slip)
jnthn I'm not sure I can see how they are. Consider |c, where c is a Capture. Or |%h, where it's a Hash. 19:04
In both cases there's named things to interpolate
Not to mention that a slip stops evaluating at the first potentially eager thing
But argument list interpolation forces eagerness
TimToady well, but those are mutually exclusive contexts 19:05
jnthn So you're saying you'd make anywhere that | is an error now create a slip?
TimToady it's an "interpolate this" in either case
yes
jnthn Well, if that's all it is then it's trivial to implement, though it's certainly a unification at a high level rather than at an implementation level. 19:07
smls m: my Int @array; say @array[10]; say ( Array[Int].new )[10]
camelia rakudo-moar 172b87: OUTPUT«(Int)␤(Any)␤»
smls ^^ why are these different?
TimToady I think it will do a better job of hiding the Slip type as well
19:08 colomon left
smls m: my Int @array; say @array.of; say (Array[Int].new).of 19:08
camelia rakudo-moar 172b87: OUTPUT«(Int)␤(Int)␤»
TimToady and I picked the | operator in the first place because it looks like a slit into which you slip something
19:09 colomon joined
jnthn TimToady: We can try it; I've a feeling it may bite somebody somewhere 19:11
Though we can't try it in my prototype 'cus it needs compiler changes
TimToady sure
you don't want to end up with God Branch 19:12
darn it, I haven't even backlogged up to the point I went to bed last night, and it's already lunchtime... 19:13
jnthn I'm going to finish up the things I wrote in comments that I should write test for
TimToady looks like real good progress
jnthn Then I'll perhaps have a go at while lazy loop/while/until will work out
uh
at *how* lazy
what on earth, brain... 19:14
Then perhaps start digging through CORE.setting to see what bits of lists it depends on
19:14 g5 joined
TimToady funzies 19:15
jnthn You're fine with a Seq passed to an @foo calling .list on the Seq and binding the result, yes?
So foo(@bar.map(...)) works out? 19:16
TimToady yup
jnthn OK, good. And the memoization of the .list result made sense to you also?
It's there in part because things like callsame(...) pass the original Capture 19:17
So we'd like subsequent .list calls to just return what was produced before 19:18
And since List is immutable that's harmless
TimToady yes, as you say, it's immutable, or sup...
yeah
19:18 smls is now known as smls_
moritz did I get this right that signature binding allows a wee bit of coercion? 19:18
jnthn OK, good.
TimToady moritz: yes, the @ now implies caching
jnthn moritz: A...wee bit in this place, yes
moritz if so, we could use that to make passing of typed arrays less awkward
TimToady give him an inch...
jnthn hah :P
moritz (though maybe that's a post-GLR topic)
jnthn Damn, I walked right into that trap :P 19:19
moritz: It is, though it's a good point.
moritz does the @ in a *@slurpy also imply caching? 19:20
jnthn Indirectly
TimToady you can do |rest if you need to
but maybe there's something inbetween, like *\rest 19:21
if we end up with some coercion semantics on positional @, then maybe some of that will carry over to *@ 19:22
dalek kudo/nom: f625b9b | jnthn++ | src/vm/moar/ops/perl6_ops.c:
Bring Scalar creation on in line with JVM.

Which also makes it usefuller for GLR.
kudo/nom: 65915b7 | jnthn++ | src/Perl6/Actions.nqp:
An nqp::op action micro-optimization.
jnthn Ah, I guess actually we will need to look through the Capture for Seqs when binding *@foo 19:24
TimToady: Oh...question I thought up
What type is (1,2;3,4) now?
hoelzro anyone have any thoughts on what I said about submethods above? I was experimenting with a library that makes submethod PRE/POST into invariants for a class, and I found it surprising I couldn't use $.variable in PRE/POST
TimToady jnthn: lemme think about that over lunch
jnthn TimToady: OK, np :) 19:25
ohhh...and I was meant to go to the corner shop before it closes...
Still time; bbi10 &
PerlJam jnthn++ TimToady++ I've only been vaguely paying attention today, but it feels like a watershed day for the GLR from here.
19:30 [particle] joined 19:31 colomon left
lizmat PerlJam: I would agree with that assessment :-) 19:34
19:34 ChoHag left 19:38 colomon joined
lucasb so... that gist will be merged into rakudo with all those GLR prefixes removed? 19:39
itz is 29 today!
TimToady we might merge in with the prefixes if we want to use the old stuff as scaffolding while we transform things, but that would be a temporary expedient if we did 19:40
virtualsue . o O ( :20<29> )
19:41 EINVAL left
TimToady fondly remembers the days when he was running a version that contained both the P4 and P5 interpreters, and converted instructions one at a time from the P4 engine to the P5 engine 19:41
PerlJam itz: happy birthday!
itz :)
moritz its++ # 30 now :-)
nine itz is lucky moritz typoed that ;) 19:42
jnthn yowser, warm corner shop was warm...
jnthn dumps one of the beers into the freezer so it'll be a nice temperature by the time he wants it
moritz erm, itz++ :-)
itz I need to write Acme::Birthday::Flattering::Base :) 19:43
19:43 FROGGS left
smls_ whats the easiest way to get a Date for the current date again? 19:44
19:44 yqt joined
jnthn m: say Date.today # total guess 19:44
camelia rakudo-moar 172b87: OUTPUT«2015-08-04␤»
jnthn win1
smls_ thanks
itz examples.perl6.org/categories/cookb...-date.html
TimToady the easiest way to get a date for the current date is to already have an SO :P
19:44 colomon left
jnthn Doesn't help when SO is the other side of the visa application process :( 19:45
moritz :(
TimToady troo :(
sorry, wasn't thinking
moritz (political borders)-- 19:46
[Coke] Is it too late to complain that the default for a new Date isn't today? ;)
jnthn is still hopeful we might get it tied up so she can enjoy the trip to Olten and Granada. :) 19:47
moritz [Coke]: it's never too late for complaints
RabidGravy itz, yoof!
lucasb yeah, Date.new should be the same as Date.today :)
moritz debian-WTF of the day: [ ok ] Reloading web server config: apache2 not running.
all I could think of was "WHAT? web server down?" 19:48
turns out it's running all right
TimToady apache2 not running YET!
19:49 colomon joined
moritz still like the Date.new christmas egg, but guesses he'll lose it in the long run 19:49
PerlJam moritz: you're meant to lose in the long run. :) 19:50
TimToady not such a long run for me anymore... :)
dalek c: 15174dd | smls++ | lib/Language/subscripts.pod:
Some rewording; make "Nonexistent elements" a separate section
 19:51
19:52 mohij_ joined, domidumont joined 19:53 mohij left
skids
.oO("ifdef" could be generalized to some sort of leapfrog topicalizer. Because .defined is not the only thing one might want to do.)
 19:54
TimToady maybe it should default to tomorrow: "a new Date is dawning"
RabidGravy o/ maybe tomorrw, maybe some day ... o/ 19:55
TimToady maybe it should be like jam, and randomly pick yesterday or tomorrow, just to keep people honest 19:56
nine No boom today. Boom tomorrow. There's always a boom tomorrow.
geekosaur .oO { that would be a Date-um } 19:57
smls_ skids: It would need to accept three arguments though (topic expression; truthiness checking expression; block)
[Coke] nine++ 19:58
smls_ skids: So how would that even look like?
timotimo TimToady: return a junction of yesterday and tomorrow "at the same time"
skids smls_: I dunno, I'm thinking about it. Whatever might play a role. 19:59
PerlJam smls_: conditionally *.defined $thing { say "wha?" } # maybe shortend that to "cond" ;) 20:00
smls_ so much TTIAR... :P
maybe 'if' could accept traismods like sub definitions do? 20:01
if $str.index($needle) id bood(*.defined) -> $index { ... }
s/id bood/is bool/
TimToady yes, I thought about those a lot, but they're awfully unweildy, expecially in the case of .defined 20:02
and you have to special-case the 'if' anyway 20:03
smls_ 'while' could support it too :) 20:04
TimToady but doing that with while doesn't help index
smls_ right
TimToady do you have any use cases for methods that are not .so or .defined? 20:05
they seem pretty rare to me
smls_ you're probably right, .defined does seem to be where I usually hit that problem. 20:06
PerlJam .elems perhaps?
lizmat it occurs to me it works like a sort of grep
TimToady PerlJam: .so works there
we really don't have the problem with lists because you can just use 'for' instead of while and actually use grep 20:07
but we don't have scalar control analogs to andthen, as we do with and 20:08
which, as I say, are largely about determining success
nine m: sub str-index(Str $str, Str $find) { my $i = $str.index($find); return False unless defined $i; return True, $i; }; if str-index("foobar", "foo") -> [$found, $i] { say "found at $i!" }
camelia rakudo-moar 65915b: OUTPUT«found at 0!␤»
tony-o thats cool 20:09
TimToady shades of Go 20:10
nine exactly
20:10 domidumont left
TimToady good reaso to avoid it :) 20:10
*son
lucasb m: say &defined; say &defined ~~ Callable
camelia rakudo-moar 65915b: OUTPUT«Sub+{<anon>}.new␤False␤»
nine I'm just wondering if all the fuss about such a tiny little use case is actually worth it
TimToady I repeat, it's not about index, it's about measuring success 20:11
(shades of Icon)
lucasb &defined didn't bound in a &c param in my sub. it doesn't know it is a callable?
[Coke] we just need to decide if Index is going out in the next release or if we kill it. 20:12
TimToady m: say &defined ~~ Callable
camelia rakudo-moar 65915b: OUTPUT«False␤»
TimToady huh
lucasb m: sub f($x,$c,&d) { $c($x) and &d($x) }; f(0,&defined,&say)
camelia rakudo-moar 65915b: OUTPUT«0␤»
lucasb ^^ anyway, I wanted to do this
TimToady m: say defined 42
camelia rakudo-moar 65915b: OUTPUT«True␤»
TimToady m: say &prefix:<defined> ~~ Callble 20:13
camelia rakudo-moar 65915b: OUTPUT«5===SORRY!5=== Error while compiling /tmp/aY486OlzZC␤Undeclared name:␤ Callble used at line 1. Did you mean 'Callable'?␤Undeclared routine:␤ &prefix:<defined> used at line 1␤␤»
TimToady m: say &prefix:<defined> ~~ Callable
camelia rakudo-moar 65915b: OUTPUT«5===SORRY!5=== Error while compiling /tmp/lQApjTlKBd␤Undeclared routine:␤ &prefix:<defined> used at line 1␤␤»
TimToady m: say defined.^mro
camelia rakudo-moar 65915b: OUTPUT«5===SORRY!5===␤Argument to "defined" seems to be malformed␤at /tmp/916F3dNr6H:1␤------> 3say defined7⏏5.^mro␤Other potential difficulties:␤ Unsupported use of bare "defined"; in Perl 6 please use .defined if you meant $_, or use an e…»
TimToady m: say &defined.^mro
camelia rakudo-moar 65915b: OUTPUT«(Sub+{<anon>}) (Sub) (Routine) (Block) (Code) (Any) (Mu)␤»
TimToady obviously a bug, if it's not Callable
lucasb TimToady++, thank you 20:14
ShimmerFairy I think 'ifdef' would work fine as a keyword, though admittedly I'd like it if we managed to find an actual word for it :) . I though 'with' was another decent choice (perhaps with $x { ... } orwith $y { ... } else { ... } ?)
20:15 zakharyas joined
TimToady we also have "without" as the unless/or analog 20:15
so it fits rather nicely
lucasb: you can probably use *.defined there instead 20:16
ShimmerFairy true, that's a nice bonus :)
20:17 darutoko left
tony-o when $x { } 20:17
TimToady taken
tony-o ah yea
with $x { } orwith $y { } without { } 20:18
TimToady we don't use "unless" to mean "else" though 20:19
ShimmerFairy Also another point for having another topicalizer: we do have a defined version of || in // , so it's clear that we've already decided defined-ness was important enough to get its own symbol in some places :)
tony-o our would without be when $x isn't defined
20:19 mohij_ left
ShimmerFairy without $x { } -> unless $x.defined { } , is the current thinking 20:20
TimToady but you've got both an $x and $y
20:21 mohij joined
TimToady anyway, else can work fine there, methinks 20:21
ShimmerFairy my 'orwith' idea could be called 'elswith' to fit with both my usage of 'else' in that hypothetical statement, as well as correlate with 'elsif'
TimToady thought of elswith earlier in the day, but I kinda like orwith better
ShimmerFairy
.oO(we could always go with 'otherwise' if we really want 'else' to stay with just one construct :P)
 20:22
TimToady: my only concern with 'orwith' is that it's a bit similar to 'orelse' , but I don't think that'd be too much of a problem in reality. 20:23
tony-o i like 'without'
TimToady me either
you don't want to have a 'without' continuer if there's also a 'without' statement control; it'd be terribly ambigous 20:24
*guous
tony-o ambigiguous
TimToady I suppose you could only get an ambigibous moon on a planet with two suns 20:25
or is that amphigibous... 20:26
tony-o lol
skids ponders how much semantic difference there really is between Int $a where {...} and "take this value if the same value -> $_ {...}" 20:28
So like, if $a where *.defined -> $a { } 20:29
ShimmerFairy TimToady: that might actually read more naturally than some uses of 'unless' I can imagine there are currently: sub foo($fname? is copy) { $fname = "default.txt" without $fname; ... } (ignoring the fact that this example should really be setting a default in the signature :P)
TimToady jnthn: I really don't know what to do with (1,2;3,4) yet, I do know that .[1,2;3,4] needs to treat the 1st dimension as non-lazy/non-flattening and the 2nd dimension as lazy/flattening, just as a normal .[1,3,5..7] needs slices to flatten when there's no ;
or maybe there's a ; implied in a subscript 20:30
ShimmerFairy: and ignoring the fact you'd just use //= there :P 20:31
skids: seems kinda backwards, insofar as the 'if' isn't testing $a 20:33
and it's still a special case
(or a 'but'-like cheat of some sort) 20:34
20:37 leont left
dalek kudo/nom: 5da3237 | lizmat++ | src/core/control.pm:
Implement leave() placeholder sub
 20:39
20:39 virtualsue left 20:41 lucasb left
skids waitasec... we already have this. 20:41
m: do { say $_ } if .defined given 0 but True
camelia rakudo-moar 65915b: OUTPUT«0␤»
TimToady we have it in spots, but it's not handy 20:42
and, as I said, you either have to rely on a special case or a cheat 20:43
here you're relying on a cheat
ShimmerFairy TimToady: btw, yesterday you said flattening was recursive, but jnthn's gist has my @e = flat @a, @b; # 5 elements [1, 2, 3, $[1, 2], $[3, 4]] in the comments. Am I misunderstanding 'flat' here, or...? 20:44
TimToady remember that arrays itemize each element
ShimmerFairy m: do { say $_ } if .defined given 0 # don't need the 'but True' in this case
camelia rakudo-moar 65915b: OUTPUT«0␤»
TimToady and flat still respects Scalar
jnthn What TimToady++ said. flat (1, (2, (3, 4), 5)) is 5 elements 20:45
ShimmerFairy Ah, so in what context would 'flat' be recursive as opposed to Slip (which is just one level)? List of Lists or Array of Arrays?
20:46 lizmat left
TimToady jnthn: so I guess in glr terms I'm thinking (stuff; and; nonsense) is a List[Seq] (assuming List is a punnable role here) 20:49
20:49 virtualsue joined
TimToady (where the Seq in question is really the flattener, which could obviousl be optimized away in simple cases) 20:56
jnthn ShimmerFairy: only lists of lists 20:58
TimToady: Hm, will have to ponder that a bit
20:58 zakharyas left
ShimmerFairy thanks, that'll be good to remember :) 21:00
jnthn Gee, that refactor took some doing... 21:05
21:06 skids left
jnthn has slayed the "assigning an infinite array to itself" issue 21:06
TimToady slayed...how...
21:06 lucasb joined
jnthn TimToady: Every problem can be solved by another level of indirection... :) 21:06
TimToady: It's basically just making sure that the version of things the iterator is reifying doesn't get confused with the current version of things in the array 21:07
It also let me move some methods out of public API on List that I really didn't think should be there 21:08
So, all win, I think.
21:08 pochi joined
TimToady just so we can defining the Hamming sequence in terms of itself someday... 21:09
21:09 virtualsue left
timotimo that'd be swell 21:09
TimToady probably entails avoiding List and doing it all with Seq somehow 21:10
jnthn m: my @a = 42 xx *; @a = @a; say @a[^10] # curious 21:11
camelia rakudo-moar 65915b: OUTPUT«This type does not support positional operations␤ in block <unit> at /tmp/aYNHS3Ehov:1␤␤»
jnthn yowser
21:11 mohij_ joined
jnthn That one works with what I have locally :) 21:11
TimToady segoi
21:11 mohij left 21:13 rurban left, rurban_ left
timotimo lisuto desu ka 21:13
lucasb m: say $_ ~~ Callable for &defined, &item, &elems 21:14
camelia rakudo-moar 65915b: OUTPUT«False␤False␤False␤»
lucasb ^^ What these functions have in common is that they all have traits, further than this I don't know, but I think maybe this is those kind of issues that only happen in core, because of bootstrapping problems
TimToady m: &defined does Callable; say &defined ~~ Callable 21:16
camelia rakudo-moar 65915b: OUTPUT«True␤»
TimToady looks like an easy fix :)
RabidGravy :-O
jnthn m: my @a = 42 xx *; @a = @a.map(* + 10); say @a[^10] # curious
TimToady we don't advertise the 'does' operator very much...
camelia rakudo-moar 65915b: OUTPUT«(timeout)» 21:17
jnthn This is probably in "getting too cute" territory, but I have that one working too :P
TimToady well, that one works in P5, except for the infiniteness
21:17 muraiki left
ShimmerFairy The more you can use infinite lists where you're not explicitly reaching for the last element (or all elements) the better, I say :) 21:17
jnthn m: my @a = 42, 43, 44; @a = @a.map(* + 10); say @a[^3] # pretty sure it works in Perl 6 already if you're not infinite :P 21:18
camelia rakudo-moar 65915b: OUTPUT«52 53 54␤»
jnthn It's easy to get right for finite things.
tony-o what happened to froggs?
TimToady P5 actually optimizes away the copy if the LHS and RHS share no identifiers
but I suppose there could be aliasing 21:19
RabidGravy TimToady, I thought it was in the docs.perl6.org/language/objects but apparently not, will do tomorrow :)
timotimo m: class Test { }; multi sub infix:<~~>(Test $a, Test $b) { say "oh lord!" }; say Test.new ~~ Test.new 21:20
camelia rakudo-moar 65915b: OUTPUT«False␤»
jnthn ~~ is too special for that 21:21
21:21 mohij joined
jnthn gist.github.com/jnthn/aa370f8b32ef98e4e7c9 # latest GLR stuff with assignment to infinite self fixes 21:21
timotimo i just wanted to make sure it doesn't accidentally actually work and the error message in that PR we just got could be wrong
jnthn lemme glanc ethe pr 21:22
21:22 mohij_ left
tony-o the p6er formally known as froggs 21:22
jnthn I love the name of that hash...and yes, patch looks good
tony-o: Last I heard from FROGGS++, life is busy with $other-things 21:23
tony-o: I'm sure I saw him come/go on IRC earlier today at least...
tony-o what is a life without perl6
and broken bones
bartolin_ jnthn++ # GLR, nice hash names ... *
21:24 kaare_ left
jnthn tony-o: Any of us who've been working on it for a while have needed to take breaks now and then. :) 21:24
jnthn certainly has
tony-o i have too :-), and i'm only hacking away on modules
i broke some bones while on rest month 21:25
RabidGravy I have to take a break from all forms of software every now and then
tony-o well, separated them
jnthn Ouch :(
21:26 rindolf left
tony-o now i need to figure out which ones of my modules are dead or dying 21:26
RabidGravy I'm quite astonished that I reached a 71 day streak on gh
tony-o nice
ugexe also does that 21:27
jnthn Think I'm doing GLR hacking for today
tony-o oh god, everything is broken
RabidGravy 100 will conveniently come a few days before we go to Ibiza ;-)
jnthn Will do the GLR'd gather/take tomorrow
tony-o HTTP::Server::Async is being brought back up to snuff, writing a blog about it during the process is painful 21:28
TimToady ++jnthn with jnthn++
tony-o ugexe.com/refactoring-httpserverasync/
TimToady: lol
21:28 mohij_ joined 21:29 mohij left
[TuxCM] blog.plover.com/prog/monad-search-2.html <= something for perl6? 21:29
21:30 jjido joined
RabidGravy I'm actually wondering whether roles could do with a whole page in the docs of their own, there is whole swathes of stuff not touched at present - as TimToady pointed out runtime application, but also parameters and so fort 21:31
smls_ /me still doesn't get why mutable Array needs to support infinitness/laziness in the first place. i.e. what it gains us that can't be done with List + scalar vars or List + array vars + binding
RabidGravy will look in the morning 21:33
jnthn smls_: Mostly that normal people tend to like assignment
tony-o did normals
jjido smls_: do you think that should be limited to non-mutable array?
smls_ mutable List's, yeah
tony-o oops, did panda break today?
smls_ immutable, I mean 21:34
jnthn smls_: Also, we basically get it for free, given Array ~~ List
smls_: At this point it'd actually be more difficult/more code to lose the capability
jjido what is the logic of having a "*" in $*PID? 21:35
smls_ jnthn: Even counting any future work needed to make Array as fast as in Perl 5? :P
TimToady jjido: all globals are accessed via the dynamic variable system, unless you want to say $PROCESS::PID instead 21:37
but for most of these, it's beneficial to occasionally allow lying to yourself in a particular dynamic scope
RabidGravy TimToady++ beat me to it 21:38
jjido there is a dynamic variable system?
TimToady well, yes, or I wouldn'ta mentioned it... 21:39
jnthn smls_: I'm not convinced Array potentially supporting laziness is where the performance challenge is.
21:39 khw left
RabidGravy well you might just be making shit up in the expectation that someone will make it work before it gets tested 21:40
jjido TimToady: all right I will find the relevant doc myself...
TimToady RabidGravy: but that would make me a retardo :)
jnthn smls_: I mean, "are we assigning to a not-yet-reified lazy area of the array" probably JITs into a a few machine instructions 21:41
smls_ jnthn: As for "people tend to like assignment": I, for one, like certainty. An assignment that magically happens to work now but can be caused to hang indefinitely through action at a distance (e.g. input no longer considered "known infinite"), seems kinda scary to me.
jnthn smls_: And for access it's just part of the bounds check.
TimToady smls_: then don't poke infinite lists into Arrays :P
RabidGravy TimToady :-O
smls_ TimToady: So why support it?
TimToady because it's nice when you need it 21:42
jnthn fwiw, the .infinite handling should be a decent bit more transitive
smls_ especially scary since it doesn't seem to be very precisely defined when Perl 6 should consider something "known infinite".
If there will be competing implementations besides Rakudo, will an array assignment that works in one hang indefinitely in the other? 21:43
TimToady there's always hope
jnthn Just 'cus it hasn't so far been precisely defined doesn't mean it can't be.
timotimo m: say do while (^3.pick) { print "." }
jnthn I suspect it's not been written down 'cus we know the current answers there haven't been good enough.
camelia rakudo-moar 5da323: OUTPUT«(timeout)Potential difficulties:␤ Precedence of ^ is looser than method call; please parenthesize␤ at /tmp/eo9Bap3ghO:1␤ ------> 3say do while (^37⏏5.pick) { print "." }␤...............................................................…»
timotimo oh, hehehe 21:44
TimToady timotimo: prec
timotimo m: say do while ((^3).pick) { print "." }
camelia rakudo-moar 5da323: OUTPUT«.0␤»
timotimo m: say do while (!(^2).pick) { print "." }
camelia rakudo-moar 5da323: OUTPUT«False␤»
jnthn With the GLR, .infinite moves into Iterator rather than being on List
jjido dynamic scoping ew
timotimo are you all fine with me throwing out the return value of the while loop to be the argument evaled to decide to stop?
we'll change the return value of while during the GLR anyway 21:45
jnthn timotimo: I think Rakudo already emits while loops with a Nil afterwards anyway
timotimo doesn't seem so, look at my eval results up there 21:46
jnthn timotimo: ergh 21:47
timotimo: That's REALLY wrong though
timotimo :)
jnthn We're meant to return the list of things in that context
timotimo yeah
TimToady we never return conditionals in P6; that's P5-think
timotimo well, "nqp::while" was implemented that way so far 21:48
jnthn TimToady: Yeah, the code-gen somehow ended up that way.
timotimo and so we ended up p6sink-ing all the flags to while loops all over the world
TimToady it seemed like a good idea at the time...
Sgeo Instead of working now, maybe known-infinite stuff should crash 21:50
So that known-infinite = crash and infinite loop = hang, they're still both failing
jnthn Sgeo: You misunderstood "known infinite"
Sgeo ? 21:51
jnthn It means "we know this thing is infinite, be sure not to try and reify all of it"
TimToady it's more like known arbitrarily long
Sgeo jnthn, I think that shouldn't be a success though, it should indicate failure or be indistinguishable from hanging, ala Haskell
HAskell's undefined causes a crash if used, but infinite loops can cause a hang if used, they're considered the same sort of thing (bottom) 21:52
jnthn Sgeo: I don't follow, but I don't think you grok what we're talking about well enough.
TimToady Haskell certainly allows infinite lists in the sense we're discussing
Sgeo But if you try to find the length of one, it loops infinitely 21:53
TimToady m: say (1..*).elems
camelia rakudo-moar 5da323: OUTPUT«Inf␤»
jnthn This is about how my @b = 1..10; my @a = @b; @b[0]++; say @a; doesn't end up with 2 because we were too lazy, while my @a = 1..*; can work too 21:54
Sgeo TimToady, if 1..* was not known infinite (but actually infinite), would it hang, or eventually output something?
RabidGravy all sorts of things have "infinite" lists in the sense there is no known upper bound when asked
TimToady yes, then you'd might have troubles, but we're not so interested in catching that at compile time 21:55
that failure mode is usually quite obvious at run time :)
Sgeo TimToady, what if it's only obvious in one implementation but another it works and returns Inf?
21:56 mohij_ left
TimToady then our tests are not testing very well 21:56
ShimmerFairy jnthn: about ~~ being too special; I reported a bug on your own multi infix:<~~> not working while investigating the fact that ~~ doesn't chain, and they both appear to arise from ~~ being handled specially. 21:57
Also, is there a provision for not being able to override ~~ in the spec? It was my understanding that operators were generally override-able. 21:58
jnthn ShimmerFairy: We've already patched it to disallow defining a multi infix:<~~>
So you don't get surprised.
21:58 TEttinger joined
jnthn No, not all of them 21:58
:= and = are two other examples of oens that aren't
TimToady Sgeo: note also that it's pretty easy to mark a list as infinite just by putting something infinite at the end
jnthn ~~ can't actually be implemented just as a multi sub 'cus it needs to set $_ to the RHS while it evaluates the LHS 21:59
ShimmerFairy jnthn: yeah, I saw them in the %specials hash, along with the flipflop ops and so on :) Unfortunately, I don't know how to make ~~ and !~~ chain like they're supposed to (trying to change the custom QAST so it calls chain caused the optimizer to crash IIRC)
Sgeo What is the scope of flip-flops? Same as state variables? 22:00
jnthn I think if you got that through the optimizer it'd still be problematic at code-gen...
Sgeo: Yes
ShimmerFairy (and removing ~~ and !~~ from the %specials hash predictably didn't work either, causing memory usage to climb until I had to kill it)
jnthn ShimmerFairy: Things aren't in the specials hash just for the fun of it ;)
ShimmerFairy jnthn: if it's too hard to make ~~ chain, and there wasn't a need for it, I wonder if it would be alright to make it the one non-chaining op at its precedence level. 22:01
jnthn ShimmerFairy: I'm not sure it's too hard...I just find it really hard to even think about what it means :)
Given the thingy that goes on with $_
TimToady perhaps we could notice a ~~ in a chain and just call .ACCEPTS without topicalizing
ShimmerFairy jnthn: basically, 1 ~~ 2 ~~ 3 -> (1 ~~ 2) && (2 ~~ 3) , like with the other chaining ops :) 22:02
TimToady since chains are for values, not fancy drancy RHSs
jnthn ShimmerFairy: Yeah, but chain makes sure to only evaluate the 2 once there
timotimo super helpful! :) 22:03
ShimmerFairy jnthn: ah, I see. (I looked at the code for nqp::chain , but only understood enough to see that it probably wasn't the problem in making ~~ potentially chain :P) 22:04
smls_ m: say ("aa123456bb" ~~ /\d+/) ~~ /3./ # left-associative would be more useful
camelia rakudo-moar 5da323: OUTPUT«｢34｣␤»
ShimmerFairy timotimo: I imagine something like 0 < $x ~~ Int being a nice shorthand for "non-zero positive integer"
jnthn: I think the worst case would be defining a custom 'smartmatch_chain' variant of chain for the precedence parser to use, perhaps (not that I know much about the uglier side of compiler writing ☺) 22:06
TimToady I think we should simply have a default non-topicalizing implementation of infix:<~~> and then treat the bare binary as a special syntactic case that also topicalizes 22:08
jnthn That works for me
TimToady then you just have to be careful to use @list X~~ *.method rather than .method 22:09
smls_ Why remove topicalization altogether? Does ~~ /regex/ need it somehow? 22:11
ShimmerFairy btw, am I right in thinking that the infix:<~~> in Mu.pm doesn't get picked up as an operator, just like custom multis?
smls_: yes, regexes match against $_ (try a bare /regex/ in a for loop ☺)
m: for <foo bar baz> { say /^b/ }
camelia rakudo-moar 5da323: OUTPUT«/^b/␤/^b/␤/^b/␤»
ShimmerFairy m: for <foo bar baz> { /^b/; say $/ } 22:12
camelia rakudo-moar 5da323: OUTPUT«Nil␤｢b｣␤｢b｣␤»
jnthn ShimmerFairy: I suspect it is used for meta-ops
smls_ ShimmerFairy: But they also have an .ACCEPTS, don't they?
jnthn Z~~ and so on
ShimmerFairy smls_: I believe so
smls_ So would still work without $_
ShimmerFairy m: class Foo { }; multi sub infix:<~~>(Foo $a, Foo $b) { 42.6 }; say Foo ~~ Foo; say Foo R~~ Foo; 22:13
camelia rakudo-moar 5da323: OUTPUT«True␤42.6␤»
ShimmerFairy jnthn: ^ I think you're right :)
jnthn smls_: 'foo' ~~ m/xxx/ needs it though, iirc
smls_ Ah 22:14
jnthn I need to get some rest :)
smls_ s/// to then probably
*too
jnthn 'night, #perl6
TimToady o/
ShimmerFairy ♞, jnthn o/
RabidGravy toodlepip
lucasb Is '$x ~~ .method' supposed to call $x.method? 22:15
timotimo no
er
yes
lucasb yes or no? :D
timotimo yes. 22:16
lucasb ok then :)
m: say 65 ~~ .chr
camelia rakudo-moar 5da323: OUTPUT«False␤»
lucasb m: say 65 ~~ *.chr
camelia rakudo-moar 5da323: OUTPUT«A␤»
lucasb What should the return value of '65 ~~ .chr' be? 22:17
TimToady False
it calls "A".ACCEPTS(65) 22:18
ShimmerFairy m: say 65 ~~ .Str
camelia rakudo-moar 5da323: OUTPUT«True␤»
TimToady "65".ACCEPTS(65) is True, yes
lucasb oh, thanks; I guess my brain is tired 22:21
TimToady --> nap & 22:28
22:28 spider-mario left
timotimo TimToady ~~ .nap 22:28
dalek p: 1b257a1 | timotimo++ | src/vm/moar/QAST/QASTCompilerMAST.nqp:
a slightly more verbose error message for localref
 22:29
p: 808ce0f | timotimo++ | src/vm/moar/QAST/QASTOperationsMAST.nqp:
throw out nqp::while return value

it used to return the value of the condition, but that doesn't help rakudo with its semantics. instead, it caused every condition of a while loop to have .sink attempted on it, even in cases where we know it doesn't make sense.
so now we either have the register be an object that's null'd or set the return kind to void.
Also, remove fossilized IMM_ARG mention.
22:31 smls joined, smls_ left 22:34 lucasb left 22:47 skids joined
RabidGravy right bed 22:56
turrah
22:59 jjido left, kurahaupo left 23:01 RabidGravy left 23:09 lizmat joined 23:11 baest left
dalek kudo/nom: 4332300 | usev6++ | src/Perl6/Grammar.nqp:
Complain when trying to override smartmach operator
 23:13
kudo/nom: b6f6c1f | lizmat++ | src/Perl6/Grammar.nqp:
Merge pull request #487 from usev6/rt125745

Complain when trying to override smartmach operator
23:16 virtualsue joined 23:17 telex left
dalek kudo-star-daily: ebe7644 | coke++ | log/ (9 files):
today (automated commit)
 23:17
rl6-roast-data: a1e4669 | coke++ | / (9 files):
today (automated commit)
23:18 telex joined 23:23 baest joined 23:24 EINVAL joined
dalek kudo/nom: 2f7ab34 | lizmat++ | src/core/Distro.pm:
Don't let absence of lsb_release spoil things

See: blogs.perl.org/users/aaron_baugher/...tos-6.html		 23:27
lizmat good night, #perl6! 23:28
23:29 smls left
EINVAL Night lizmat 23:30
23:30 baest left
timotimo goes to sleep, too 23:31
23:42 Zoffix left 23:47 Zoffix joined
[Coke] ~~ 23:48
23:49 baest joined 23:51 EINVAL left 23:52 Zoffix left 23:53 captain-adequate left
[Coke] r: (0, {last if $^a==$^b; (1..5).pick} ... *).perl.say 23:55
camelia rakudo-{moar,jvm} 2f7ab3: OUTPUT«(ListIter.new,)␤»
23:55 Zoffix joined 23:56 baest_ joined 23:57 baest left

Powered by Cro and the Raku® Programming Language.

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

Thank you!