»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.
00:02 cognominal left
timotimo a rakudo built with mast_localref_3 (which i just rebased onto latest nqp) spectests fine 00:21
i mean ... why wouldn't it
but i'm still checking to be sure
something's wrong ... "perl6" doesn't start any more .. ?! 00:24
perhaps i need to Configure.pl a few times again 00:25
indeed, that was it 00:27
dalek p: d125c71 | timotimo++ | / (2 files):
initial implementation of localref X local access

as well as three simple failing tests		 00:31
00:31 dalek left, Averna joined
timotimo poor dalek 00:31
00:32 dalek joined, ChanServ sets mode: +v dalek 00:45 cognominal joined
dalek kudo/optimizer_lexicalref_lowering: cbe2eca | timotimo++ | src/Perl6/Optimizer.nqp:
the first obvious changes towards lowering lexicalref
 01:08
timotimo jnthn: ^- this branch currently gives me Stage mast : Cannot take a reference to non-native local __lowered_lex_723 01:09
(the name of the local gets output in the exception via a local patch)
maybe you remember what obvious part i've missed that'd cause references to obj primspec'd locals be emitted
.o( and also: shouldn't we actually be able to handle obj registers be localref'd? ) 01:10
i'll try to get some good rest now
01:11 rangerprice joined 01:19 ggoebel joined, rmgk_ joined, rmgk is now known as Guest8167, Guest8167 left, rmgk_ is now known as rmgk 01:23 ggoebel left 01:30 TEttinger joined 01:38 amurf left 01:40 Quom left, [Coke] left, Util left 01:41 dalek left, sergot left, masak left, PerlJam left, pmichaud_ left 01:44 BenGoldberg_ joined 01:46 dalek joined, ChanServ sets mode: +v dalek 01:47 JimmyZ joined, avuserow joined 01:48 Mouq joined, Util joined, [Coke] joined, sergot joined, masak joined, PerlJam joined, pmichaud joined, psch joined, masak is now known as Guest11116 02:03 chenryn joined 02:07 chenryn left 02:08 chenryn joined 02:11 colomon joined 02:13 amurf joined 02:14 colomon_ joined 02:16 colomon left, colomon_ is now known as colomon 02:19 colomon_ joined 02:20 colomon left, colomon_ is now known as colomon 02:21 rangerprice left 02:25 colomon_ joined 02:26 colomon left, colomon_ is now known as colomon 02:29 laouji joined 02:30 colomon_ joined 02:31 colomon left, colomon_ is now known as colomon 02:34 colomon_ joined 02:35 colomon left 02:37 colomon joined 02:38 yqt left, colomon_ left 02:41 AlexDaniel left, colomon left 02:42 colomon joined 02:44 cognominal left 02:45 colomon_ joined 02:47 colomon left, colomon_ is now known as colomon 02:51 colomon left 02:54 noganex joined 02:57 noganex_ left 03:02 yeahnoob joined 03:24 schmooster left 03:27 cibs joined
census why did cognominal leave? 03:30
03:30 gfldex left 03:32 BenGoldberg_ left
TEttinger census: cognominal has quit (Quit: This computer has gone to sleep) 03:43
03:46 dayangkun left 03:48 atroxaper joined 03:52 laouji_ joined, laouji left 03:54 dayangkun joined 04:10 schmooster joined 04:13 chenryn left 04:18 Averna left 04:43 chenryn joined 04:53 [Sno] left 04:54 atroxaper left 04:59 sftp_ joined 05:02 khw left 05:05 sftp_ left 05:07 sftp joined 05:12 chenryn left 05:16 atroxaper joined 05:31 mr-foobar left 05:32 laouji_ left 05:33 laouji joined 05:36 atroxaper left 05:37 laouji left 05:40 cognominal joined 05:42 chenryn joined 05:43 atroxaper joined, laouji joined 05:48 diana_olhovik_ joined 05:57 baest_ is now known as baest 06:01 atroxape_ joined 06:03 atroxaper left 06:04 atroxaper joined, atroxape_ left 06:13 jkva joined 06:17 [Sno] joined 06:22 FROGGS__ left 06:23 atroxaper left 06:31 atroxaper joined 06:46 atroxaper left 06:47 atroxaper joined 06:51 Woodi left 06:52 Woodi joined 06:53 FROGGS joined
FROGGS o/ 06:53
nwc10 \o 06:55
06:56 zakharyas joined 07:02 rurban joined 07:04 kaare_ joined 07:06 amurf left 07:08 darutoko joined 07:13 espadrine_ joined
Guest11116 morning, #perl6 07:14
07:14 Guest11116 is now known as masak 07:17 atroxaper left
FROGGS morning masak_grrr 07:17
:o)
masak :) 07:18
apparently I missed the _grrr part this time.
07:18 telex left
jkva I think most of us are _grrr on mondays :) 07:19
nwc10 good Monday, #perl6 :-)
lizmat good *, #perl6! 07:20
FROGGS nwc10 masak jkva o/
07:20 telex joined 07:21 mr-foobar joined 07:24 atroxaper joined 07:25 xinming left 07:26 xinming joined 07:27 atroxaper left 07:29 atroxaper joined
lizmat PSA: it appears I'll be doing the P6W again tonight 07:31
masak ++lizmat 07:32
jkva lizmat: P6W? 07:35
lizmat Perl 6 Weekly: p6weekly.wordpress.com
jkva lizmat: Thanks 07:36
nwc10 lizmat: I thought that irclog.perlgeek.de/perl6/2015-07-23#i_10942545 was cool enough to get a P6W mention
07:36 [Tux] joined
lizmat yeah, that one was cute :-) 07:36
masak :) 07:37
07:38 bjz joined 07:41 pdcawley joined
dalek kudo/nom: 9ed640c | lizmat++ | src/core/ (2 files):
Implement .Pos <-> .Int coercing
 07:43
lizmat b2gills: my experiences with branches in rakudo has not been the best
reverting multiple commits is easy
07:44 RabidGravy joined
lizmat re. using a hypothetical "ifdef" statement in stead of having Pos 07:44
so you could say something like:
ifdef "foo".index("f") -> $pos { }
is nice and useful in and of itself, but will not help ternaries 07:45
anyway, I'm holding off writing tests for Pos until we have reached consensus
meanwhile, this patch gives: 07:46
$ 6 'say ?0.Pos'
True
jnthn: regarding batching of method calls using >>.foo
I don't think we can do anything but .cue a job for each method call 07:47
as we can have no idea what that method call will do
for example, if you have a list of URL's that you want to scrape
@urls>>.scrape
each scrape could take several seconds\ 07:48
you're using >>. for its parallelization
you don't expect them to be batched
perhaps we could have a method attribute "is batched(100)" 07:50
masak lizmat: "but will not help ternaries" -- that part was not clear to me. could you please clarify?
lizmat say "foo".index("f") ?? "found" !! "did not find"
masak my $found = "foo".index("f") != -1; 07:52
lizmat re "is batched": that would indicate how you would like to have that method batched in >>. situations
masak say $found ?? "found" !! "did not find"
sometimes I think we're a little bit too eager to reduce those final keystrokes ;)
lizmat masak: that still doesn't solve the issue if $found is an Int and 0 07:53
masak it doesn't solve the semipredicate problem, no
lizmat m: say 0 ?? "found" !! "not found"
camelia rakudo-moar ed945e: OUTPUT«not found␤»
lizmat m: say 0.Pos ?? "found" !! "not found"
camelia rakudo-moar ed945e: OUTPUT«Method 'Pos' not found for invocant of class 'Int'␤ in block <unit> at /tmp/3T2ONNdPBM:1␤␤»
masak you have to make *some* value exceptional to mean "not found" -- unless you go all-out with option types 07:54
lizmat $ 6 'say 0.Pos ?? "found" !! "not found"'
found
masak that value for .index is -1
the API user is expected to know this
lizmat m: say "foo".index("bar")
camelia rakudo-moar ed945e: OUTPUT«(Pos)␤»
lizmat it's *not* -1 07:55
and hasn't been for a *long* time
it's undefined, either from being a Failure or from being a Type object
masak oh, right.
lizmat .index and .rindex actually differed in that, I harmonized that
07:57 azawawi joined
masak yeah, sorry about that. I'm being confused. 07:57
azawawi hi
master of confusion is here :)
masak lizmat: I agree with you that there's a risk of confusion, since 0 and undefined values are both falsly, and that can trip people up. 07:58
lizmat: I'm not too sold on the solution of overloading ints with bools.
lizmat: partly because we've been down that road design-wise before with other things, and it hasn't worked out so well
lizmat such as ? 07:59
was that from the time that "0" was still true ?
masak no.
it was return values from processes, the prototypical example of `0 but True`
ShimmerFairy masak: Well, ideally Pos can be used for strings too (the idea was that StrPos could be useful to more than just strings) 08:00
masak it ended up with TimToady saying "I never spec'd that" :)
08:00 mrf joined
masak ShimmerFairy: I'm wary of premature generality, too. 08:00
lizmat masak: well, it's not mixing in any roles into instances, it is minimally subclassing Int 08:01
masak ("let's abstract into positions of all the things!")
08:01 dakkar joined
ShimmerFairy I'll have to look at the GLR draft again and see what Pos can help in there. 08:01
08:02 atroxaper left, dakkar left
masak solution looking for a problem. 08:02
08:02 dakkar joined
lizmat masak: the reason I did this, was a discussion on #perl6 last week 08:02
it is *not* a solution looking for a problem
masak lizmat: sorry, was unclear. 08:03
ShimmerFairy leafing through GLR in order to motivate it being a general positions-of-things solution, that's a solution looking for a problem.
ShimmerFairy masak: And the problem with finding more uses for something is...?
08:04 Fleurety joined
lizmat masak: "let's not think outside the box, please" ? 08:04
masak lizmat: my primary worry is still that the cure (overloading ints with bools) will be perceived as a bigger wart than the problem it solves (semipredicate problem for .index)
lizmat: I'm all for innovative thinking. I just want to end up with solutions that don't come back and bite us.
lizmat: it *is* confusing to have a 0 that boolifies to True. 08:05
that's what we learned from the return codes thing.
azawawi 0 => True... what ?
lizmat azawawi: 0.Pos => True
masak ShimmerFairy: the problem with finding more uses for something is you've already *decided* that this solution needs to be general enough to support more uses, before you've seen those other uses.
ShimmerFairy: that's what "premature generalization" means. 08:06
ShimmerFairy masak: How else shall I see other uses if not with the hope that I can generalize something? :)
I disagree with the notion that "solution looking for a problem" is _always_ a bad way to figure things out. 08:07
masak ShimmerFairy: that's usually not how you arrive at a solid design. you're supposed to notice a number of uses that can be unified, and *then* generalize. doing it the other way triggers warning bells.
lizmat needs to be afk for a few hours, will backlog later 08:08
masak lizmat: \o
ShimmerFairy masak: IIRC I initially agreed with the suggestion to reintroduce StrPos by saying it could even be useful for lists and such. 08:09
azawawi confused 08:10
masak ShimmerFairy: granted, .first-index and such have the same semipredicate problem. 08:11
(but that's still assuming that the proposed Pos solution is something we should adopt) 08:12
I think another part of my worry is that Pos intentionally mixes up two concepts: boolean truth, and definedness.
all in the name of trying to make it easier for the user.
in my experience, it's easier when those concepts are not mixed up. 08:13
08:13 Hor|zon joined
ShimmerFairy masak: One difference between return types and this, is that this is a separate type (and not just a "0 but Bool" being returned). I think you'd be expected to know Pos.new(0) == True in the same way you'd be expected to know index() returns Int:U in its Pos-less version. 08:19
masak well, neither the old nor the new solution is hard to use if you know what you're expected to know. :P 08:20
so the only interesting cases to discuss are the failure modes that happen when the user does not have full knowledge of the API
and in that case, an *undefined* value will warn and make noise in various ways 08:21
a truthy 0.Pos will just silently be accepted everywhere
ShimmerFairy masak: Now knowing what "semipredicate problem" means, that's something I try to avoid all the time (and I hate whenever it has to happen). Haskell's "Maybe" type comes to mind as a possible, if unlikely by now, solution to that issue :) 08:23
masak yes, that's why I mentioned option types above
ShimmerFairy I must have missed that. 08:24
masak the semipredicate problem runs deep. any language that has IEEE 754 floating-point values also has NaN, which gets in *everywhere*. 08:26
ShimmerFairy I imagine something like if "foo".index(bar) -> $a { my Int $idx = $a.just; ... } potentially
masak maybe try it out in a fork of rakudo 08:27
ShimmerFairy Maybe I will :)
vendethiel thinks this is the wrong solution to the wrong problem
nwc10 masak: sorry for being slow/unable to read scrollback, but why does NaN end up as causing semipredicate problems? (I'm aware that NaN != NaN etc, and it's pretty much like NULL in SQL, rather than Perl's undef)
ShimmerFairy (Wouldn't be the first time we were inspired by Haskell :P) 08:28
vendethiel The solution should be 1-indexing, that's obvious *g*
masak nwc10: NaN is not on the real number line. it's a floating-point value with the semantics "I am not a floating-point value".
nwc10 masak: aha yes. That's a lovely way to look at it.
masak it's undefined, but for floats 08:29
nwc10 it's not even on the complex plane.
(which is the image in my head when you say "real number line")
vendethiel doesn't know why NaN exists
masak you won't find it among the hypercomplex numbers, either
nwc10: also, NaN + NaN\i is a lovely thing to ponder.
nwc10: it's undefined in two dimensions!
08:30 atroxaper joined
masak vendethiel: IEEE 754 is full of sensible compromises. I'd rage against NaN too, but we live in the real world, which often (rightly) prefers "practical" to "pure". 08:30
vendethiel is tired of hearing that line as well :) 08:31
especially about something designed in '85
masak sorry :/
what's wrong with '85?
I do think we did a good thing standardizing on floats in that way.
vendethiel pretty much everything we did (programming-wise) then has changed. Or at least, it certainly looks like it from my point of view.
masak I don't think all that many people need floats these days, however. 08:32
vendethiel *40 years ago*. If we go the other way, that's 1945, and "programming" is ...
masak I'm not sure what kind of fallacy that is. "if we go twice as far back..."
08:32 bjz left
masak :) 08:32
vendethiel '85 is closer to cobol's birth than to us.
whoops, I meant 30, not 40. anyway, not twice
masak vendethiel: I'd much much rather have IEEE 754 than no standards for floats (and most of the designed by amateurs) 08:33
vendethiel I'm just taking '85 and going backward instead of forward
oha perhaps the point is that NaN makes sense in some scopes, but in a higher level language it is more easy to handle exceptions?
ShimmerFairy I don't see anything wrong with NaN, but then again I'm not the world expert on floating point :)
masak er, "lots of standards", I mean.
ShimmerFairy: you said you tried to avoid the semipredicate problem at all times. ShimmerFairy, meet NaN.
oha or actually, failures
08:33 inokenty joined
RabidGravy in 1985 I only knew COBOL, BASIC and some Assembly language 08:34
masak NaN's purpose is to propagate up your calculations, blanking them all out as it goes.
RabidGravy oh and Logo
ShimmerFairy masak: It's more for when I'm designing something, I really like making sure the failure value can't be a valid value too.
masak ShimmerFairy: there are many situations where one has to express the idea of (a) "there is no value here", or (b) "I don't know this value". 08:36
08:36 g4 joined 08:37 espadrine_ left
masak Ovid has some nice blog posts about how NULL can mess up your SQL logic, because boolean logic invisibly turns into three-value logic. 08:38
and he's right. but there is often a real, practical need for NULL in some columns too.
why? because the world refuses to be as neat and filled with perfect knowledge as we would like. 08:39
DrForr Multivalue screws up the underlying Coad-Yourdon model. 08:40
08:40 daxim joined, yeahnoob left
jkva DrForr: Never seen that before, thanks 08:41
masak DrForr: I didn't know what "Coad-Yourdon model" was, so I looked it up. looks vaguely UML-y.
DrForr I might have the name wrong actually.
It's the mathematical underpinnings to SQL.
And I think actually 'multiset' is the problem I was looking for.
masak you're not thinking of Edgar F. Codd, are you? 08:42
DrForr Codd, yes, thanks.
masak his paper set off the SQL thing.
www.seas.upenn.edu/~zives/03f/cis550/codd.pdf
DrForr One damn character.
jkva Huh, I ended up at conceptdraw.com/a2283c3/preview
masak DrForr: I seem to recall you're right. Codd never liked NULL at all.
DrForr: but the *industry* did, and even though that's not all that great perhaps, I would say it's significant. 08:43
DrForr And of course it made it practical rather than a pure model.
masak aye. 08:44
I would say though that SQL has still benefited from being based on a mathematical model as opposed to, say, OOP.
DrForr Oh, I wouldn't question that. 08:45
08:52 yakudza joined
RabidGravy you can always tell which database engine someone firsted learned SQL on by whether they always tend to do "foo is not null and foo = bar" or not 08:52
08:55 amurf joined
masak unless there are only two database engines, I'm not sure you can :P 08:55
nwc10 country *and* western? :-)
masak www.youtube.com/watch?v=YSTJ5Xe-E8c 08:56
08:58 abraxxa joined
RabidGravy There are only two 08:59
Oracle and Informix
08:59 amurf left 09:00 araujo joined 09:03 laouji left 09:04 bjz joined
RabidGravy there's a comment in IO::Handle that the $whence argument should be an Enum, is there any reason why that hasn't been done? 09:09
masak m: say Pair ~~ Enum # note that the type "Enum" may not be what you think it is 09:13
camelia rakudo-moar 9ed640: OUTPUT«True␤»
RabidGravy oh it is 09:14
unfortunate mis-capitalization 09:15
09:15 espadrine_ joined
masak if I remember my S12 correctly, it encourages people to call the whole thing an "enumeration", and its individual constituent parts "enums" or "enum values". 09:16
this despite the fact that the *keyword* for the whole thing is just `enum`.
a little similar to the fact that the name of the thing you create in Perl 6 using the keyword `given` can be referred to as a "switch statement". 09:17
RabidGravy (just documenting seek() and tell() - I'll leave the other bit to later) 09:19
azawawi m: class Foo { has Int $.a; }; my $o = Foo.new(:b(1),:a(1)).a;
camelia ( no output )
azawawi should rakudo warn about parameter b? 09:20
ShimmerFairy Not currently, and that's thanks to the implicit *%_ on methods :) 09:21
masak azawawi: there's been a persistent wish for that, yes.
azawawi: ShimmerFairy has it right.
RabidGravy also shouldn't IO::Handle.slurp-rest be a multi? (i.e. one with :bin returning a BUf and one without returning a Str) 09:23
ShimmerFairy That actually wouldn't work IIRC, because of implicit *%_
masak RabidGravy: that sounds reasonable, yes.
ShimmerFairy: if does if you mark the :$bin! as required 09:24
RabidGravy indeed
masak m: class C { multi method foo() { say "without" }; multi method foo(:$bin!) { say "with" } }; given C.new { .foo(); .foo(:bin) } 09:25
camelia rakudo-moar 9ed640: OUTPUT«without␤with␤»
ShimmerFairy right, my mistake :)
masak not everything gets screwed up by *%_ :P
RabidGravy I'll do that later (and leave documenting slurp-rest until then.)
azawawi thanks
ShimmerFairy masak: Still haven't seen what good *%_ does either :P 09:26
09:26 laouji joined, laouji left
masak if it were unambiguously bad, we'd have removed it ages ago. 09:26
dalek c: 63fb6eb | RabidGravy++ | lib/Type/IO/Handle.pod:
document seek/tell somewhat
 09:27
09:27 laouji joined
ShimmerFairy masak: It's been complained about for a long time now, and I've never been able to understand (or been shown) what it helps with. 09:27
09:29 Alina-malina left 09:30 araujo left
masak ShimmerFairy: it makes a method in a base class resilient to named arguments it doesn't know but which a deriving class might know what to do with. 09:31
I wish I had a real example to demonstrate it with.
ShimmerFairy Shouldn't it be the other way around? Or are there cases where the base class' methods are tried first? O_o
masak for now, maybe imagine `class Coord3d is Coord2d`, and you can initialize either one with .new(:$x, :$y, :$z) 09:32
ShimmerFairy: it matters for things like `nextsame`
in fact, `nextsame` (I now recall) tends to be the argument that makes *%_ not be removed :) 09:33
ShimmerFairy masak: I think my question then is why isn't .new($x, $y, $z) afforded the same resiliency?
09:34 araujo joined
masak ShimmerFairy: I think it's grounded in the fact that positional arguments and named arguments are used differently in practice. 09:35
a positional is required by default, and tends to be part of the necessary/primary information going into a call.
a named is optional by default, and tends to be used for "extra" information, or tweaks, or settings. 09:36
ShimmerFairy What about optional positionals then, at least? (Imagine if Coord3D made the $z positional optional and defaulted to zero when not supplied)
Couldn't the argument be made that nextsame should treat foo($x, $y, $z?) and foo($x, $y) equivalently? 09:37
09:40 dayangkun left
masak it could. 09:40
but it's probably significant that it hasn't ;)
ShimmerFairy masak: I just think that *%_ has historically caused far more problems than it solves, on top of the fact that I've never even seen a solution showing how useful having implicit *%_ in the lang is :) 09:41
09:42 atroxaper left
ShimmerFairy I'm not so sure the problem of "extraneous nameds" is nearly as prevalent as implicit *%_ would suggest. 09:42
(The other inconsistency question aside from positionals is why subs don't get the same implicit *%_ ? Can't we use the nextsame bunch outside of classes too?)
masak classes have a different story to them. someone else can subclass your base class. 09:43
whereas multi subs and the like are tied to nested lexical scopes, all of which tend to be under your control. 09:44
09:45 espadrine_ left 09:48 atroxaper joined
ShimmerFairy masak: I think my problem with implicit *%_ is that it purports to fix something I've never seen be a problem. I understand the concern to not die on extra, unknown nameds by default. Problem is, I don't know when that ever is likely to happen. 09:49
Only in cases like foo(|c) (where you're passing a collection of things as args), but then you also have to worry about positionals. 09:50
Er, I can only see it potentially happening in cases like that, to be clear. 09:51
RabidGravy I can see a case like 'my %conf = get-config-from-file(); my $foo = Foo.new(|%conf)' where thee config may be shared between multiple types and a given type is only interested in a sub-set 10:01
ShimmerFairy Essentially I don't see nameds and positionals as being fundamentally different (they're just different ways of accepting arguments from callers), so I don't understand why just nameds get this special and unintuitive treatments, in just methods too. 10:03
10:04 espadrine_ joined
masak I find optional positionals to be quite similar to nameds :) 10:06
required positionals, not so much. YMMV
10:06 spider-mario joined
ShimmerFairy masak: I don't doubt their typical usages, btw. I only doubt that those typical usages should be upheld in design to the extent that *%_ does. (Look at the Coord2d/3d example you shared earlier, for an example of nameds not being "extra" info) 10:08
RabidGravy y'see for me as an application designer I would probably favour a multi over optional positionals
10:12 renormalist joined 10:13 renormalist left, renormalist joined 10:17 baest_ joined
jdv79 lizmat: define "batched" 10:19
10:19 Spot___ joined, laouji_ joined 10:20 mr-fooba_ joined 10:21 mr-foobar left, robinsmidsrod left, Spot__ left, baest left 10:22 jkva left, laouji left 10:23 Spot___ is now known as Spot__
jdv79 tadzik: sup wit smoke runs? 10:29
RabidGravy the oom killer got my last smoke on raspberry pi 10:30
nine RabidGravy: so what about "foo is not null and foo = bar"?
10:31 chenryn left
RabidGravy some databases comparison of a null to a non null yields false, some it's undefined 10:33
roughly split between Sysbase/MS-SQL/Informix and the rest 10:34
tadzik jdv79: I have no idea, I don't smoke :)
oha tadzik, you don't smoke? mh... let's see... 10:36
10:36 spider-mario left 10:37 jkva joined
RabidGravy just multi'd slurp-rest in github.com/rakudo/rakudo/pull/480 if anyone cares to take a look 10:40
dalek kudo/nom: 8a46049 | RabidGravy++ | src/core/IO/Handle.pm:
Make IO::Handle.spurt-rest a multi
 10:41
kudo/nom: e3314c8 | FROGGS++ | src/core/IO/Handle.pm:
Merge pull request #480 from jonathanstowe/slurp-rest-multi

Make IO::Handle.spurt-rest a multi
vendethiel lgtm, it seems :)
RabidGravy last great trans-dimensional meatload? 10:42
nine masak: I understand your caution regarding Pos(0) being True. I just really like that it's the first time I see such an API where the intuitive or naive approach to using it is actually the right one. 10:45
Up until that, I've only seen bad solutions and worse. PHP gives you False which is harder to differentiate from 0 in boolean context. Python and Javascript are horrible in that they are both dynamic languages yet feel the need to invent yet another error indicator (-1) when both already have ways to express undefinedness + exceptions 10:47
10:47 rurban1 joined
timotimo o/ 10:47
masak nine: JS gets a lot of flak because it has an == operator in which 0 == "0", 0 == "", but "0" != "". it feels to me that we're making a similar thing, breaking a common expectation somewhere. 10:48
nine: I'm usually quite proud of Perl 6 for making a proper distinction between true/false and defined/undefined 10:49
ShimmerFairy masak: If we think of the index as being defined when it can find what it's looking for, and undefined otherwise, then the problem with something like Int is that you can't simply check if it's a type object or not without creating a temporary variable. 10:53
nine For use cases where the exact position is not needed, I'd rather see a .contains method with a Bool result. That leaves two cases: one where we just want to know the index. And the other, where we don't know yet if the string is actually contained and need the exact index if it is. 10:54
ShimmerFairy masak: if the statement if "foo".index("bar").defined -> $idx { ... } put the actual index in $idx (not necessarily .defined exactly, but something like it at least), I suppose there's a bit less reason for Pos.
timotimo right; that's what lizmat means with "if foo.index(bar) -> $idx { ... }" 10:55
FROGGS m: if "foobar".index("foo").grep(*.defined) -> [$idx] { say $idx; say $idx.WHAT } 10:57
camelia rakudo-moar 9ed640: OUTPUT«0␤(Pos)␤»
timotimo ouch :D
FROGGS it is a workaround, surely
dalek c: e7c25b1 | RabidGravy++ | lib/Type/IO/Handle.pod:
Document IO::Handle.slurp-rest
RabidGravy there, now back to writing my own stuff 10:59
nine Please note, that we currently don't have a .contains, so this needs covering, too: if $string.index("foo") { ... }
timotimo don't we have has-substr?
ShimmerFairy One of my biggest problems in any language is when you basically need two values from something. This feels similar to me to when you want to do for @items -> $item { ... } but you also need the index, so you end up having to do something like for ^+@items -> $idx { $items = @items[$idx] }
10:59 atroxaper left
timotimo m: "foobar".has-substr("oob").say 10:59
camelia rakudo-moar 9ed640: OUTPUT«Method 'has-substr' not found for invocant of class 'Str'␤ in block <unit> at /tmp/6TOlZNfCu2:1␤␤»
timotimo hm
RabidGravy boo
timotimo ShimmerFairy: well, we do have .kv for that particular case
11:00 leont joined
timotimo and python has enumerate(...) together with index unpacking 11:00
other languages ... not so much
ShimmerFairy timotimo: Huh, I didn't think of .kv , probably because that's a very Associative method for something I'm not using in a hash-like fashion :) 11:01
leont Can someone tell me where coersions are documented, and what their current state is?
timotimo ah!
well, we're "proud" of our Cool and Any things
m: say "hello".kv.perl
camelia rakudo-moar 9ed640: OUTPUT«(0, "hello")␤»
timotimo m: say "hello".pairs.perl
camelia rakudo-moar 9ed640: OUTPUT«(0 => "hello",)␤»
ShimmerFairy timotimo: maybe .iv or .ie ("index/value" or "index/element") for the array-related use :P :P 11:02
timotimo not only do single items behave like lists of that single item, our lists also behave like hashes of index to value
ShimmerFairy timotimo++ # reminding me of .kv is gonna help me out a lot in the future, since I seem to do stuff requiring index + value somewhat often :) 11:03
timotimo yay
RabidGravy leont, design.perl6.org/S02.html#Coercive_...clarations 11:05
and currently only work for argument types IIRC
11:05 atroxaper joined
leont Does not appear to work on Array arguments :-s 11:08
timotimo yeah, that won't work
typed arrays are more of a "all or nothing" thing 11:09
11:10 robinsmidsrod joined 11:12 Begi joined 11:13 chenryn joined
RabidGravy I think that is stated explicitly elsewhere in the design docs 11:14
jdv79 tadzik: i meant why is smoke.perl6.org stalled at 7/21 11:16
Woodi hi today :)
FROGGS jdv79: ups 11:18
ohh, unups
I can't fix that
jdv79: ask colomon 11:19
11:20 robinsmidsrod left
Woodi whatever "chaining" is $a ~~ (2..3) is strange. it should be the same with () and without. probably via DWIM. we need GLR to remove DWIM *inside* core but keep it user facing. maybe redefining some user things too... 11:20
11:21 robinsmidsrod joined
Woodi but anyway, it is Perl ! sooner or later someone will remove need for () :) 11:21
m: (2 ~~ 1 .. 3 ~~ 0..4) ~~ True # chaining... 11:22
camelia rakudo-moar e3314c: OUTPUT«5===SORRY!5=== Error while compiling /tmp/slSEl48ldW␤Operators '~~' and '..' are non-associative and require parentheses␤at /tmp/slSEl48ldW:1␤------> 3(2 ~~ 17⏏5 .. 3 ~~ 0..4) ~~ True # chaining...␤ expecting any of:␤ infix…»
Woodi m: (2 ~~ (1 .. 3) ~~ (0..4)) ~~ True # chaining... 11:23
camelia rakudo-moar e3314c: OUTPUT«5===SORRY!5=== Error while compiling /tmp/QIMZ1PXu50␤Operators '~~' and '~~' are non-associative and require parentheses␤at /tmp/QIMZ1PXu50:1␤------> 3(2 ~~ (1 .. 3) ~~7⏏5 (0..4)) ~~ True # chaining...␤ expecting any of:␤ inf…»
Woodi m: (2 ~~ ((1 .. 3) ~~ (0..4))) ~~ True # chaining... 11:24
camelia rakudo-moar e3314c: OUTPUT«Potential difficulties:␤ Smartmatch against True always matches; if you mean to test the topic for truthiness, use :so or *.so or ?* instead␤ at /tmp/bLgqlc4yQx:1␤ ------> 3(2 ~~ ((1 .. 3) ~~ (0..4))) ~~7⏏5 True # chaining...␤»
Woodi m: (2 ~~ ((1 .. 3) ~~ (0..4))) so True # chaining...
camelia rakudo-moar e3314c: OUTPUT«5===SORRY!5=== Error while compiling /tmp/4hnVkhEDRu␤Two terms in a row␤at /tmp/4hnVkhEDRu:1␤------> 3(2 ~~ ((1 .. 3) ~~ (0..4)))7⏏5 so True # chaining...␤ expecting any of:␤ infix␤ infix stopper␤ statement …»
Woodi m: (2 ~~ ((1 .. 3) ~~ (0..4))) :so True # chaining...
camelia rakudo-moar e3314c: OUTPUT«5===SORRY!5=== Error while compiling /tmp/CDXph6L_gs␤You can't adverb that␤at /tmp/CDXph6L_gs:1␤------> 3(2 ~~ ((1 .. 3) ~~ (0..4))) :so7⏏5 True # chaining...␤ expecting any of:␤ pair value␤»
RabidGravy right off out for a bit. play nicely. 11:25
leont It seems Proc.status is returning crap :-/ 11:26
Probably tries to perform math on Int:D + Any:U 11:27
Yeah, 􏿽xAB($!exitcode +< 8) +| $!signal􏿽xBB, with $!signal being Any:U 11:30
11:32 skids left
leont I think submethod BUILD should handle $!signal (and possibly $!pid) 11:32
azawawi Atom::Electron windows events are now working :)
11:33 RabidGravy left
azawawi github.com/azawawi/perl6-atom-elec...e_launcher # Finished Atom::Electron example 11:35
Woodi azawawi: ! some tutorial or blog post pleas ! :)
azawawi Woodi: on it :) 11:36
Woodi should read-first
jdv79 .tell colomon smoke is stalled
yoleaux jdv79: I'll pass your message to colomon.
azawawi Woodi: basically write Perl 6 desktop webapps and control chromium (aka google chrome) windows :) 11:37
Woodi azawawi: bookmarks too ?
azawawi Woodi: you run your own chromium instance. you dont control google chrome 11:38
Woodi: so you can run your own web app and create a window and load that web app and handle its events directly 11:39
what's the tool we use to generate module documentation in Perl 6? 11:41
moritz what tool? :- 11:42
)
there's Pod::To::HTML, but it just generates one HTML file, and you have to manually take care of running it for each module 11:43
11:43 baest_ is now known as baest
masak <leont> It seems Proc.status is returning crap :-/ 11:44
leont: please submit a rakudobug.
azawawi moritz: API to wrap github.com/atom/electron/tree/master/docs/api ... it is a fair daunting task :) 11:45
11:46 rurban left
Woodi so it is GUIs future ? :) 11:47
tadzik heh, I hope not, but I'm afraid yep
masak we are all afraid 11:48
Woodi anyone have examples or some pointers to that technology that "in 100 year we will not use code" ?
masak Woodi: that doesn't square too well with Perl 6 being a 100-year language ;) 11:49
awwaiid naw, just need to add the NLP grammar 11:50
Woodi masak: too :) but I read this few times here as sure thing :)
awwaiid or maybe the robots will still program in perl6
moritz Woodi: I've herad the interesting notion that, eventually, we'll stop writing code, and start growing it instead
*heard
awwaiid so... if "we" == humans... there are LOTS of reasons why humans won't be programming in perl6 :)
moritz like, guiding program auto-assemble from many small parts 11:51
Woodi for sure programming could be less university-grade occupation... artisanal web pages should be by ppls without engeenering degree...
azawawi or machines writing code and then figure out it is easier to outsource it to humans :)
Woodi as soon some code is generated noone is interested in quality or performance... 11:52
I mean things like XML configs not compiler output :) 11:53
11:54 cognominal left, cognominal joined 11:55 itz joined, rurban1 left, freeze joined 11:57 itz_ left 11:58 Begi left 11:59 itz left
masak moritz: I like that. I think I would prefer a future where I discussed the design of a system with an expert system, and then the expert system coordinated the implementation. 12:00
12:06 itz joined
Woodi anyone know how to exit from emacs ? 12:07
12:07 gfldex joined
nine Woodi: kill -9? 12:07
Woodi nine: this is not from inside it :) 12:08
Woodi asked before running emacs... but stackoverflow knows everything :) 12:09
nwc10 Woodi: control-X control-C 12:10
and if someone has unmapped *that*
try
ESC x save-buffers-kill-emacs
but why would anyone want to leave emacs?
or just hit control G a lot. 12:11
Woodi nwc10: exiting is first reaction after entering vi, so I wanted to be prepared :) 12:13
of course it *do not* tell anything about vi quality :) 12:14
nine Well on pressing Ctrl-C in vim, it at least tells you exactly how to exit.
Woodi someone have some advice about C developmnet on Windows ? eg. compiling Moar. looks gcc hallow.c is not all... 12:16
azawawi Woodi: C/C++ development on windows = Visual Studio Express Community edition 12:21
FROGGS but not the 2015 version 12:22
Woodi azawawi: yes, but Express probably do not allow eg. mysql libs linking or I just have problems...
azawawi Woodi: you link the .lib and you are set 12:23
Woodi: but then again i used professional in the past 12:24
12:25 llfourn joined
Woodi in *nix you say gcc file.c and you have a.out... in windows you need ide tons of swithes and libraries for simple "Hallo World"... 12:25
12:26 cognominal left
FROGGS Woodi: no, you don't 12:26
you can also use gcc or cl directly 12:27
12:27 atroxaper left
Woodi I just tried once :) 12:27
12:28 atroxaper joined
masak I would just like to say about 0.Pos and all that -- I'm not staunchly against it. many people like the change, and I can see why. the design feels to me like it replaces risk by magic, which I'm not sure is the right thing to do. but it may be that it's still a net win. 12:28
again one might argue that good tutorials and idioms trump the actual language choice we make.
12:29 FROGGS left 12:32 atroxaper left 12:33 amurf joined
jdv79 why does calling $*SCHEDULER.queue.elems always return 1? 12:35
Woodi num of treads ?
masak jdv79: the answer to that lies in learning what $*SCHEDULER.queue is. what type is it?
jdv79 its seems to be a "ConcBlockingQueue" repr 12:36
trying to print it seems to always yield "Queue.new" 12:37
:(
12:37 amurf left
masak jdv79: so my guess is that it simply doesn't have its own .elems, and so the method call falls back to Any.elems 12:40
dalek kudo-star-daily: 2232970 | coke++ | log/ (8 files):
today (automated commit)
masak m: say Any.new.elems
camelia rakudo-moar e3314c: OUTPUT«1␤»
jdv79 ah 12:42
masak jdv79: I agree with you that a Queue should be expected to have its own .elems, though
jdv79 i see elems in the .c file
masak jdv79: so essentially, you've found a bug.
jdv79 i tried nqp::elems($queue) and nope
any ideas?
masak jdv79: please submit to [email@hidden.address]
jdv79 Well Queue is an empty class in ThreadPoolScheduler so idk... 12:43
but i'll but it
*bug
timotimo if it's empty, it probably gets all of its functionality from the REPR it uses
jdv79 i just don't know how to work with opaue vm-ish things yet
[Coke] (you have to make some value exceptional) ... if only we had... exceptions. 12:44
masak [Coke]: :) 12:46
[Coke]: not really applicable, since the use case is "if it's in the string (then give me the position of its first occurrence)"
jdv79 timotimo: any idea how to call its elems func?
masak [Coke]: it's not "I'm going to assume it's in the string, and please blow up if it isn't" 12:47
azawawi masak: is it possible to run n Bailador servers within the same Perl 6 process using start { baile; } ? 12:49
masak: with different ports that is
masak azawawi: no idea; maybe ask tadzik?
12:51 chenryn left, yqt joined, chenryn joined 12:52 llfourn left
azawawi github.com/azawawi/perl6-atom-elec...example.p6 # it works sometimes lol :) 12:53
Woodi Perl6 should have something like emacs --daemon, then using emacsclient just connect to server :) would solve startup time problems...
jnthn [6~ 12:54
Woodi azawawi: I saw WebContent.undo in Electron API, what it can do ? 12:55
azawawi Woodi: link it please 12:56
moritz Woodi: we have that for the JVM
timotimo jdv79: you can try "use nqp; nqp::elems(nqp::decont($thequeue))" or something 12:57
Woodi github.com/atom/electron/blob/mast...ntentsundo
[Coke] I have no problem with other languages where the sentinel value used isn't possible, like -1.
[Coke] re-reads what he just wrote. 12:58
jdv79 timotimo: segfault
Woodi moritz: looking
jdv79 i'll ignore it for now
timotimo oh my
masak m: my $i = "foo".index("y"); say $i.perl 12:59
camelia rakudo-moar e3314c: OUTPUT«Pos␤»
masak m: my $i = "foo".index("y"); say $i
camelia rakudo-moar e3314c: OUTPUT«(Pos)␤»
masak :/ 13:00
[Coke] , having had at least some coffee, is still ok with that.
masak I didn't realize Pos was (re?)applied to nom. I wanted to check what kind of undefined value we used to return from .index
moritz well, its .perl method needs some improvement 13:01
masak guess I can check the source
ShimmerFairy masak: It was (Int)
moritz aye
masak ShimmerFairy: thank you.
moritz star: say 'foo'.index('why?')
camelia star-m 2015.03: OUTPUT«(Int)␤»
azawawi Woodi: i think it is for controlling text fields (editing commands), undo/redo/replace/etc...
masak moritz++ 13:02
well, if people are OK with this change, then I'll be silent. my unease could be just a measurement glitch or something. 13:03
azawawi jnthn: any idea why gist.github.com/azawawi/901a7c4e086615ea6721 is not capturing prompt output from either thread?
[Coke] If we're going with Pos, I assume there's a ton of places where it would make sense to use. It's basically the old StrPos from the syn with a new name, yes? 13:04
masak yes.
jnthn You're really not meant to use what .queue returns directly. It's only for passing off to various other VM-level APIs, and decidedly not intended for use in userspace.
masak though I always saw the advantage of StrPos in being that it abstracted a numerical position into a more "cursor-like" thing that could straddle/ignore the bytes/chars/graphs levels. 13:05
that doesn't seem to be the case with this new Pos. it seems to be a subtype of Int.
[Coke] I am torn. It handles the use case it was introduced for very nicely, but it's yet another type. I am probably just an old fart at this point, but having to save the result of an expression if I want to inspect it multiple ways seems like "the way to do it" not a big problem.
jdv79 i was just poking at the async stuff 13:06
jnthn azawawi: Using sync IO handles from threads other than the one that created them is currently not supported
jdv79 i'll find a better way later i guess to do what i was attempting
jnthn (It's on the to-fix list.)
jdv79: What were you trying to do?
tadzik azawawi: it may not be
azawawi: but probably you can read what start() does and just replicate it with different values for port and stuff :)
jdv79 just fan out and collect - experimenting in general
13:07 chenryn left
jdv79 the particular issue was detecting when the queue is exhausted 13:07
azawawi jnthn: so `use Bailador; start { baile; }; start { baile; }` # wont be supported?
jdv79 but its the wrong way about it anyway for anything real
jnthn jdv79: my @results = await do for @work { start { ...computation... } } 13:08
13:08 cognominal joined
jdv79 that's higher level than i wante 13:08
jnthn Why?
jdv79 just wanted to see how much extra taht is from a minimal impl 13:09
just for fun
jnthn ah
13:09 ggoebel joined
jnthn You can't do it by looking at the queue 13:09
Not reliably
That'll be full of races
13:09 llfourn joined
jnthn You need a semaphore or cond var somewhere so you can wait efficiently 13:10
azawawi: That's almost certainly the wrong way to do it
jdv79 yeah, i know. i was trying to do the absolute min and then build up
less min tehn 13:11
thanks
moritz well, you can start by startings Tread objects and joining them in the end
jdv79 yeah but i didn't want to code a pool mgr
i'll fool later - have to run though. thanks! 13:12
masak .oO( run, you fools! ) 13:13
moritz fool, you, run!
masak [Coke]: well, the expectation we're breaking is essentially that ?$pos == ?+$pos 13:14
Woodi moritz: eval-client.pl reads file and sents to perl-j listening on port and doing eval ?
azawawi jnthn: thanks... run Bailador in an async process ftw github.com/azawawi/perl6-atom-elec...example.p6 13:16
moritz Woodi: iirc it named socket/pipe 13:17
azawawi moritz: with that bailador example, we could run perl6 REPL in a very nice GUI and control the browser window shortcuts (no more Ctrl-W closing the window ...etc) 13:22
moritz: we can even run web irc in a separate browser tab. 13:23
13:24 laouji_ left, Begi joined 13:25 ribasushi left
moritz azawawi: nice 13:27
13:28 ribasushi joined
azawawi is panda install HTTP::Server::Async failing on your dev box? where is the testers link? 13:31
13:38 laouji joined
Woodi can be electron binary packaged with app ? 13:39
13:40 JimmyZ_ joined
azawawi Woodi: yes we can :) 13:40
Woodi: just make it available in your $PATH 13:41
13:42 raiph joined
azawawi Woodi: did you download the appropriate runtime for your plaform from github.com/atom/electron/releases/tag/v0.30.1? 13:43
Woodi azawawi: not downloaded yet. but thinking about making Windows apps becouse on linux I usually don't have X desctop 13:46
*desktop
curently there is a fashion for making game launchers that rsync conntent for you :) 13:47
13:49 jkva left
timotimo do you mean content as in game updates or as in save games and such? 13:55
13:56 cognominal left
Woodi timotimo: content. eg. DayZ mod for Arma2 13:57
azawawi Woodi: cool
timotimo mhm
Woodi azawawi: can I replace firefox with electron and have my own session manager ? :)
timotimo you would have your own session manager, but not much else 13:58
Woodi timotimo: but having GUI for app builded on SDL would be nice
13:58 raiph left
timotimo very difficult to embed, i assume 13:58
Woodi what to embed ? SDL window with menu, status bar and grid... 13:59
dalek c: 53189e2 | (Lloyd Fournier)++ | lib/Language/modules.pod:
1. Addressed some of the tense and passive voice issues

3. Fixed some mistakes
c: 33b9276 | (Lloyd Fournier)++ | lib/Language/modules.pod:
small touch ups and sentence rewrites on modules
c: ccf8229 | paultcochrane++ | lib/Language/modules.pod:
Merge pull request #108 from LLFourn/master

Attempted to address passive vs active voice and added examples for need,use,import and require
timotimo i thought you meant SDL + electron 14:00
Woodi no, electron is 30Mb, SDL 3 ?
timotimo SDL doesn't load different image and video formats and doesn't do text layouting and 3d stuff ... 14:01
14:01 raiph joined
nine [Coke]: unfortunately if $string.index('foo') { ... } is such an easy trap to fall into. 14:01
14:02 khw joined 14:03 census left
azawawi Woodi: it is Atom::Electron. i have not seen a way to embed firefox while doing research 14:06
timotimo well, you can use gecko 14:07
14:07 laouji left
azawawi true 14:07
but then again electron is more actively developed because of the Atom editor atm :) 14:08
azawawi going home in 10 min :)
timotimo right 14:09
14:11 skids joined 14:14 ggoebel left
azawawi Woodi, timotimo: Enjoy i.imgur.com/Nbf57YO.png # Electron + Bailador :) 14:17
Woodi azawawi: maybe I do not understand how electron is working... it framework for making desktop apps. it can load url. so I can have many windows like in firefox -> session manager, bookmarks :) maybe not tabs
masak azawawi: wow! 14:18
azawawi++
azawawi Woodi: you can make new BrowserWindow(s) :)
Woodi: and there is Session management but not wrapped yet
Woodi I think about something on Perl side...
azawawi Woodi: github.com/atom/electron/blob/mast...ioncookies
Woodi azawawi: that "File View Help" is standard thing ? 14:19
azawawi off to get a Frappuccino to celebrate :)
Woodi: can be disabled, it is for development 14:20
Woodi azawawi: I want to have it :) windows on the right could be hidden...
azawawi masak: the fun thing is that we can also use clipboard api, frameless transparent windows, tray api and shell api :) 14:22
14:22 amurf joined
azawawi Woodi: BrowserWindow->new(:show(False)); # should work :) 14:22
masak azawawi: nice! 14:24
14:24 atroxaper joined
azawawi masak: so basically it is a wxwidgets app but with a HTML/CSS layout engine :) 14:25
masak: cross platform, hopefully lol
azawawi home & 14:26
14:27 amurf left 14:29 atroxaper left 14:30 cognominal joined, azawawi left
timotimo cute 14:34
the image
14:34 jkva joined
dalek ast: f30dec1 | skids++ | S06-currying/named.t:
Remove unused sub from test file.
 14:37
Begi azawawi : nice ! I'll try that ! 14:41
14:42 g4 left
skids So... question for those who think removing the implicit *%_ is a good idea: "class A { has $.a }". What should be the signature of the implicitly defined A.new and A's submethod BUILD? Will it still be Mu's, or an autogenerated one? Curious just how radical you rebels are. 14:42
14:43 telex left 14:44 telex joined
moritz skids: I still think Mu.new can accept any named param 14:44
skids Right, but would A have Mu's new, or an autogenerated one based on attributes? 14:47
moritz it can have Mu's new 14:48
skids OK, so anyone who wants to NOT take unnown named attributes to new will have to define their own new anyway.
moritz yes 14:49
skids (just as they do now)
moritz it's all the other methods I'm concerned about right now :-)
skids does not share that concern frankly. But that's good to know. 14:50
14:56 brrt joined
vendethiel m: [1] @ [2] 14:57
camelia rakudo-moar e3314c: OUTPUT«5===SORRY!5=== Error while compiling /tmp/EE4eHIqWNs␤Two terms in a row␤at /tmp/EE4eHIqWNs:1␤------> 3[1]7⏏5 @ [2]␤ expecting any of:␤ infix␤ infix stopper␤ statement end␤ statement modifier␤ …»
vendethiel ^ @ is now an operator python has, and we don't :P
brrt \o
vendethiel o/ brrt 14:58
brrt \o vendethiel
15:00 yqt left 15:01 RabidGravy joined
RabidGravy boom 15:02
15:05 jkva left 15:06 jkva joined 15:08 diana_olhovik_ left 15:14 eMBee left 15:20 brrt left 15:29 dustinm` left 15:31 zakharyas left 15:32 mtj_- left, dustinm` joined 15:35 petercommand left, agentzh left 15:36 mtj_- joined, petercommand joined, agentzh joined 15:37 Begi left
vendethiel looking at PEP-492, it looks like me that we... "folded" the generator feature that languages added/are adding (php5.6, python, es6) into gather thanks to infinite lists, and I like it :D. 15:38
one thing "interesting" is that "await" really applies to "one expression". Like: "await fut + 1" will be parsed as "(await fut) + 1". Basically, you're really "calling a function with a modifier", it's not like our "await {}" 15:39
15:40 [Tux] left 15:42 captain-adequate joined
nine Assuming I can fix all emerging issues, would it actually be a good idea to have Perl5Object inherit from Mu instead of Any? One one hand it would fix issues with Any's methods shadowing the Perl 5 object's methods. On the other hand those objects would feel less like Perl 6 objects. 15:43
15:43 Begi joined 15:46 diana_olhovik joined
masak vendethiel: do you mean "one term"? 15:47
vendethiel masak: where? 15:48
oh, yes. I did. 15:49
15:54 [Tux] joined
timotimo nine: i think it's a bit problematic because we basically limit the big majority of parameters to Any 15:54
by default, that is
for pointy blocks, parameters are Mu, but for subs and methods, they are Any by default
nine timotimo: sounds like an excellent argument for keeping Any as base class. Especially since I hit that even in Inline::Perl5's code itself. 15:55
timotimo: I wonder how this works with custom meta objects? 15:56
15:57 khw left, yqt joined
timotimo haven't tried 15:57
dalek c: 7485ab2 | (Lloyd Fournier)++ | lib/Language/functions.pod:
added example of type coercions for non builtin types
 15:59
c: c874e11 | LLFourn++ | lib/Language/functions.pod:
Merge pull request #109 from LLFourn/master

added example of type coercions for non builtin types
16:02 rurban joined, jkva left 16:03 rurban1 joined 16:06 smls joined
smls gist.github.com/jnthn/a56fd4a22e7c43080078 looks pretty cool, jnthn++ 16:08
I especially like supply/emit as the observable equivalent to gather/take
masak me too. 16:09
and whenever as an async loop construct.
smls I admit I don't quite understand why whenever $supply { } is needed when for $supply.list { } already works 16:11
or after pmichaud's GLR changes, simply for $supply { } 16:12
masak I can't imagine that those have the same semantics.
(the first two)
the for loop looks blocking to me
smls and whenever doesn't? 16:13
oh right, "asynchronous looping construct"
16:13 xinming_ joined
dalek line-Perl5/mu: f1d4668 | (Stefan Seifert)++ | / (2 files):
Failed attempt to use Mu as base for Perl5Object

Using Mu instead of Any as base class for Perl5Object and Perl5Package classes might have helped cases where Any's methods shadow the Perl 5 object's methods.
But as timotimo++ pointed out, the big majority of parameters is restricted to Any (so Perl5Objects would not be accepted). So this would probably cause more harm than good.
Push it into a branch to document this fact and what changes would have been necessary.		 16:14
smls Though inside a supply { }, using a blocking loop should be fine too shouldn't it? If it works like gather { }
16:17 xinming left
masak yeah, but two `whenever`s can work in parallel without any extra instructions, which is nice. 16:28
smls yeah, I think I see the benefits now
lizmat is back 16:32
and has read all of the backlog
(of today)
I like the idea of Str.contains(Str) returning a boolean 16:33
I wonder if .contains could be more general
and a replacement for the $int ~~ ^10 meme
(^10).contains($int) 16:34
smls What's wrong with .index().defined ?
lizmat if "foo".index("f").defined -> $pos # doesn't work 16:35
smls Neither will .contains -> $pos if it returns a Bool will it?
16:35 espadrine_ left
lizmat no, but if you use .contains you declare not being interested in the position 16:35
there are a few places in the core where .index is used in that way 16:36
.contains will *always* return a bool
smls index().defined too :) 16:37
lizmat I wonder whether we should maybe change .defined in such a way that it will pass on the value but True ?
m: if "foo".index("f") -> $pos { say $pos } # says $pos even if it is 0 and therefore false in the Int interpretation 16:38
camelia rakudo-moar e3314c: OUTPUT«0␤»
flussence re: *%_ -- the kernel devs do a similar thing, with `flags` params in all new syscalls for future-expansion stuff. That's one reason to have it, but it'd be nice to fix the awkward things it causes too.
lizmat ISTR that jnthn at least has an idea for complaining about unused named parameters
smls Maybe .defined needs a shorter name (akin to .so) to make encourage its use?
lizmat that would at least solve some of the issues
flussence that'd help a lot, yeah :)
lizmat smls: it's not about the name, it is about the functionality 16:39
if you find something *and* you want to know its position, you needed (before Pos) to go through hoops
*or* forget the case it was found at the first position
m: my $pos = "foo".index("f"); if $pos.defined { say $pos } # old situation 16:40
camelia rakudo-moar e3314c: OUTPUT«0␤»
smls right, but introducing .contains would not help at all with that
lizmat no, but it *would* clarify the intended use better... 16:41
smls maybe
16:41 cognominal left
lizmat and it could be made to be slightly faster than .index 16:42
basically: nqp::index( nqp::unbox_s(self.Str), nqp::unbox_s($needle.Str), nqp::unbox_i($pos.Int) ) >= 0 16:43
smls For some reason, use of .index() for merely checking existence does not bother me much, but using .first/.first-index on a list for merely checking existence does bother whenever I see it 16:44
lizmat I would say: another case for .contains :-)
smls or, make $needle == any(@haystack) really short-circuiting and optimized :P 16:45
leont would like a short-cutting contains on lists
smls What comparison would List.contains use? eqv? 16:46
leont Yeah, that is the tricky bit
If junctions in boolean context can be optimized to short-cut, that may be helpful too, but I'm not sure they could be
16:47 kst left 16:48 atroxaper joined
smls I was always under the impression that the reason why Perl 6 does not have a .contains (or "in" operator like Python), was that we are supposed to combine junctions with a comparison operator for that 16:50
But I vaguley remember that junctions also are supposed to be allowed to parallelize, which would be mutually exclusive with short-circuiting 16:51
16:52 atroxaper left
smls lizmat: Also, overloading .contains for both strings and lists, would be unusual in the context of Perl 6 16:53
16:53 JimmyZ_ left
smls e.g. $string.grep does not grep characters, it treats $string as a single-element list 16:54
lizmat good point, smls++
dalek ast: beded9e | moritz++ | S32-io/slurp.t:
Remove a test that does not work anymore
 16:57
lizmat starts backlogging the past 2 weeks to be able to write the P6W 16:58
moritz lizmat: did you do a roeast cleanup? 'cause a "make spectest" with RAKUDO_DEPRECATIONS_FATAL=1 seems pretty clean
lizmat moritz: I haven't done much in the past 2 weeks being on the road and such
I *did* remove some deprecations before the release and their associated tests 16:59
maybe that could account for the difference?
moritz lizmat: uhm, no, it seems that RAKUDO_DEPRECATIONS_FATAL=1 doesn't do much
lizmat oh?
moritz export RAKUDO_DEPRECATIONS_FATAL=1; make t/spec/S17-su│··
pply/rotor.t
passes
even though there's a deprecation warning 17:00
lizmat hmmm..... 17:01
moritz ah no
it's a dies-ok { ... }, 'no param version illegal'
and that triggers the deprecation warning 17:02
and RAKUDO_DEPRECATIONS_FATAL=1 makes it fatal, thus passing the otherwise TODOed test
lizmat maybe we should have a RAKUDO_DEPRECATIONS_EXIT ?
moritz don't bother 17:03
is a sufficiently obscure case that we shouldn't optimize for it
lizmat okidoki :-)
[Coke] I thought we absolutely had an in operator for sets. 17:08
leont There's unicode for that :-p
TimToady .tell mdinger the way to think about it is that all assertions are about the current cursor, so <?after foo> means "is the cursor after foo?", and <?before baz> means "is the cursor before baz?"
yoleaux TimToady: I'll pass your message to mdinger.
leont Yeah, that was initially confusing for me too 17:09
smls [Coke]: Right, but sets are special-purpose and have a settled on === comparison for everything 17:10
moritz not more or less confusing that look-ahead and and look-behind
if you look ahead to something, it's before you, no?
smls [Coke]: The ASCII version is called $needle (elem) $haystack I think 17:11
leont "Is foo after this" versus "is this after foo"
17:11 dakkar left
[Coke] moritz: now you're confusing meanings of before. :P 17:12
moritz [Coke]: am I?
[Coke] are you referring to looking ahead in time or place? 17:15
lizmat
.oO( when it's Daylight Saving Time, is it earlier or later than before? :-)
 17:16
[Coke] and did it happen before now, or does it lay before your current position? I agree with OP that this is potentially confusing. I think TimToady answered it fine.
lizmat: I don't know, but I'm sure all the IT people in the room are grumpier when it's DST. :P 17:17
Juerd A thought crossed my mind, and please don't take this seriously: shouldn't $_ be renamed to $- now? ;) 17:27
[Coke] m: say $-; 17:29
camelia rakudo-moar e3314c: OUTPUT«5===SORRY!5=== Error while compiling /tmp/ZgNU7T3_VG␤Missing required term after infix␤at /tmp/ZgNU7T3_VG:1␤------> 3say $-7⏏5;␤ expecting any of:␤ prefix␤ term␤»
lizmat Juerd: no, because you cannot have a trailing -
Juerd I know :)
lizmat or an identifier starting with - 17:31
b2gills $- The number of lines left on the page of the currently selected output channel. ( from perldoc -v '$- 17:34
Juerd b2gills: That's Perl 5 17:35
raiph TimToady: did you suggest a Maybe thing sitting alongside junctions that does stuff like collapsing to a Bool but value?
TimToady don't recall that 17:36
I did suggest an Each that autothreads lists though
nine smls: .index().defined is stating how it should be done, while .contains() states what should be done. 17:38
17:38 yqt left 17:41 llfourn left
vendethiel metacpan.org/pod/Dios Damian Conway is just insane :o). In a good way,obviously 17:44
leont Obviously, he is! 17:49
vendethiel is now reading the source code 17:51
leont says goodbye to vendethiel's sanity 17:53
vendethiel leont: I could do with a definition of "sanity" before leaving :D.
17:57 rurban1 left 17:58 zakharyas joined
leont Being able to distinguish up and down :-p 17:58
17:58 abraxxa left 18:00 amurf joined 18:04 amurf left 18:05 mr-fooba_ left, ChoHag joined
ChoHag How's perl 6's support of unicode wrt. modules and their filenames? 18:06
In particular considering that perl 5 doesn't allow non-ascii modules because of, largely, Windows' filesystem. 18:07
lizmat ChoHag: they should be supported, but aren't yet afaik
S11:420 18:08
ChoHag "internal names are Unicode-aware, even if the filesystem isn't" 18:11
Уау! 18:12
dalek kudo/nom: 174049f | TimToady++ | src/Perl6/Grammar.nqp:
Revert "move ~~ from chaining to structural"

This reverts commit 81ad2c0db9910fe7093c1db1d0e25e4710f3e67e.
Consensus that the damage inflicted by changing the precedence is worse than the unnaturalness of considering ~~ a chaining op.		 18:13
ast: 687df77 | TimToady++ | S (9 files):
Revert "benign prep for ~~ going chained -> structural"

This reverts commit ab605d96f42e531f9d544a167eef590a381a80b0.
Don't need parens around ranges in ~~ again		 18:14
18:15 Begi left
leont ChoHag: in perl5 that could be solved, but no one @people.grep(Tuits&Win32-knowledge) eqv () 18:26
18:27 kst joined 18:28 cognominal joined
ChoHag That's what I figured from the reading I was able to do on the subject. 18:32
Except in English.
jnthn evening, #perl6 18:36
Few bits from backlog (mostly gut feeling, I'm too tired for anything else :)) 18:37
18:39 zakharyas left
jnthn On the eternal *%_ complaining (I guess 'cus named params are overused and Liskov is underunderstood :P): I wonder if we should make it be invocations with callsame/nextsame/callwith/nextwith that tolerate dropping extra named args, since that's where you actually want it. And normal invocations don't. 18:40
Basically, a failover on the "do we have any extra named args" check that realizes the invocation was done through callsame and friends.
And fails to complain in that case.
18:41 muraiki joined
lizmat jnthn o/ 18:41
jnthn Probably we have to make $foo.Bar::baz() also behave the same way too
o/ lizmat
18:42 FROGGS joined
jnthn On 'for $supply.list { }': it's an anti-pattern to do this most of the time. It's occasionally useful. I would say it's useful in demos, but last time I did the .Net equivalent in a live demo my code deadlocked on itself. :P 18:43
The reason being that it is an async/sync transition. Such boundaries are inherently dangerous.
On hyper op parallelization semantics: I think we probably should have the operator form of hyper try to DWIM, and let named args be used to the method equivalents for when you really need the escape hatch. 18:45
18:46 bin_005 joined, pdcawley left
jnthn Experience tells that too many folks think they can beat the runtime's decisoins without actually seeing if that's true...and even when they're right on the system they test the code on, that can all be undone by deploying on different hardware or on a loaded system (vs. their unloaded development machine). 18:46
So I'm inclined to dehuffmanize the "I want to over-specify things" path. 18:47
lizmat jnthn: so I understand, you're talking about @list>>.foo ? 18:48
jnthn lizmat: Yeah
Well, I'm talking more generally, but that's an example. 18:49
18:49 Hor|zon left
lizmat my idea was that methods would specify how they would like to be batched in hypers (if at all) 18:49
jnthn The method is only one axis of the decision making, though. 18:50
18:50 inokenty left
jnthn And you might legitimately want to make different decisoins in different uses of the method. 18:51
lizmat what is the other axis then? Number of threads / cores available ?
jnthn Size of data set, and whether you care more about throughput or latency.
raiph isn't batching (1 or 100) what lazy vs mostly lazy was about?
lizmat but how can the system know from the outside how much CPU / wallclock a method will take ? 18:52
jnthn raiph: No; hyper is a form of eager that can thread.
raiph thanks
jnthn (We have used the word batching in the context of laziness though, so I understand the confusion. :)) 18:53
lizmat: You can't know, but you can measure and use the past to predict the future.
lizmat: And even then you have to be somewhat dynamic about it. 18:54
lizmat ok, that would imply starting with one and adapting along the way ?
jnthn Crappy example 'cus I'm tired, but (^1000)>>.&a-stupid-recursive-fib-impl has vastly different costs for different calls over time.
lizmat: Yes. 18:55
lizmat: Thankfully there's lots of prior art we can look at on this. :)
lizmat: I think hill-climbing algos are common in this space, though. 18:56
I think that's what we should do as the default, anyways. But I'm all for a control know too. I just thing it's most useful on a use-case basis. 18:57
For example, if I'm throwing results onto a UI then I'd like to work in small batches to get the latency down. If I'm just crunching away non-interactively, I'd prefer bigger batches to keep the thread communication down. 18:58
uh, an control *knob*
and an a...
18:58 estrabd_afk is now known as estrabd
jnthn en.wikipedia.org/wiki/Hill_climbing is the thing I'm talking about, fwiw 19:00
19:00 Alina-malina joined 19:07 spider-mario joined
lizmat moritz: shall I mention the Rakudo * Release Candidate at hack.p6c.org/~moritz/r​akudo-star-2...RC0.tar.gz ? 19:14
moritz lizmat: uhm, I could also just release it :-)
lizmat Fine by me too :-) 19:15
.oO( long live P6W driven development :-)
19:15 beastd joined
timotimo long live the lizmat 19:17
moritz hack.p6c.org/~moritz/rakudo-star-2015.07.tar.gz tests welcome 19:20
lizmat moritz: you want me to use that URL ? 19:23
moritz lizmat: no
lizmat okidoki
moritz release tagged, tarball uploaded 19:25
lizmat: you can link to rakudo.org/2015/07/27/announce-raku...e-2015-07/ :-) 19:27
lizmat moritz++ (done)
19:28 jkva joined
timotimo is [release 2015.07.2] supposed to have brackets around it? 19:29
Notable changes in modules in modules: 19:31
moritz timotimo: linkified
timotimo there's one in modules in modules too modules in
dalek ar: 60ecaef | moritz++ | docs/announce/2015.07.md:
Grammar fix, timotimo++
 19:32
19:34 mr-foobar joined 19:36 [Sno] left, zakharyas joined, espadrine_ joined 19:37 ChoHag left, darutoko left, yqt joined, llfourn joined
timotimo moritz++ # rakudo star release 19:37
jnthn going for some rest...back tomorrow & 19:39
dalek ar: f5f2ca3 | moritz++ | tools/star/release-guide.pod:
mention 2015.07 release in release-guide.pod
 19:41
moritz jnthn: have a RESTful night :-) 19:42
19:42 llfourn left, zakharyas left
moritz R* release process finally winding down 19:43
... except for the missing .msi 19:44
19:45 Hor|zon joined 19:48 dha joined 19:49 brrt joined, rurban1 joined 19:50 espadrine_ left, Hor|zon left
dha I'm sure this is just me being confused, but where the heck is panda in my rakudo distribution? 19:50
lizmat it's not 19:51
panda lives separately, or in rakudo star
or installed with rakudobrew
dha I keep getting confused about rakudo vs. rakudo star.
lizmat generally, we don't announce rakudo compiler outside the echo chamber, just rakudo star 19:52
dha *nod*
19:55 raiph left
RabidGravy but if you already installed rakudo & panda via rakudobrew then reinstall rakudo with it you'll get the new panda and all the modules reinstalled 19:55
19:56 smls_ joined, smls left, cognominal left
moritz panda intall Task::Star 19:56
20:00 [Tux] left
dha I've not been using rakudobrew. In fact, I've not installed rakudo. Just been using it in its own directory. 20:03
rakudobrewing now, though.
Which, using the airplane's wifi, should be tons of fun.
lizmat
.oO( building in the very fast lane )
 20:04
nine has never used rakudobrew
I just have a git checkout of rakudo and panda and do make install in rakudo 20:05
lizmat looking at $a.?foo and $a.*foo and $a.+foo
I've seen the first abused a lot when in fact we only want to call the method if the object is instantiated 20:06
so sometimes I think we need something like:
20:06 Hor|zon joined
lizmat $a//.foo 20:06
this would call "foo" on $a if $a was defined
20:06 jkva left
lizmat m: sub infix:<//.>($obj,$method) { $obj.DEFINITE ?? $obj.$method !! Nil }; my $a; sub b { $^a + 2 }; say $a//.&b # something like this 20:07
camelia rakudo-moar 174049: OUTPUT«Nil␤»
lizmat m: sub infix:<//.>($obj,$method) { $obj.DEFINITE ?? $obj.$method !! Nil }; my $a = 40; sub b { $^a + 2 }; say $a//.&b # same, now defined
camelia rakudo-moar 174049: OUTPUT«42␤»
lizmat note the messy &b is needed to make sure we have a callable there
was this functionality discussed in the past in some way ? 20:08
20:10 diana_olhovik left
ShimmerFairy lizmat: Any reason you couldn't use Callable $method or &method in the parameter list? (But for a sub call, you need & to use it as a method anyway) 20:12
lizmat m: sub infix:<//.>($obj,$method) { $obj.DEFINITE ?? $obj.$method !! Nil }; my $a = 40; say ($a//.Str).perl # this actually works ??? 20:14
camelia rakudo-moar 174049: OUTPUT«"40"␤»
lizmat probably not very optimized, as $method is a Str there
m: sub infix:<//.>($obj,$method) { dd $obj, $method; $obj.DEFINITE ?? $obj.$method !! Nil }; my $a = 40; say ($a//.Str).perl 20:15
camelia rakudo-moar 174049: OUTPUT«$obj = 40␤$method = Str␤"40"␤»
lizmat so we would need some syntactic sugar there
ShimmerFairy lizmat: Now I see why you thought you needed the & , and now I'm surprised it works too. 20:16
lizmat m: sub infix:<//.>($obj,$method) { $obj.DEFINITE ?? $obj.$method !! Nil }; my $a = 40; say ($a//.foo).perl # interesting
camelia rakudo-moar 174049: OUTPUT«5===SORRY!5=== Error while compiling /tmp/HepXEsCotM␤Undeclared routine:␤ foo used at line 1␤␤»
lizmat so, if it is not known, it thinks it's a sub 20:17
ShimmerFairy m: sub infix:<//.>($obj,$method) { dd $obj, $method; $obj.DEFINITE ?? $obj.$method !! Nil }; my $a = 40; say ($a//.pred).perl
camelia rakudo-moar 174049: OUTPUT«5===SORRY!5=== Error while compiling /tmp/bgsotB_n_R␤Undeclared routine:␤ pred used at line 1␤␤»
lizmat and if it *is* known, it is passed as a string
ShimmerFairy lizmat: It always thinks it's a sub, you were just trying methods with sub versions, it turns out :) 20:18
brrt \o 20:19
ShimmerFairy o/
brrt hmm 20:20
i'm not sure i can formulate my question clearly enough to expect an answer
rurban1 Dios is nice, but with its PPI dependencies too big. t/alias.t compiles to a 18M c files which runs out of memory. Do you like it's syntax? The {{{ }}} interpolation with <{ … }> expansion syntax is esp. ugly. I used `` for expansion. 20:21
brrt i'm trying to prune the combinations of rules to those that will actually appear
nine m: role Foo { $?CLASS.^add_method(list => method (|args) {}); }; my $class := Metamodel::ClassHOW.new_type( name => "Foo" ); $class.^add_role(Foo); $class.^compose; 20:23
camelia rakudo-moar 174049: OUTPUT«Cannot find method 'collisions'␤ in any apply at src/gen/m-Metamodel.nqp:2633␤ in any compose at src/gen/m-Metamodel.nqp:2810␤ in block <unit> at /tmp/lTIz0X99Nd:1␤␤»
nine why?? ^^^
m: role Foo { $?CLASS.^add_method("list", method (|args) {}); }; my $class := Metamodel::ClassHOW.new_type( name => "Foo" ); $class.^add_role(Foo); $class.^compose;
camelia ( no output )
ShimmerFairy nine: probably because the Pair is gobbled up by *%_ , would be my guess 20:24
m: role Foo { $?CLASS.^add_method((list => method (|args) {})); }; my $class := Metamodel::ClassHOW.new_type( name => "Foo" ); $class.^add_role(Foo); $class.^compose;
camelia rakudo-moar 174049: OUTPUT«Cannot find method 'collisions'␤ in any apply at src/gen/m-Metamodel.nqp:2633␤ in any compose at src/gen/m-Metamodel.nqp:2810␤ in block <unit> at /tmp/KiczotfpnX:1␤␤»
nine This "Cannot find method 'collisions'" appears in some constellations and not in others. But I cannot find a system
20:24 atroxaper joined
nine It's not the =>. That's just one way to provoke it. In Inline::Perl5 it fails without a => 20:25
ShimmerFairy nine: and I'm guessing there's a reason you can't just do role Foo { method list (|args) { } } and class Foo { ... } :)
nine ShimmerFairy: of course :) The class name is only known at runtime and the methods I add are Any.^methods 20:26
Oh, I think I know the system! It appears on any error during method creation: 20:29
m: role Foo { for Any.does-not-work -> $name { $?CLASS.^add_method($name, method (|args) {}); }; }; my $class := Metamodel::ClassHOW.new_type( name => "Foo" ); $class.^add_role(Foo); $class.^compose;
camelia rakudo-moar 174049: OUTPUT«Cannot find method 'collisions'␤ in any apply at src/gen/m-Metamodel.nqp:2633␤ in any compose at src/gen/m-Metamodel.nqp:2810␤ in block <unit> at /tmp/9KSnt7YtMK:1␤␤»
20:29 atroxaper left
ShimmerFairy m: role Foo { $?CLASS.^add_method((list => method (|args) {})); }; my $class := Metamodel::ClassHOW.new_type( name => "Foo" ) but Foo; say $class; # mixing in the role with "but" 20:30
camelia rakudo-moar 174049: OUTPUT«'Foo+{Foo}' cannot inherit from 'Foo' because 'Foo' isn't composed yet (maybe it is stubbed)␤ in any add_parent at src/gen/m-Metamodel.nqp:774␤ in any generate_mixin at src/gen/m-Metamodel.nqp:1290␤ in any mixin at src/gen/m-Metamodel.nqp:1250␤ …»
nine m: role Foo { for <foo foo> -> $name { $?CLASS.^add_method($name, method (|args) {}); }; }; my $class := Metamodel::ClassHOW.new_type( name => "Foo" ); $class.^add_role(Foo); $class.^compose; 20:33
camelia rakudo-moar 174049: OUTPUT«Cannot find method 'collisions'␤ in any apply at src/gen/m-Metamodel.nqp:2633␤ in any compose at src/gen/m-Metamodel.nqp:2810␤ in block <unit> at /tmp/cYVtXaM62J:1␤␤»
nine I think this is what happens in Inline::Perl5 ^^^ 20:34
I'm trying to add a method that is already inherited from somewhere. Ironically this works just fine when I crate the class using EVAL "class GLOBAL::$module does Perl5Package[\$p5, \$module] { }"; 20:35
ShimmerFairy nine: maybe you can use a macro that creates a method instead of trying to go through the metamodel? (I'm not familiar with doing such complex language trickery, so I'm afraid I can't be of much help) 20:36
20:37 llfourn joined 20:42 llfourn left 20:43 bin_005 left 20:44 dalek left 20:45 rurban left, dalek joined, ChanServ sets mode: +v dalek
nine Ok, I think the loop adding methods runs multiple times. No idea why though. A next if $?CLASS.^declares_method($name); in the loop seems to help. 20:48
20:54 bin_005 joined
nine "An exception occurred while evaluating a BEGIN" 20:55
Is it me or has rakudo's error reporting become drastically worse in the past few days? I'm now getting such usefull messages...
skids
.oO(If we implemented failures to spec we could just have my $nosink = Any.method.fails.down.to.this.semicolon; # and not deal with syntax sugar.)
 20:56
20:57 brrt left
ShimmerFairy nine: There was something about error reporting over the last few days. Though if this is your first time messing with the MOP, I'd blame delving deeper into the internals as a cause for more cryptic errors :P 20:59
21:00 dha left, RabidGravy left
nine ShimmerFairy: unfortunately it's not only the messing around with the MOP 21:00
m: class Foo is Mu { }; say Foo.new ~~ Pair; 21:01
camelia rakudo-moar 174049: OUTPUT«X::Multi::NoMatch exception produced no message␤ in block <unit> at /tmp/JcPiZmeThD:1␤␤»
nine ^^^ another example of a pretty useless error message
skids Because someone did not put a .message in that.
nine And I have the feeling that "Cannot find method 'collisions'" is also new
There is code that should tell me that the message already exists 21:02
s/message/method/
21:02 bin_005 left
skids Hrm what did you do to get the X::Multi::NoMatch? 21:03
OIC nm
I'll look at that when I get home. That should not happen.
it has a .message method. 21:04
21:04 dha joined
skids (And is likely my fault somehow.) 21:04
21:05 [Sno] joined
nine Ok, the "Cannot find method 'collisions'" was already there on July 18th 21:06
class Foo is Mu { }; say Foo.new ~~ Pair; gave an equally unhelpful "Invocant requires a 'Failure' instance, but a type object was passed. Did you forget a .new?" 21:07
skids I've only seen that one when doing something truly abusive with roles/scoping.
nine skids: so actually your changes really were an improvement. There's just more to be gained :)
skids Well, if there really is a type object being passed in, then no, that would not be an improvement. Anyway decommute and then I'll dig at it a bit. 21:09
21:09 telex left
nine Since I only do Foo.new ~~ Pair, I think the source of the error is a missing ACCEPTS that Any would provide. 21:09
X::Multi::NoMatch is a mild indication to that effect.
smls_ lizmat: "was this functionality discussed in the past" -- I tried to. :P I think TimToady told me to use 'andthen'. (Once it works correctly. But even then, the low precedence will be bothersome.) 21:10
21:10 telex joined
smls_ i.e. ($obj andthen .method) 21:10
My biggest gripe with "andthen" is actually, that its name makes it sounds like a version of "and" (which checks for truthiness), but it instead checks for definedness. 21:12
lizmat m: my Int $a; say $a andthen .succ # how is this supposed to work ? 21:13
camelia rakudo-moar 174049: OUTPUT«(Int)␤Method 'succ' not found for invocant of class 'Any'␤ in block <unit> at /tmp/9tjkbHrpId:1␤␤»
21:13 skids left
lizmat m: my Int $a; say $a andthen $a.succ # huh? 21:14
camelia rakudo-moar 174049: OUTPUT«(Int)␤Invocant requires a 'Int' instance, but a type object was passed. Did you forget a .new?␤ in block <unit> at /tmp/s8WBKf681l:1␤␤»
ShimmerFairy m: my Int $a; say ($a andthen .succ); # low precedence caught you here, looks like
camelia rakudo-moar 174049: OUTPUT«(Int)␤»
smls_ andthen is supposed to make the LHS available as $_ before evaluating the RHS 21:16
diegok Hello, is there any config or anything to make p6 repl more friendly? 21:23
dha define "friendly". :-) 21:24
smls_ diegok: If you install the Linenoise module, the REPL will automatically use it and you'll get proper line editing, tab completion, and history support. 21:25
21:25 Fleurety left
diegok smls_: thanks!, installing it :-) 21:26
21:26 paulmcquad joined, paulmcquad left 21:27 egrep joined, egrep left
smls_ Maybe the REPL should print a corresponding hint when started... 21:28
Python's REPL prints 3 lines of intro messages when starting the REPL; Perl 6 prints none so far... 21:29
diegok smls_: looks like a good idea... now I should understand why it isn't working for me :-( 21:30
dha Ironic. I'd expect Python to print 3 lines of whitespace.
;-)
diegok dha: or tabs...
smls_ "Rakudo version 2015.07.1-11-ge3314c8 [no line-editing support enabled]. Type 'help' for info." -- an intro message like this could really increase discoverability of the Linenoise feature... 21:31
dha Yeah, I just found something about Linenoise today. I would have liked to have found it some time ago. :-) 21:32
smls_ diegok: With "isn't working" you mean that even after installing the Linenoise Perl 6 module, the REPL doesn't have line editing support for you? 21:33
diegok smls_: right :-/
21:34 Fleurety joined
smls_ How did you install it? I installed Rakudo and panda via rakudobrew, and Linenoise via "panda install Linenoise", and it works for me. 21:35
diegok smls_: Linenoise module seems to be installed as I can 'use' it but has no effect :-/
smls_: I've installed rakudo-star using homebrew today 21:36
dha Huh. I just got this error message when running the perl6 after installing Linenoise: "Cannot locate native library 'liblinenoise.dylib': dlopen(liblinenoise.dylib, 1): image not found"
?
diegok smls_: and installed Linenoise with panda
smls_ strange
21:37 amurf joined
diegok dha: rjbs.manxome.org/rubric/entry/2081 ? 21:37
dha Thanks. I'll see if that helps. 21:41
21:42 amurf left
dha Out of curiosity, where would panda be installing things? 21:46
Ok. used rjbs's fix, and I'm not getting the error, but the line editing still isn't working. :- 21:53
/
21:53 colomon_ joined
dha Ah. Now it's working. yay. 21:53
diegok hm... looks like I should use rakudobrew again :-) 21:54
21:55 dolmen joined
lizmat has wrought another P6W: p6weekly.wordpress.com/2015/07/27/...e-present/ 21:57
21:58 colomon joined 21:59 dha left 22:00 colomon_ left
lizmat good night, #perl6! 22:01
colomon o/
yoleaux 11:36Z <jdv79> colomon: smoke is stalled
22:06 colomon left 22:09 dha joined 22:15 kaare_ left 22:19 Begi joined, beastd left 22:20 fling left 22:26 fling joined, rurban joined 22:27 dolmen left, rurban1 left 22:30 Begi left 22:34 bjz left 22:38 llfourn joined 22:39 rurban left
diegok smls_: it just works with a fresh rakudobrew install, yay! 22:41
22:43 llfourn left, skids joined 22:48 dha left 22:51 dha joined 22:54 fling left 22:58 Mhoram left, fling joined 22:59 Mhoram joined 23:00 khw joined
skids m: class Foo is Mu { }; say Foo.new ~~ Failure; 1; 23:01
camelia rakudo-moar 174049: OUTPUT«X::Multi::NoMatch exception produced no message␤ in block <unit> at /tmp/pPbuN9_oSG:1␤␤»
23:02 smls_ left
skids Yeah. Because one of the things X::Multi::Nomatch.message does is take the Foo.new and ~~ Failure with it. And off in circles we go. 23:02
Similar to the problem that necessitated github.com/rakudo/rakudo/pull/479 23:04
23:05 amurf joined
skids nine: ^^^ 23:05
Really some of the core X:: classes are way too optimistic about how many language features they are allowed to get. 23:09
*use
23:16 spider-mario left, lsm-desktop left 23:28 lsm-desktop joined 23:30 dha left 23:43 BenGoldberg_ joined 23:46 Mhoram left 23:47 leont left 23:48 Mhoram joined 23:54 bloomburgers joined
bloomburgers anyone know if rakudo supports method stubs in roles? I seem to be getting errors 23:55

Powered by Cro and the Raku® Programming Language.

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

Thank you!