»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg p6eval perl6: ... | irclog: irc.perl6.org/ | UTF-8 is our friend!
Set by sorear on 4 February 2011.
00:11 benabik left 00:22 benabik joined
colomon sorear: I mean lc, uc, tc, foldcase, etc. 00:23
sorear oh! 00:25
TimToady colomon: irclog.perlgeek.de/perl6/2012-07-30#i_5861524 seems like a bug; shouldn't the a's total to 5? 00:27
oh, no, it 00:29
colomon n: my $x = KeyBag.new({a => 3}); my $y = KeyBag.new({a => 2}); my $z = ($x ⊎ $y); say "z = ", $z; 00:30
p6eval niecza v19-15-g051783d: OUTPUT«z = bag("a" => 5)␤»
TimToady it's the absence of ⨄ that is maybe the bug
wait, where did you find that
colomon TimToady: which?
TimToady nevermind, got the wrong op 00:31
got the "uppercase" on
*one
which presumably means the same as [⊎] :)
sorear colomon: wondering if you might be interested in this month's release 00:32
colomon has to admit that the various bag union operators are virtually indistinguishable in the IRC font he uses
sorear: when should it be?
00:32 skids joined
TimToady much like Σ is just [+] 00:32
colomon TimToady: wait, what? (I mean, I understand the sum, but not the set/bag operator equivalent.) 00:33
TimToady .u ⨄ 00:34
phenny U+2A04 N-ARY UNION OPERATOR WITH PLUS (⨄)
TimToady I take the N-ARY to mean "reduction"
colomon huh
TimToady as opposed to the lowercase one, which is binary 00:35
.u ⊎
phenny U+228E MULTISET UNION (⊎)
colomon agh, what?!?
TimToady I'm not suggesting anything is wrong
I just forgot ordinary union with bags was "max" 00:36
colomon wildly increases the size of his font, trying to tell the difference between the two.
I've gone as big as I can, and I still can't tell the visual difference between those two symbols. 00:37
TimToady maybe your font just uses the same glyph
the N-ARY one is just uppercase sized here instead of lowercase sized 00:38
colomon which one does niecza implement? (he asked, dreading the answer)
.u ⊎
phenny U+228E MULTISET UNION (⊎)
colomon \o
\o/
TimToady "I'm not suggesting anything is wrong"
colomon stops panicking 00:39
.u ⊍
phenny U+228D MULTISET MULTIPLICATION (⊍)
colomon I seem to recall mjd thought multiset multiplication was completely useless.
TimToady I seem to recall thinking the same thing :) 00:40
colomon in the discussion we had back when I added ≼ and ≽
I just assumed that if someone bothered to assign a unicode codepoint to it, it must be used somewhere. :) 00:41
TimToady all I can figure is that, if you're doing multiset multiplication, the numbers you have are the viewpoint from a different dimension
so looking from the x direction I see 2 a's, and looking from the y direction I see 3
so it's only good for when counting projection images
moritz japhb_: I didn't notice we had non-generated files in html/ 00:42
japhb_: I'll probably undo the commits and simply place favicon.ico into html/ directly
TimToady btw, I don't want people to have to say "use Set", on the "math just comes" principle
japhb_ moritz, yeah, the bits needed to support the autocomplete searching
colomon sorear: about the case routines, the C# ones appear to just be wrong. I've implemented more correct versions the same way I did tc, but that approach (looking up the table for each character) is wildly inefficient. 00:43
TimToady if unicode slows down our compilers drastically, we need to fix our compilers
japhb_ Fair enough. I'm happy either reverting, or moving forward to having all static files in html-files, and html/ being always completely generated, depending on whether you want it to act like a blib/ or not.
colomon TimToady: It's just my implementations which suck. 00:44
There are very obvious ways to improve the code, they just make it a bit trickier to implement.
moritz japhb_: mostly I want an easy way to clean out html/ when some pod files or routines are deleted or renamed
japhb_ Time for a Makefile with 'html' and 'clean' targets? 00:45
Maybe 'sync' and 'missing' too
00:52 am0c joined
dalek c: a792008 | moritz++ | html (2 files):
Revert "[htmlify] add copying of static files; add favicon"

This reverts commit d5477a86d741516eac6225eb4e75e7d1be49f3e9.
There are still lots of static files in html/, so toss this attempt which just makes things more complicated		 00:56
c: d0f8a02 | moritz++ | html/favicon.ico:
add favicon
c: ee7934d | moritz++ | .gitignore:
make .gitignore more specific
moritz r: say Grammar.^mro 00:58
p6eval rakudo c1bfbb: OUTPUT«Grammar() Cursor() Any() Mu()␤»
01:02 benabik left
dalek c: 8a04f29 | moritz++ | lib/Grammar.pod:
document Grammar
 01:05
01:06 scott___ joined, japhb_ left
moritz n: say Grammar ~~ Cursor 01:06
p6eval niecza v19-15-g051783d: OUTPUT«True␤»
01:08 Guest47300 left
dalek c: c03644e | moritz++ | lib/Cursor.pod:
document Cursor
 01:11
moritz if Cursors have a public API, and anybody knows it, please add it :-) 01:12
01:12 Khisanth joined
ruoso great news on the switch to QAST... I imagine the ammount of PIR in rakudo is now much smaller 01:13
moritz ruoso: not in rakudo directly, but in its toolchain
01:13 vinian joined
moritz the PAST -> QAST and POST -> PIRT switches got rid of the last big remaining chunks of PIR code 01:14
ruoso that's nice...
so porting rakudo to different runtimes is now a real possibility
moritz aye 01:17
QAST also has less parrotisms than PAST
ruoso so I guess the first step in such adventure would be to port nqp itself 01:21
moritz well, one possible first step 01:22
you can also run the nqp-on-parrot to generate whatever code you want to generate
ruoso I guess that would be an intermediate step anyway 01:23
since nqp would need to emit code for other vm prior to running in other vm
(i mean, for the port to be useful)
moritz right 01:24
ruoso is there much pir emitting in rakudo? or is that all hidden behind QAST?
pmichaud there's very little pir in rakudo itself. 01:25
ruoso includding in rakudo's output?
pmichaud there are a few places that hide it behind "pir::". There are one or two Q:PIR's left. 01:26
I don't quite understand "in rakudo's output". We still compile to PIR, yes.
moritz and most references to PIR in Perl6::Actions are wrapped in QAST::VM.new(pirop => $whatever)
pmichaud but compiling to PIR is a function of the backend, not of anything specific to rakudo.
moritz pmichaud: I guess the motivation behind the question was, how much would you need to modify Actions.pm to generate something else than PIR 01:27
ruoso yes...
moritz a quick glance tells me that most direct PIR stuff is related to exceptions
which is pretty VM specific
pmichaud not too much, although you still have to replace the custom C-ops, too.
moritz tries to sleep again, now that $daughter has finally done the same 01:28
pmichaud TimToady: STD.pm6 has a few instances of things like token trait_mod:handles { <sym>:s <term> } 01:33
am I understand correctly that :s there doesn't have any effect other than to add a <.ws> after <term>; i.e., the above is equivalent to 01:34
token trait_mod:handles { <sym> <term> <.ws> }
?
(I realize that particular instance can now be better written as rule trait_mod:handles { <sym> <term> } ... I just want to confirm that the whitespace after :s isn't treated as sigspace, per S05) 01:35
afk, walk, bbiaw 01:36
01:37 plobsing joined 01:44 vinian left
TimToady I think when that was written :s did introduce <.ws> directly after 01:50
in any case, that seems to be the intent there
and, in fact, the 'endsym' for handles requires whitespace there 01:54
since term doesn't traverse whitespace, I'd say STD is still turning on :s directly after
iirc, the :s not turning off whitespace right after was a brief transitional form to the current formulation of not doing <.ws> unless something real was matched 01:55
diakopter is #perl6 in a punctuation or equilibrium 01:56
TimToady Gould only knows 01:57
01:59 plobsing left 02:00 vinian joined
TimToady should switch STD to the new sigspace policy.. 02:02
02:06 benabik joined 02:07 gongyiliao left, gongyiliao joined 02:36 Psyche^ joined 02:38 Patterner left, Psyche^ is now known as Patterner
sorear colomon: release today-ish. 02:40
colomon sorear: oh. I've got to get to bed in the next couple of minutes, I didn't get enough sleep last night and had a long day of three-year-old wrangling today. And if I actually had any brainpower at the moment, I'd really need to be $working anyway... 02:42
sorear colomon: fair 02:44
02:58 orafu joined 03:02 gongyiliao left 03:14 scott___ left, scott___ joined 03:20 adu joined
pmichaud TimToady: so, STD.pm6 is transitional and S05 is correct (whitespace after :s and other modifiers isn't sigspace)? That would definitely be my preference. 03:25
03:41 vinian left 04:01 gongyiliao joined 04:05 vinian joined 04:06 larks joined 04:08 larks left, larks joined 04:17 telex joined
TimToady spec is correct, sigspace is sigspace only after a "real" match of some kind 04:24
working on a STD patch currently
heh, now it won't parse /:i foo/ :) 04:28
04:28 telex left
TimToady hmm, roast has a lot of this error: 04:29
Placeholder variable $^k may not be used here because the surrounding block takes no signature at /home/larry/perl6/roast/S03-operators/bag.t line 7:
------> $x.keys.sort.map({ $^k ~ ':' ~ $x{⏏$^k} }).join(' ')
maybe my t/spec is out of date though... 04:30
04:30 cognominal__ joined
TimToady actually, there are still 5 instances of that 04:31
04:33 cognominal_ left 04:37 telex joined 04:40 cognominal_ joined 04:43 cognominal__ left 04:44 adu left 04:55 jaldhar joined
dalek ast: 4d00912 | larry++ | S0 (4 files):
$^k not legal inside hash subscript
 04:57
d: f43a358 | larry++ | STD.pm6:
1st whack at new sigspace semantics

Only regex tokens that set $*SIGOK will enable following sigspace.		 05:05
05:21 birdwindupbird joined 05:33 aloha joined 05:44 kaleem joined 06:04 xinming joined 06:24 SamuraiJack joined
arnsholt r: my Int foo { ... }; say 'alive' 06:27
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤Malformed my␤at /tmp/bB4PKvFJYa:1␤»
arnsholt r: Int sub foo { ... }; say 'alive' 06:28
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤Confused␤at /tmp/DHNdeoiJOX:1␤»
06:44 xinming left 06:50 wtw joined 06:51 kaare_ joined 07:09 brrt joined 07:12 cognominal_ left, cognominal joined 07:47 brrt left, bbkr joined, bbkr left 07:48 xinming joined 07:56 crab2313 joined 07:57 fhelmberger joined
moritz r: my Int sub foo() { }; say &foo.returns 08:01
p6eval rakudo c1bfbb: OUTPUT«Mu()␤»
jnthn morning o/ 08:05
moritz \o
japhb o/
08:09 GlitchMr joined
dalek kudo/io-pipe: 271741a | (Geoffrey Broadwell)++ | src/core/IO.pm:
Absolute minimum changes needed to support opening pipes for reading and writing
 08:14
japhb I need to head to bed, and have only a very slow computer here ... can someone please spectest the above commit? 08:15
moritz japhb: will do
japhb ++moritz # Many thanks
good night & 08:16
moritz sleep weel japhb++
*well
dalek c: 4d92362 | (Geoffrey Broadwell)++ | type-graph.txt:
Fill in role lists for Perl6::Metamodel::*HOW
 08:18
08:19 daxim joined
masak good antenoon, #perl6 08:25
tadzik good noonteraf, masak 08:28
crab2313 masak: \o 08:29
dalek href="https://perl6.org:">perl6.org: fce57f6 | (Gabor Szabo)++ | source/documentation/index.html:
link to the perl6maven site instead of the old pages on szabgab.com
 08:30
masak tadzik: I see what you did there 08:32
moritz \o masak, tadzik, crab2313 08:33
szabgab++ # blogging, p6maven
and just yesterday I wrote docs for Junction :-)
08:33 birdwindupbird left 08:35 birdwindupbird joined
dalek c: 792ea01 | moritz++ | lib/Junction.pod:
[Junction] add See Also section with links
 08:36
08:39 dakkar joined 08:51 dayangkun joined 09:03 kaare_ left 09:11 bbkr joined 09:14 Coleoid joined 09:19 fgomez left 09:20 dayangkun left
arnsholt moritz++ # trait-exceptions 09:29
09:42 bbkr left 09:51 Vakaris joined
moritz arnsholt: does it work now for you? 09:51
arnsholt The of case looks like it works as expected 09:54
Same segfault issue in the returns handler though
09:59 sergot joined
sergot hi o/ 10:00
tadzik \o
moritz arnsholt: returns is a trait_mod on Routine, right?
masak sergocie! \o/ 10:06
sergot \o/ 10:10
10:11 vinian left
arnsholt moritz: Yeah 10:12
The problem seems to be triggered by the exception throwing, somewhere above the trait_mod stuff
I'm digging around in gdb, but I don't have a good idea what's going on yet 10:13
moritz arnsholt: want to push the whole thing to a branch, so that I can play with it?
not that I have a track record in fixing segfaults or anything, but I did work with exceptions before :-)
arnsholt I think the patch in gist.github.com/3206774 is all there is to it 10:16
My findings so far are:
1) The segfault happens because something that's not a sixmodelobject gets passed to Rakudo_cont_decontainerize 10:17
2) The decontainerization is part of inspecting the arguments for multidispatch
3) There's a single argument, and a single md candidate which can only take exactly one argument 10:18
moritz could it be that nqp::isnull($returns); returns a parrot integer, and prefix:<not> dispatching dies?
arnsholt Oooh. That may very well be it! 10:19
moritz tries a fix
... unless nqp::p6bool(nqp::isnull(...))
arnsholt Oooh, looks like it 10:21
moritz it gets through the parse stage that way
(it could even be that 'unless' doesn't need the nqp::p6bool thingy, being provided by the compiler. But prefix:<not> is a normal multi) 10:22
phenny: tell japhb spectest on the io-pipe branch is clean here. Feel free to merge! 10:23
phenny moritz: I'll pass that on when japhb is around.
arnsholt Yep. Looks like it's working 10:24
moritz $ ./perl6 -e 'sub f(--> Str) returns Int { }'
===SORRY!===
Redeclaration of return type for f (previous return type was Str)
at -e:1
arnsholt Awesome! 10:25
10:25 scott___ left
moritz spectesting now 10:25
jnthn arnsholt++, moritz++
arnsholt Looks like the problem is that nqp::isnull became turned into a Parrot PMC 10:27
I'm guessing assignment to an intermediate P register turned it into an int PMC
moritz in Integer probably
s/in/an/ 10:28
arnsholt Yeah
10:32 crab2313 left, araujo joined
moritz spectest passes 10:35
10:38 mucker joined 10:39 JimmyZ joined
JimmyZ hello, anchor with space doesn't suppot firefox, ie: doc.perl6.org/type/Array#Methods%20...Positional 10:49
jnthn I'm pretty sure it's firefox that's doing the non-supporting :P 10:50
masak it's also not supported by HTML 4.01. 10:51
mhasch no, I mentioned it a few days ago. The anchor would have to have spaces where the URL has %20 for the link to work.
better no spaces at all
JimmyZ well, they should be replaced by '-' or something 10:52
mhasch precisely. moritz might accept a patch :-) 10:53
masak I'm absolutely sure he will. 10:54
actually, either of us will. :) 10:55
mhasch preferring - over _ would be the perl6ish thing to do, too ...
arnsholt jnthn: One thing that mystifies me, now that I've thought about it: Why does the code explode in trait_mod:<returns> but not trait_mod:<of>? They had identical bodies and signatures...
jnthn arnsholt: That is mysterious...maybe just that one gets triggered in the setting compilation but not the other? 11:07
11:07 MayDaniel joined 11:09 sftp joined, Vakaris left 11:14 Vakaris joined
arnsholt Right. The world behaves a bit differently when you're compiling the setting, I suppose 11:21
11:22 dbr joined
arnsholt tries something 11:24
masak Fortress project wrapping up: blogs.oracle.com/projectfortress/e...rapping_up -- lambda-the-ultimate.org/node/4570 11:26
I've always liked Fortress for what it tries to do.
"There is much work yet to be done in designing and implementing programming languages, and we believe that many of the ideas mentioned above will be important, in some form, in future language designs." +1 11:28
arnsholt jnthn: Yeah, that's it. Changing a returns to of in setting code triggers the segfault again 11:29
11:29 kaare_ joined 11:31 Vakaris left
masak from the lambda-the-ultimate thread: "Productizing languages is difficult". 11:32
moritz arnsholt: do you want to commit the patch, or should I? 11:38
arnsholt I can commit it
dalek kudo/nom: 1b2db80 | (Arne Skjærholt)++ | src/core/traits.pm:
Throw an exception when a sub gets assigned several return types.
 11:39
11:40 Vakaris joined
arnsholt I tried to implement the "Int sub foo()" case as well, but it required more grammar tuits than I had right then 11:40
moritz arnsholt: usually the typ constraint is in $*OFTYPE somehow 11:41
moritz tries a patch 11:42
arnsholt std: Int sub foo { ... }
p6eval std f43a358: OUTPUT«===SORRY!===␤Two terms in a row (listop with args requires whitespace or parens) at /tmp/id5I2dGN4D line 1:␤------> Int ⏏sub foo { ... }␤ expecting any of:␤ infix or meta-infix␤ infixed function␤ statement modifier
..loop␤Parse f…
moritz r: my Int sub foo() { }
p6eval rakudo c1bfbb: ( no output )
moritz r: my Int sub foo() { }; say &foo.returns 11:43
p6eval rakudo c1bfbb: OUTPUT«Mu()␤»
arnsholt Ah, right
jnthn It always has to has a scope declarator first
arnsholt Quite 11:44
jnthn
.oO( my oh my, the state of my English... )
moritz it might be as simple as $code.signature.set_returns($*OFTYPE.ast) if $*OFTYPE;
moritz tries that patch
11:44 JimmyZ left
arnsholt Yeah, that should do it, and an exception if the signature already has a value 11:44
moritz right 11:45
interesting that the Fortress folks found STM to be a good choice 11:46
because the last things I've heard about STM was that it was overhyped and doesn't work too well in practise
maybe they "just" did it right :-) 11:47
the "Symmetric multimethod dispatch" is probably close to what p6 has
$ ./perl6 -e 'my Int sub f() { }; say &f.returns' 11:49
Int()
now for exceptions...
11:49 JimmyZ joined 11:52 skids left 11:58 Vakaris left 12:03 gongyiliao left 12:08 Chillance joined 12:09 Coleoid left 12:13 Vakaris joined 12:15 sftp left 12:24 spider-mario joined
pmichaud good morning, #perl6 12:24
Pleiades` morning! 12:25
JimmyZ morning, pm
moritz good am, pm
jnthn o/ Pm 12:26
12:29 GlitchMr left, Vakaris left
moritz my currently-in-the-works patch already found a roast bug 12:31
arnsholt Nifty!
12:31 GlitchMr joined
moritz my List sub f { return () } # and it expected a call to that to live 12:31
12:32 JimmyZ left 12:33 JimmyZ joined 12:41 JimmyZ_ joined 12:44 JimmyZ left, JimmyZ_ is now known as JimmyZ
dalek kudo/trait-exceptions: 3a20b76 | moritz++ | / (2 files):
remove some deprecated subs and methods
 12:48
kudo/trait-exceptions: f4b83df | moritz++ | tools/build/NQP_REVISION:
bump NQP revision to get dyncall callbacks, arnsholt++
kudo/trait-exceptions: 705fabb | (Felix Herrmann)++ | src/core/Main.pm:
make Main.pm output a usage message on an unexpected named argument instead of bind failing

Signed-off-by: Moritz Lenz [email@hidden.address]
kudo/trait-exceptions: 07faa45 | moritz++ | / (4 files):
Merge branch 'nom' into trait-exceptions
kudo/nom: 6441778 | moritz++ | src/ (3 files):
refactor redeclaration of return type exceptions, enable "my Type sub f" return types

this refactors the check if a signature has a return type into a has_returns method in Signature, which reduces the monkey-getattr accesses into Signature
ast: 9db8364 | moritz++ | S06-routine-modifiers/scoped-named-subs.t:
fix a return type bug
ast: 864c54b | moritz++ | S04-exceptions/fail.t:
fudge a test for rakudo which we passed for the wrong reasons before
ast: 75c081f | moritz++ | S32-exceptions/misc.t:
tests for multiple return type exceptions
moritz oops, I didn't mean to push to trait-exceptions, sorry 12:49
12:50 atrodo joined
arnsholt Ah, good refactor. moritz++ 12:50
moritz: I recently set git config --global push.default current to avoid that =) 12:51
moritz arnsholt++ # thanks
dalek kudo/nom: 85541f0 | moritz++ | docs/ChangeLog:
update ChangeLog
 12:58
12:59 PacoAir joined 13:01 mscha joined
mscha perl6: my $a == 2|3; say $a==2; say $a!=2; 13:01
p6eval niecza v19-15-g051783d: OUTPUT«===SORRY!===␤␤Preceding context expects a term, but found infix = instead at /tmp/IeJyCpetGD line 1:␤------> my $a =⏏= 2|3; say $a==2; say $a!=2;␤␤Parse failed␤␤»
..rakudo c1bfbb: OUTPUT«===SORRY!===␤Preceding context expects a term, but found infix = instead␤at /tmp/uwbWTlW7pm:1␤»
mscha perl6: my $a = 2|3; say $a==2; say $a!=2; 13:02
p6eval niecza v19-15-g051783d: OUTPUT«any(Bool::True, Bool::False)␤False␤»
..rakudo c1bfbb: OUTPUT«any(True, False)␤False␤»
mscha perl6: say ((2|3) != 2); 13:03
p6eval rakudo c1bfbb, niecza v19-15-g051783d: OUTPUT«False␤»
moritz mscha: negated operators are special-cased for autothreading
eg $a != $b behaves as !($a == $b)
mscha But is this correct? Any of 2 or 3 IS unequal to 2, no?
moritz otherwise it would be very confusing
13:04 PacoAir left
moritz people have written things like if $a != any(1, 2, 3, 4) { } 13:04
and wondered why it was always True
mscha Ah, that's a good point. Very confusing if !(a == b) != (a != b)
moritz I must admit that in the beginning I was opposed to the change, but I now see the wisdom of it. TimToady++ 13:06
13:06 JimmyZ_ joined, PacoAir joined 13:08 JimmyZ left, JimmyZ_ is now known as JimmyZ
moritz p6rd: upcoming 13:19
p6rd niecza 2012-07-30, parrot 2012-08-21, rakudo 2012-08-23
mscha perl6: my @grades = <A B A A F C C>; say any(@grades) ne "A"; # I'm a straight A student!
p6eval rakudo c1bfbb, niecza v19-15-g051783d: OUTPUT«False␤»
moritz eeks, reports a date from the past :(
perl6: my @grades = <A B A A F C C>; say all(#grades) eq 'A'; # not so :-) 13:20
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤Unable to parse postcircumfix:sym<( )>, couldn't find final ')' at line 2, near ""␤»
..niecza v19-15-g051783d: OUTPUT«===SORRY!===␤␤Unable to parse argument list at /tmp/2kDjVuHHpX line 1:␤------> my @grades = <A B A A F C C>; say all(⏏#grades) eq 'A'; # not so :-)␤Couldn't find final ')'; gave up at /tmp/2kDjVuHHpX line 1 (EOF):␤-----…
moritz perl6: my @grades = <A B A A F C C>; say all(@grades) eq 'A'; # not so :-)
p6eval rakudo c1bfbb: OUTPUT«all(True, False, True, True, False, False, False)␤»
..niecza v19-15-g051783d: OUTPUT«all(Bool::True, Bool::False, Bool::True, Bool::True, Bool::False, Bool::False, Bool::False)␤»
moritz perl6: my @grades = <A B A A F C C>; say so all(@grades) eq 'A'; # not so :-)
p6eval rakudo c1bfbb, niecza v19-15-g051783d: OUTPUT«False␤»
mscha I guess it's impossible to make everything DWIM with junctions. 13:22
13:22 bluescreen10 joined
moritz it's impossible to make everything DWIM, period :-) 13:23
which hilights the need for good docs 13:24
13:25 skids joined 13:28 JimmyZ_ joined 13:30 JimmyZ left, JimmyZ_ is now known as JimmyZ
dalek c: de7f6db | moritz++ | lib/Junction.pod:
explain autothreading and negated operators
 13:33
[Coke] me; 13:34
masak mscha: any(...) ne is to be read as "not any equal" (which was false in your case), not as "any not equal" (which was true). 13:35
[Coke] ww
moritz masak: we've discussed that before mscha++'s example 13:36
[Coke] reads gabor's maven artile about junctions and is left wondering "if they're like sets, why do we have sets and junctions".
*article
moritz they aren't like sets :/
pmichaud yeah, we've been through that a few times before. 13:37
I wonder if it makes more sense to think of them as predicates.
13:37 benabik left
pmichaud or existential quantifiers 13:37
13:39 benabik joined 13:40 tokuhiro_ joined 13:43 tokuhiro_ left, tokuhiro_ joined
dalek p/sigsp: ba8d67c | pmichaud++ | src/QRegex/P6Regex/Grammar.nqp:
Update to handle leading spaces in p6regex arglist.
 13:45
13:52 JimmyZ left
pmichaud adds a comment re junctions/sets to the p6maven article 13:53
moritz doc.perl6.org/type/Junction now explains negated ops + autothreading 13:59
masak what's the url to szabgab's article? 14:00
pmichaud perl6maven.com/perl6-is-a-value-in-...-of-values
masak thank you.
pmichaud reads the doc 14:01
14:02 fhelmberger left
masak in szabgab's defense, he just says "Similar to a set." -- and only when talking about the un-orderedness of junctions. 14:03
moritz is there a way to specify fallbacks for lexical lookups in Perl 6? like AUTOLOAD/CANDO for methods
(not that I particularly want one, just curious)
masak r: my $weekdays; say $weekdays.^name
p6eval rakudo c1bfbb: OUTPUT«Any␤»
masak r: my $weekdays = any <Monday Tuesday Wednesday Thursday Friday Saturday Sunday>; say $weekdays.^name 14:04
p6eval rakudo c1bfbb: OUTPUT«Junction␤»
jnthn moritz: Don't believe so. A lot hangs off the lexpad being immutable from CHECK time on.
Like 14:05
r: foo()
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤CHECK FAILED:␤Undefined routine '&foo' called (line 1)␤»
masak moritz: if you specified such a fallback for lexical lookups, would it then be impossible to get an error when typo'ing a variable?
moritz masak: only if the fallback creates the error
pmichaud masak: I agree that "similar to a set" is appropriate. However, my experience has been that without the added caution against thinking about junctions as sets, a lot of newbies make the association "junction" == "set" and things get complicated from there.
masak pmichaud: I agree about that, too.
moritz masak: just like you only get errors for failed method dispatches in the presence of AUTOLOAD if the fallback creates the error 14:06
masak I agree that 'junctions == sets' doesn't capture what junctions are, and will create misunderstandings.
moritz: sounds reasonable. but for methods it happens at invoke time. for lexical lookup, at parse time.
moritz: (that is, I understand if the more static lexical lookups don't support this) 14:07
moritz well, such a fallback could be present at compile time
PerlJam moritz: what would you do with a lexical lookup fallback? 14:08
moritz ie at lookup time, just the same as with method fallback
PerlJam: not sure. Maybe evil DSL hackery :-)
pmichaud moritz: what's an example of looking up a lexical that cannot be resolved at compile-time, ooc?
moritz 16:03 < moritz> (not that I particularly want one, just curious)
14:08 GlitchMr left
moritz pmichaud: my $name = prompt; say ::($name); 14:09
pmichaud we do have things like OUTER:: and stuff.
moritz but that's kinda beside my point. If there'd be a fallback, it would operate at compile time too
PerlJam moritz: so ... give no use-case, it probably shouldn't exist :)
pmichaud OUTER::{$name} might work.
moritz you need it parse time to determine if it's followed by a term or operator 14:10
masak er, what?
pmichaud r: my $a = 'alpha'; my $name = '$a'; { say OUTER::{$name}; }; # just curious
p6eval rakudo c1bfbb: OUTPUT«alpha␤»
masak \o/ 14:11
pmichaud so, you want a fallback for
r: my $a = 'alpha'; my $name = '$b'; { say OUTER::{$name}; }; # just curious
p6eval rakudo c1bfbb: OUTPUT«Any()␤»
pmichaud looks like //
r: my $a = 'alpha'; my $name = '$b'; { say OUTER::{$name} // 'beta'; }; # just curious
p6eval rakudo c1bfbb: OUTPUT«beta␤»
pmichaud r: my $a = 'alpha'; my $name = '$b'; { say OUTER::.exists{$name}; }; # just curious 14:12
p6eval rakudo c1bfbb: OUTPUT«Not enough positional parameters passed; got 1 but expected 2␤ in method exists at src/gen/CORE.setting:6465␤ in block at /tmp/8Lf9ULYrLk:1␤␤»
pmichaud hmmmm.
might want to add .exists to Stash
or... something.
14:12 cognominal left
pmichaud oh, wait. 14:13
r: my $a = 'alpha'; my $name = '$b'; { say OUTER::.exists($name); }; # just curious
p6eval rakudo c1bfbb: OUTPUT«0␤»
jnthn Stash inherits from Hash. :)
pmichaud from Parrot Hash? 14:14
jnthn Huh?
geekosaur has been sitting there wondering if those braces were supposed to be parens...
jnthn No, from Hash :)
pmichaud .exists returned zero there
jnthn Perl 6 Hash :)
14:14 cognominal joined
pmichaud looks. 14:14
moritz maybe a hash bug
14:14 tokuhiro_ left
jnthn Oh, though 14:14
That's not a Stash
pmichaud I would've expected False.
jnthn It's a PsudoStash I guess
moritz r: say Hash.new.exists('foo')
p6eval rakudo c1bfbb: OUTPUT«False␤»
14:14 wtw left
jnthn r: say OUTER::.WHAT 14:14
p6eval rakudo c1bfbb: OUTPUT«PseudoStash()␤»
moritz r: say PseudoStash.^mro 14:15
p6eval rakudo c1bfbb: OUTPUT«PseudoStash() EnumMap() Iterable() Cool() Any() Mu()␤»
jnthn r: say OUTER::.^methods(:local)
p6eval rakudo c1bfbb: OUTPUT«new at_key bind_key exists␤»
jnthn And it overrides exists, and that's probably where le bug is.
I blame the guy who wrote PseudoStash :P
14:15 sergot left
jnthn ah, I see it 14:16
pmichaud okay, I just patched up Rakudo's grammar for the new leading space issues, and it compiled on the second try. Now I guess I have to go ahead and kill my laptop battery with a full spectest run.
wasn't expecting that to happen this morning :-)
moritz the first branch just goes to nqp::existskey
jnthn yeah
compiling a fix 14:17
pmichaud nqp::p6bool that.
jnthn right
moritz pmichaud: you can push to a branch and let me do the spectest, if you want to save battery live :-)
pmichaud will do that... but I also think I'll have enough battery to last the rest of my son's soccer practice :) 14:18
pmichaud is currently working from the middle of a park.
dalek kudo/sigsp: 7300248 | pmichaud++ | src/Perl6/Grammar.pm:
First pass at sigspace fixes (for others to test).
 14:19
pmichaud there, pushed as rakudo/sigsp. note that you have to build that branch with the sigsp branch of nqp.
tadzik heh, that reminds me of i.imgur.com/nD0Hn.jpg :) 14:20
moritz pmichaud: willdo
tadzik (that's me working on last year's gsoc :P)
was a lot better than sitting in the office
moritz building nqp now 14:21
14:21 mscha left
jnthn Patch seems to work. 14:24
Will spectest.
(the PseudoStash one)
pmichaud I get a couple of spectest fails in s05; I'm guessing the tests may need updating also. 14:25
moritz might very well be the case
and you'll probably find that the now failing tests are TODOed for niecza :-) 14:26
speaking of which
sorear: CAN HAZ RELEASE PULESE?
pmichaud p6rd: next
p6rd niecza 2012-07-30, parrot 2012-08-21, rakudo 2012-08-23
moritz it's a bug in p6rd that it reports niecza's release date from the past 14:27
14:27 mucker left 14:30 sergot joined
moritz $ ./perl6 -Ilib t/spec/S05-grammar/protos.t 14:31
===SORRY!===
Unrecognized regex metacharacter < (must be quoted to match literally) at line 15, near "...> }\n\n "
that's proto token fred { <...> }
I guess that's deprecated proto syntax, but it wasn't intential that it's not parsed anymore 14:32
pmichaud correct.
moritz though Perl6::Grammar itself uses <...> quite a bit
14:33 GlitchMr joined
pmichaud looking. 14:33
moritz probably role regex_dev in Grammar.pm
s/role/rule/ 14:34
'{'[
| ['*'|'<...>'|'<*>'] <?{ $*MULTINESS eq 'proto' }> $<onlystar>={1}
pmichaud yes, that's likely it.
moritz yes, looks like that could be affected by the change
pmichaud need a space after the '{'
14:34 benabik left 14:35 kaare_ left
pmichaud oh, wait, no. 14:35
hrm.
14:36 kaare_ joined
pmichaud fixing. 14:36
dalek kudo/nom: ce9ae4e | jnthn++ | src/core/PseudoStash.pm:
Fix PseudoStash.exists to always return a Bool.
pmichaud compiling, testing. 14:37
moritz gist.github.com/3217474 14:38
14:38 JimmyZ joined
moritz the failures fro type.t and S32-exceptions/misc are likely due to sigsp being behind nom 14:38
pmichaud yes. 14:39
moritz ss/likeley //
dalek kudo/sigsp: 9c63f57 | pmichaud++ | src/Perl6/Grammar.pm:
Better handle <...> and other special forms in regex_def. moritz++
pmichaud see if that patch resolves the S05 stuff. 14:40
moritz recompiles
14:40 thou joined
pmichaud \o/ fixed it for me 14:43
moritz here too. t/spec/S05-metasyntax/angle-brackets.rakudo still aborts after test 35 14:45
pmichaud that's ... weird
moritz $ ./perl6 -e 'my @second = rx/\.**2/, rx/'.**2'/; "a" ~~ /<@second>/' 14:46
===SORRY!===
Unrecognized regex metacharacter } (must be quoted to match literally) at line 2, near ""
pmichaud probably another <?> after square bracket or | issue 14:47
in regex_var
looking
GlitchMr perl6: print srand
p6eval niecza v19-15-g051783d: ( no output )
..rakudo c1bfbb: OUTPUT«===SORRY!===␤CHECK FAILED:␤Calling 'srand' will never work with no arguments (line 1)␤ Expected: :(Int $seed)␤»
GlitchMr I guess it's impossible to receive current seed? 14:49
pmichaud I'm not even sure what test 36 is meant to test.
oh, I see. 14:50
kinda.
PerlJam GlitchMr: if you need a specific seed (for reproducability presumably), then you just set it to some known value. 14:51
moritz pmichaud: I guess it's really a NYI thing that we've been passing or failing for entirely the wrong reasons 14:52
pmichaud --ll-exception says it's in MAKE_REGEX
moritz still it might point to something that change in regex parsing which might or might not be what we want 14:53
pmichaud that's gotta be where the '}'
is
pmichaud installs 'ack' on his notebook, which for some reason hadn't been done yet.
hoelzro woo 14:54
ack++
GlitchMr ack?
hoelzro betterthangrep.com/
GlitchMr I've tried it, but I don't like how it works on regexpes
14:54 nightlord joined
hoelzro GlitchMr: what don't you like about it? 14:54
moritz I like very much that I can use it with p5 regexes, which I happen to know very well
hoelzro I happen to have an in with the developers ;)
GlitchMr ack '\*\+\?'
But, I guess I can use \Q, so whatever 14:55
moritz ack -Q '*+?'
GlitchMr Unrecognized escape \Q passed through in regex
or not
moritz ack -Q '*+?'
pmichaud I'm guessing we must've been not-really-passing that test.
pmichaud looks.
GlitchMr Interesting how "ack -Q '+'" doesn't work... probably I should return this bug 14:56
report*
moritz r: my @a = rx/.+/, rx/'.+'/; say ~@a
hoelzro GlitchMr: which version of ack are you on?
p6eval rakudo c1bfbb: OUTPUT« ␤»
pmichaud hahahahah 14:57
moritz pmichaud: probably related to those regexes stringifying empty
hoelzro GlitchMr: good catch; please do report it =)
pmichaud indeed, that's undoubtedly it.
moritz pmichaud: and then our failure to detect null patterns
GlitchMr ack-grep 1.92
...
hmmm... yeah... odl
old
hoelzro GlitchMr: btw, you should be able to do ack -Q -- '+'
pmichaud okay, so not really a sigsp regression.
hoelzro that works with my ack (2.00a01) 14:58
pmichaud cool, so the sigsp branch appears to pass all of the relevant spectests
GlitchMr ok... I'm now installing ack 2.00a01 14:59
14:59 birdwindupbird left
pmichaud okay, when I get back home I'll run a few more tests and then merge sigsp back into the master branches 14:59
(<30 mins left here in the park) 15:00
dalek ast: 7c36bd0 | moritz++ | S0 (2 files):
rakudo refudges

  * unfudge return type tests that now pass
  * skip regex tests that have been passing/running for the wrong reasons
GlitchMr Actually, I think I was using recursive awk instead of ack before 15:01
hoelzro all the more reason to try ack again =) 15:02
moritz urks, JSON::Tiny relies on the old regex behavior 15:03
GlitchMr I would like ack more if it would use finite automata (for example re::engine::RE2, but whatever)
15:04 cognominal_ joined
pmichaud looks 15:04
GlitchMr I guess I could modify ack to use re::engine::RE2 instead 15:05
hoelzro GlitchMr: well, throw it in the issue tracker for ack 2 =)
pmichaud rule TOP wants a space between the ^ and the [ 15:06
15:06 cognominal left
GlitchMr Because you know, grep uses finite automata 15:07
timotimo ack2 will be a perl6 program?
spider-mario more like ack6
timotimo yeah, because that's how long it'll take, right? ;) 15:08
GlitchMr Actually, I should make ack version for yourself
hoelzro timotimo: ack2 is already written =)
GlitchMr Which for example replaces space with regular expression working similarly to Perl 6 rules 15:09
moritz pmichaud: that alone isn't enough
pmichaud moritz: hmmm 15:10
I probably need to look at the code gen for the % and ~ items, then.
timotimo oh, didn't realize. my local ack is apparently outdated
pmichaud it can be of course "fixed" with the leading <?> trick, but we probaby want to make sure there's not a regex engine bug lurking there 15:11
moritz pmichaud: github.com/moritz/json/commit/1f01...b3c7af7766 fixes all but two errors
not ok 43 - JSON string «{ " s p a c e d " :[1,2 , 3\n...[42]» parsed 15:12
not ok 47 - JSON string «[ 0.5 ,98.6\n...[46]» parsed
are the remaining failures
pmichaud I'll have to look at it when I get back home
hoelzro timotimo: it's still in alpha, though
so distros haven't likely packaged it up
pmichaud I really would prefer to not have the <?>'s there. 15:13
15:13 kaleem left
moritz too 15:13
pmichaud at the moment I'm suspicious of the <.ws> handling around the ~'s
or maybe the %'s
moritz in the long run I'd even prefer <.ws> over <?>
more explicit in intent 15:14
pmichaud yeah, but it ends up calling <.ws> twice.
unless you do <.ws>token instead of <.ws> token
moritz -rule pairlist { <?> <pair>* % \, } 15:15
+rule pairlist { <?> <pair> * % \, }
should that make a difference?
pmichaud it might, depending on how the <.ws>'s are distributed inside of the quantification 15:16
15:16 spider-mario left
pmichaud that's the part I think I need to review. 15:16
moritz in general I want to allow <.ws> everywhere here (ie between <pair> and \, and \, and <pair>)
15:17 cognominal joined
pmichaud in the case of <foo>**3 and <foo> **3 the second includes <.ws> between repetitions 15:19
and yes, S05 says there's a difference between <pair>* % \, and <pair> * % \, 15:20
15:20 cognominal_ left
geekosaur (that is just a bit ugly) 15:22
moritz decommutes
pmichaud time fo rme to leave also 15:25
bbl
15:27 JimmyZ left
pmichaud TimToady: github.com/perl6/specs/issues/22 when you get a chance 15:29
bbl
15:37 nightlord left 15:59 Coleoid joined, sergot left
dalek kudo/nom: 271741a | (Geoffrey Broadwell)++ | src/core/IO.pm:
Absolute minimum changes needed to support opening pipes for reading and writing
 16:09
kudo/nom: 37f890a | (Geoffrey Broadwell)++ | src/core/IO.pm:
Merge branch 'io-pipe' into nom
16:11 SamuraiJack_ joined
pmichaud ....is :p the spec'd method for opening a pipe, ooc? 16:12
moritz no 16:14
16:14 SamuraiJack left, sudokode joined
moritz the part of S32::IO that talks about pipes is in the design insanity part 16:14
pmichaud I'm a little concerned about adding :p if it doesn't turn out to be spec. 16:15
it just feels wrongish.
TimToady we don't have to call all our IO constructors 'open'
moritz so, name it &pipe? 16:16
16:16 thelazydeveloper joined
TimToady can't think of a better name with only half a cup o' coffee in me 16:16
16:19 sergot joined
masak seems a good name to me. +1 16:19
pmichaud my only (very minor) concern with &pipe is confusion with the unix syscall of the same name.
at some point it's just IO::Pipe.new(...) 16:20
or even IO::Pipe(...)
moritz I don't know if we even need a separate IO::Pipe
pmichaud we don't have to give all of our IO constructors specialized function-call forms :-) 16:21
moritz actually IO::Path.pipe might be good fit 16:22
masak aye.
pmichaud I'd definitely like that better.
moritz senses another IO spec refactoring, though a much smaller this one 16:23
in fact I'd like Str.io to return an IO::Path 16:24
16:24 sammii joined, sammii left
moritz then IO::Path can support .open, .slurp, .spurt, .pipe 16:24
and since IO as a class is going away (to be replaced by IO::Handle), we teach people to say $str.io instead of $str.IO 16:25
or maybe $str.path
there's just a wee problem with that
IO::Path is currently Cool, and I'd like it to stay that way 16:26
so IO::Path.lines currently does what Str.lines does (break a string up into lines)
so, either we leave it at that, then we have to force people to write $str.path.open.lines
instead of $str.path.lines
or we make IO::Path.lines like IO::Handle.lines 16:27
but then not all "things that look like Str" (ie Cool) do the same when you call .lines on them
16:27 orafu left
moritz any opinions which way is less bad? 16:30
japhb just got here, so reading back 16:39
phenny japhb: 10:23Z <moritz> tell japhb spectest on the io-pipe branch is clean here. Feel free to merge!
16:39 am0c left
TimToady why should IO::Path be Cool? 16:42
japhb moritz, If IO::Path supports .slurp and .spurt, it feels like .lines makes some sense from the perspective of "methods that operate on the entire contents of the file" 16:43
And FWIW, I just did :p for pipe open because A) it's actually what's happening at lower layers, B) it made sense as a parallel to the other flags, and C) because I'd been told at one point a few weeks ago to assume that the syn wording for pipes was essentially ignorable 16:45
My apologies if I misinterpreted C), of course.
16:48 Coleoid left
moritz TimToady: because &dir returns IO::Path objects, and people expect to use them like strings 16:49
japhb: I still agree with C) :-)
japhb: and you see, even if we decide the patch was "wrong", it has still spawned an interesting discussion 16:50
japhb True enough. :-)
16:50 kaleem joined
TimToady can't we make them stringify without making them Cool? 16:51
moritz TimToady: stringify yes, but not otherwise behave like a string 16:52
like, regex-match, for example
well, we could, but we already have a mechanism for that, and that is Cool 16:53
16:53 dakkar left
moritz we *can* break the assumption that the return values from &dir behave string-like, but it'd make much file name handling code more verbose 16:54
TimToady well, maybe we can intercept lines instead 16:55
the primary purpose of IO::Path is to represent the identity of a file
strings are secondary
losing sight of that is a good way to get race conditions that lead to security holes 16:56
japhb TimToady, how do we make an IO::Path object that A) does not open or stat the file when it is created and B) does not fall prey to race conditions? 16:57
pmichaud japhb: C) is definitely a truth -- the existing IO spec is very loose at the moment. My question was more along the lines of "is a :p flag really what we want here?" 16:59
16:59 daxim left
TimToady readdir does not just return strings; it also returns inodes, and IO::Path should track that 17:00
it can at least check that the inode stays the same
japhb pmichaud, I'm torn between the part of me that dislikes having to say ':r, :p' or ':w, :p' and the part of me the dislikes lying to upper-layer programmers about what the underlying API actually is. I suppose I could assuage both parts with explicit ':rp' and ':wp' flags, but then that feels a little ugly from the orthoganality angle. 17:01
pmichaud japhb: by "underlying API" do you mean "Parrot", ooc? 17:02
TimToady I think if you open a path, and the inode changes, it should fail by default 17:04
one can defeat this by opening the string name, of course
japhb pmichaud, at least several layers down through parrot, yes. Though it does eventually have to split execution path to do popen(), I suppose. And of course, Parrot's design decision shouldn't matter for Perl 6's design.
pmichaud japhb: I'm sure this will come across totally wrongly to people that read it, but in some areas Parrot's design choices were deliberately anti-patterns from Perl 6. 17:05
17:06 kaleem left
japhb pmichaud, what are you thinking of when you say that? 17:06
pmichaud exceptions, IO, basic data types.... a variety of things 17:07
calling conventions
japhb thinks on that a moment
pmichaud there were many times when the responses I got were along the lines of "Parrot shouldn't follow the Perl [6] model in this case..." 17:08
japhb Well, intentionally different (I assume because of a desire to be more language-agnostic), or just baked in before anyone really understood Perl 6's designs?
pmichaud I'm not saying those choices were wrong for Parrot, just that Parrot's goals for a long time were very different from Rakudo/Perl 6 ones.
desire to be language-agnostic, yes. 17:09
TimToady desire not to be seen as favoring Perl
pmichaud anyway, these days I'm especially cautious about choosing an API because "that's what Parrot does in this case". The track record there isn't that good. 17:10
japhb shrugs
I can buy that.
pmichaud I should probably stop there. :)
17:11 Sorensen_ joined
TimToady respects it greatly when people stop there. :) 17:11
japhb In any case, I'm certainly open to better ideas, and happy to hack something in. I actually need working pipes, but I don't strongly care what they look like, TBH.
pmichaud japhb: (need working pipes) -- yeah, I agree we need something for now, so I'm not calling for immediate change or reversion of the patch :) 17:13
IO is one of those areas where we actually do need some exploration; it's just that :p doesn't feel quite right to me. It's good enough for interim, though.
Sorensen_ Hello, all. I think I may have a bug to report for Rakudo. But it may already be in the bug list, I don't know... It's a bit tricky to search with small keywords 17:14
moritz Sorensen_: what's the bug?
Sorensen_ ...How do you make the IRC bot guy do the perl codes?
moritz (yes, RT isn't easy to search for)
Sorensen_ like, "Hey Rakudo, eval me this" 17:15
moritz r: say 'OH HAI, Sorensen_'
p6eval rakudo c1bfbb: OUTPUT«OH HAI, Sorensen_␤»
moritz like this
Sorensen_ cool, okay, let me try this out...
17:15 vmspb joined
Sorensen_ r: my $any = any <a b c>; say $any; $any++; say $any; $any++; say $any; 17:15
p6eval rakudo c1bfbb: OUTPUT«any(a, b, c)␤any(b, c, d)␤any(c, d, e)␤»
Sorensen_ Okay, never mind. I guess the bug has been fixed! 17:16
I was using Rakudo 2012.04.1, I guess that is out of date now?
flussence very
moritz Sorensen_: there's a reason we do monthly releases :-)
Sorensen_ On my machine, I get (a b c), (a b c), (b c d). The first increment doesn't do anything.
moritz that's weird 17:17
flussence b: my $any = any <a b c>; say $any; $any++; say $any; $any++; say $any;
p6eval b 922500: OUTPUT«any("a", "b", "c")␤any("b", "c", "d")␤any("c", "d", "e")␤»
Sorensen_ Well, cool to see that four months makes such a difference.
flussence b: say $*PERL
p6eval b 922500: OUTPUT«name rakudo␤version 2011.07␤␤»
flussence interesting
Sorensen_ hmm
moritz well, b was an entirely different beast anyway :-)
Sorensen_ wait wait... maybe it's an issue with the REPL. Ha, I should have tried it all on one line like I did in here. It works when it's on one line; but the first time I write "$any++" and hit enter, the REPL gives me the old version 17:19
Like this...
> my $any = any <a b c>; any(a, b, c) > $any++ any(a, b, c) > say $any any(b, c, d)
eh, that didn't paste nicely...
moritz Sorensen_: note that $any++ returns $any, not the incremented any 17:20
Sorensen_ OH DUH
the smacking sound you hear the in background is my hand against my forehead
moritz and I thought for a moment an airplane had crashed :-)
Sorensen_ ++$any works as expected
(I was playing around with stuff from this post: perl6maven.com/perl6-is-a-value-in-...-of-values ... Yeah, I wasn't thinking there. The way the REPL works makes sense, it just wasn't how I was thinking about it.) 17:21
moritz is glad that szabgab brings people to play with p6 17:22
japhb pmichaud, perhaps methods/subs 'pipe-to' (the common write-to-process pipe), 'pipe-from' (the common read-from-process pipe), and 'pipe' (make a raw pipe)
moritz what does the raw pipe do?
japhb goes to check if that's already what the syn said
moritz likes pipe-to and pipe-from
japhb moritz, you get both ends. Do with them what you will 17:23
Essentially pipe(2) in *nix
masak Sorensen_: welcome. glad to see that your bug from four months ago was (1) fixed, and (2) not even a bug ;)
so, really not (1), I guess... :P
Sorensen_ Ha, I like that. Ciao all, thanks for the quick feedback. Best of luck with your programming. 17:24
17:24 Sorensen_ left
japhb Interesting, another item for :p as a flag on pipe -- .p is the test for 'is a pipe'. Perhaps pipe-to and pipe-from implemented in terms of open :p, which is considered the "low-level" way to do it. ;-) 17:27
s/flag on pipe/flag on open meaning pipe/
TimToady it bothers me slightly to have a flag that changes the meaning of a string in hindsight 17:28
also, we have to worry a bit about combinatorics if we get too many interacting options 17:29
japhb Ah, S32 has 'IO::Pipe.to', 'IO::Pipe.from', and 'IO::Pipe.pair'. I suppose I could do pipe-to, pipe-from, and pipe-pair ....
TimToady, yeah, we already have some silliness of which of :r :w :a is used with something that is or is not '-' 17:30
TimToady
.oO(open-to, open-from, open-pair...)
japhb TimToady, I think that is too general a set. Meaning, it sounds like it wants really magical open, of the "opens sockets if you ask it to" variety 17:31
TimToady I'm just point out the inconsistency 17:32
*ing
japhb Which inconsistency? I can see several angles of strange there, but I'm not sure what you're actually referring to.
TimToady why pipe-to, but open :w ? 17:33
japhb Hmmm, well, pipe's don't have the append/overwrite dichotomy. And you can have a R/W pipe, but it's a hectare of brambles. 17:34
17:34 orafu joined
japhb Yeah, I dunno. What is your preference, or do you have one yet? 17:35
TimToady no, I just enjoy complaining :)
moritz would like either IO::Path.pipe() with :r/:w or IO:Path.pipe-{to,from,pair}
japhb I will say that I find the way Perl 5 and bash both go, with all sorts of special combinations of <>+&| and so forth all meaning various things, gets confusing enough that for the rarer stuff I always have to check perlfunc to be sure. 17:37
IO::Path.pipe() doesn't make sense to me for two reasons: 1) You don't have a path involved in a pipe pair, and 2) the command you pipe to may be much more than just a path (e.g. having options/arguments) 17:39
flussence maybe IO::Path.pipe should call mkfifo()
TimToady it's more like IO::Identity in my mind, which why it was just IO before 17:40
so IO::Handle is closer to the truth than IO::Path
unless, of course, we redefine 'path' 17:41
but it represents "I already said how to get to this thing, and I don't want to say it again" 17:42
moritz well, IO::Handle is a handle, IO::Path is a path 17:44
and in my world, reading a directory returns paths, not handles
maybe your world is different :-)
and I don't see how inodes are any less prone to race conditions than paths, for that matter 17:45
japhb moritz, in my world, a directory returns "things I can open and query for name, path, and stat info".
It happens that Perl 5 gave me string representations of name, and expected me to get the others through contortions on that. But that always felt like an extra step. 17:47
Mind you, it was an extra step that in some sense reflected the underlying API.
17:48 fgomez joined
japhb Humm, I guess returning path info from a directory-contained object depends recursively on the path of the directory. Makes me wonder about symlinks, mounts, canonical names, and so forth. 17:49
moritz currently I see my approach (something simple, implementable very soon, easy to use) at odds with TimToady++'s musings (which seems to target for perfect and easy to use) 17:50
17:52 sftp joined 17:53 simcop2387_ joined 17:54 simcop2387_ is now known as simcop2387
masak I don't believe in aiming for perfect. I believe much more in mimicking things out there that already work well. 17:55
surely we're not the first ones to implement this functionality in a language library.
diakopter but "aiming for perfect" is just an expression of trying to define "well" in your statement about "work well" 17:56
masak if that's true, then how do moritz' and TimToady's approaches differ?
diakopter just amount of time willing to devote to the definition, I guess 17:57
japhb Because IIUC TimToady thinks paths-as-strings does not qualify as "working well"
17:57 SHODAN joined
masak I interpreted moritz as "get it working now", and (his description of TimToady, I haven't been following along too closely) as "get it right first". 17:58
mhasch fwiw I think the notion of an "atomic" open (taking a path, returning a handle) should at some point be exposed to the program; things like string munching for paths need abstraction but not "implicit-ness".
masak that was the worst use of parentheses, ever. sorry. :/ 17:59
mhasch :P
japhb masak: so ... many ... LISP ... jokes
masak :P 18:00
moritz well, the problem with "aiming for perfect" is that the IO specs have been trying that for years, and see what it brought us 18:01
just read the whole thing, and try not to cringe 18:02
(and remember that I've thrown out the worst 30% already)
masak I'm not sure the IO spec has been aiming anywhere in particular for years. 18:03
mostly it's been the victim of many chefs, and no clear vision.
moritz IMHO it's the victim of no implementation, many chefs, and no clear vision 18:04
japhb moritz, I'm essentially saying this as a man on the outside, but I'm not sure the problem was "aiming for perfect", but rather "design up front in an implementation vacuum"
moritz japhb: right, that was a big part of the problem
sorear good * #perl6 18:05
moritz \o sorear
japhb o/
masak sorear! \o/
flussence I can't help but wonder if all this would be easier after a big s/IO:://g 18:08
(i.e. stop trying to get all the subclasses playing nice and just have them work well on their own) 18:09
moritz flussence: try it and see :-)
flussence goes off to re-read that bit of S32... 18:10
18:16 crab2313 joined 18:18 crab2313 left
tadzik o/ 18:22
japhb o/ 18:25
masak o/
colomon o/ 18:31
18:31 fhelmberger joined
sorear o/ 18:31
jnthn \o 18:32
masak .oO( a non-conformist! )
jnthn :D 18:34
18:34 fridim_ joined
masak .oO( you are unique, just like everyone else ) 18:35
colomon
.oO( not me! )
 18:37
masak :P
sorear 'but some are more unique than others'
japhb I am utterly the same, unlike everyone else. 18:38
masak 'individual uniqueness is already here, it's just not evenly distributed'
sorear colomon: since you seem to be the only person who contributed to niecza (git log v19..HEAD) this month, would you mind summarizing what you've done for the announcement?
18:38 cognominal_ joined
TimToady well, if anything ever needed duck typing, it's I/O 18:39
my main concern is that we don't make the lazy way something we have to recant later in the interests of security (see 2-arg open) 18:40
18:42 cognominal left
TimToady and it's just really easy in the original Unix model of I/O for race conditions to sneak in between the name and the fd you think came from a particular name but didn't 18:42
colomon sorear: Added floating point support for printf/sprintf. Added simple versions of tc, tclc, and tcuc. 18:43
TimToady that's why the newer APIs are full of fdthis and fdthat
18:44 fhelmberger left
TimToady no true identity is in filenames (or commands) until they are opened 18:44
japhb Too bad open can be rather expensive. :-(
TimToady the race condition mostly shows up in doing stat when you should fdstat 18:45
hence my desire that $filename.IO ~~ :w work against an fd by default, not against a filename 18:46
or .io, or whatever it ends up
japhb Too bad inode ids aren't globally unique.
TimToady yes, well, you need at least dev as well, but a directory should only be in a single dev 18:47
I also recognize that this becomes more of a problem in the context of universal naming schemes
18:48 sivoais joined
TimToady to the extent that the only safe-ish ID that you can have in the long run is a cryptochecksum like git does 18:48
japhb Well, I was thinking about inode reuse, actually. Even on a single device ... the inode number space isn't large enough not to reuse them. Especially on filesystems that treat them as indices in some pre-allocated array.
TimToady there's really no guarantee that two probes of a URI will produce the same thing
japhb True.
TimToady japhb: but at least if the inode changes you have some clue that something happened 18:49
moritz TimToady: $filename.IO ~~ :anything working through a fd doesn't buy us anything at all
TimToady: because that file handle is GC'ed immediately
and what remains is the file name again
TimToady right, unless there's caching of some sort
sorear TimToady: except git uses SHA-1, which is widely regarded as almost broken *ducks* 18:50
TimToady *ducks* typing...
admittedly caching is not the solution either to security
japhb TimToady, there are some filesystems (I gather ext2 is one of these) that like to immediately reuse the inode of a just-deleted file.
TimToady makes you want one-use filename, so if you say $filename.IO ~~ :w and $filename.open it says "You can't open that now" 18:52
japhb sorear, IIRC Linus said back when SHA-1 was first seriously attacked that since he was using it for uniqueness amongst cooperative people, not for security against an attacker, that wasn't an issue for Git. Has this been revised?
18:53 cognominal joined, cognominal_ left
TimToady it's a bit more important for the Perl 6 Official Library to have something that can't be spoofed 18:53
(maybe)
masak ooh, HPMoR tommorrow? :)
sorear japhb: I think gitland has gotten a bit wider since then
japhb Of course. I just figured "sorear is usually more up to date on this sort of thing", so I thought I'd ask. :-) 18:54
TimToady it's just that, having been burned multiple times in the past, I know in my guts that file security cannot be a retrofit
japhb masak, hopefully not another delay like July 1. :-/
sorear there's also the question of 'how far will SHA-1 cryptanalysis get'?
masak japhb: one can hope.
sorear MD4 is so badly cracked that you can find collisions with pen & paper
Tene masak: hpmor progress update tomorrow
masak right. 18:55
japhb TimToady, in any case, I'm curious if there is prior art to putting a safe-but-still-fast-and-powerful interface on top of the painfully finicky unix IO API. I'd hate to lose all the power or performance in trying to gain the ability to save the programmer from themselves. 18:58
TimToady
.oO(use UNSAFE_STAT)
 18:59
moritz also, going through FDs is only possible if you have permissions to open the stuff you're talking about
TimToady yeah, I'm just grouchy today 19:00
masak TimToady: I'd hate to see us crawling up out of the ditch of perfection (if that's what happened to S16), only to fall back down again.
TimToady can usually figure out a way to have it both ways, and it doesn't seem possible here
japhb moritz, sadly, that's a clear fail of the Unix API right there (the lack of separation between "get a unique handle to" and "open for read or write") 19:01
moritz also let's not forget that there are non-UNIX systems out there 19:02
japhb moritz, I'm not. But unless this changed, the old Perl philosophy was "the world's APIs -- seen from the Unix API point of view" 19:03
At least, that's how I always saw it.
TimToady well, an fd is the only guarantee of identity in Unix, so I guess we have to live with that one way or another 19:05
also, it's object identity, not value identity, since files are mutable 19:06
19:11 cognominal left
TimToady I guess it bothers me that someone might do: $handle = $filename.IO; and think $handle ~~ :x is working on the fd when it's not 19:11
that situation doesn't arise in P5 because stat works directly on filename strings 19:14
IO in that case really has very little to do with I/O 19:16
almost want $filename.stat ~~ :x instead 19:17
I think that would reflect what's happening much better 19:18
japhb Well, the objection about $handle = $filename.IO is part of the reason for having $path = $filename.path instead. Handle is a crazy name for something that isn't actually a handle.
moritz TimToady: then let's go for $filename.path ~~ :x 19:19
TimToady part of the reason for the original formulation was to cache the statbuffer for the filename
moritz that fits very well with my plans so far
japhb Do you object to $filename.path ~~ :x and $filename.open ~~ :x doing the string and fd tasks?
TimToady I think that's clearer than what we have
19:21 SamuraiJack_ left 19:24 GlitchMr left
TimToady in retrospect .io is too confusing to be the default way to do things 19:27
japhb I agree.
19:27 vmspb left
TimToady though we want to scrap .open we could have .io, .input, and .output to go with pipe-io, pipe-input, and pipe-output :) 19:29
or .io, .in, .out to go with pipe-io, pipe-in, pipe-out 19:30
gets rid of all the :r, :w, and :rw flags at least
would still need :a and such though 19:31
japhb But I guess we still need to keep :a, :bin, :chomp, :enc
... right
moritz also for UNIX systems we might want to have .open on Buf too
PerlJam stubbornly clings to :r :w and friends.
TimToady .sysopen :P 19:32
moritz because file names are really binary on linux
TimToady $filename.in.lines ain't so bad
japhb TimToady, we need an escape hatch for magical '-' (both wanting the magic and wanting to explicitly not use the magic) -- how did you want to factor that in this line of thinking? 19:33
TimToady well, '-'.in is obviously the opposite of '-'.out
moritz: yes, encodings of filenames is a complete disaster across OSes 19:34
japhb Sure, but does it open the standard handles, or a path named '-'?
TimToady I'd think '-'.in would just return $*IN and '-'.out would just return $*OUT 19:35
japhb So is this where we then need '-'.sysopen(:r) and '-'.sysopen(:w)? 19:36
oh, and: ew.
.oO( .sys-in, .sys-out, .sys-io )
 19:37
(That 'ew' was to my own suggestion)
TimToady I do like that the -io variants don't commit to whether they return a pair or just something two-ended like a socket
geekosaur $go.sysin DD ...
TimToady isn't there supposed to be some slashes in there somewhere 19:38
well, I never learned JCL
geekosaur s/.../*/ and all you need is an appropriate DD operator. and maybe syntax fudging to get the slashes in there ;} 19:39
dalek p: 0fdbf28 | jnthn++ | src/QAST/ (5 files):
Fix over-sharing of children.
 19:40
geekosaur just used the $ in place of //. just define the twigil! 19:42
19:43 cognominal joined 19:44 birdwindupbird joined
TimToady wonders if people could get used to: my $handle = in $filename; 19:44
19:44 colomon joined
moritz thinks "in" is a bit too generic for that purpose 19:45
japhb Yes, feels like an over-huffmanization
TimToady but open-in is a bit too long for what is the default 19:46
moritz is fine with open
TimToady considered open, open-out, and open-io
dalek c: 995c843 | moritz++ | type-graph.txt:
[type-graph.txt] remove IO::Dir and IO::File
TimToady then wondered if pipe, pipe-out, pipe-io could be forced to wrk 19:47
or maybe open is just a synonym for open-in
sorear mmm JCL
japhb Well, you can make open still be the truly generic thing (pile of options to do all sorts of stuff), and at the same time be the "open for reading" sub/method -- that overloading at least works with my brain just fine.
TimToady so open still defaults to input, but open-in is also there 19:48
japhb Yeah, I can dig that.
sorear "well, O_RDONLY is 0, so it all just works"
TimToady so we can agree on that huffman coding for the listop form, but what about the methods? .in vs .open-in vs .open; .out vs .open-out 19:52
can argue that both ways 19:53
moritz I'm firmly against .in and .out
because there IO isn't the only boundary where somethiing goes in and out
skids
.oO(isn't this what adverbs are for?)
PerlJam skids: I was thinking the same thing.
moritz we also have scopes, modules, containers etc. where something could go in or out 19:54
19:54 sftp left
PerlJam and in could nicely be a predicate too :) 19:54
japhb
.oO( feeds )
TimToady then the same methods as listops 19:55
moritz aye
TimToady .open, .open-in, open-out, .open-io
masak rn: module A { class B {}; our $c = B.new }; say $A::c.^name
p6eval rakudo c1bfbb, niecza v19-15-g051783d: OUTPUT«B␤»
masak rn: module A { class B {}; our $c = B.new }; say A::B 19:56
p6eval rakudo c1bfbb, niecza v19-15-g051783d: OUTPUT«B()␤»
TimToady I do think that it's a good thing to put the direction into the name, because that's a user-centric viewpoint, not an implementor-centric viewpoint
"I am creating this file", not "I am opening a handle"
19:57 sftp joined, benabik joined
masak rn: module A { class B {} }; class A::C {}; say "alive" 19:59
p6eval rakudo c1bfbb, niecza v19-15-g051783d: OUTPUT«alive␤»
masak \o/
moritz erm, why doesn#t that give a redeclaration error? 20:00
s/#/'/
jnthn Which symbol is being redeclared? 20:01
skids moritz: B vs C... but shouldn't it need an augment, or "is also" or whatever it is these days?
oh wait, module, not class... 20:02
sorear moritz: 'A::C' generates a very "soft" declaration of A, which can merge with anything else
PerlJam masak: Do you have a state-of-the-art guide to macros somewhere? 20:03
moritz oh, I thought it was A::B 20:04
dalek p: 14a8815 | jnthn++ | src/QAST/InlinePlaceholder.nqp:
Don't lose namedness of arguments when inlining.
moritz never mind then
dalek kudo/inliner: 5cd3686 | jnthn++ | src/Perl6/Ops.pm:
Mark some Perl 6 ops as inlinable.
 20:06
kudo/inliner: 958a3ea | jnthn++ | src/ (2 files):
We will keep around the QAST tree of inlinable things now, not some string thingy.
kudo/inliner: fdcee65 | jnthn++ | src/Perl6/Optimizer.pm:
Tweak optimizer to look for inlining information properly (though not actually using it yet).
kudo/inliner: 44fef37 | jnthn++ | src/Perl6/Actions.pm:
Update the inlinability analysis and start attaching inlining information to routines again.
masak PerlJam: what, more exactly, are you looking for? 20:13
PerlJam masak: Just to know what's implemented and works today. 20:14
masak PerlJam: oh! hold on, then. 20:15
20:15 bruges joined
masak PerlJam: strangelyconsistent.org/blog/macros...-d1-merged 20:15
PerlJam: strangelyconsistent.org/blog/macros...ak-shaving
PerlJam ah, danke 20:16
dalek c: d510e13 | moritz++ | lib/Whatever.pod:
document Whatever
 20:17
moritz nqp: sub f(:$i) { say($i) }; f(:i) 20:19
p6eval nqp: OUTPUT«1␤»
moritz jnthn: you can write :inlineable(1) as :inlinable in nqp :-)
jnthn moritz: Oh :) 20:22
masak .oO( it's less inhospitable than you think! ) 20:23
jnthn
.oO( it's never hospitalized me )
20:29 cognominal left
dalek kudo/inliner: 60117ea | jnthn++ | src/Perl6/Optimizer.pm:
Implement inlining, now using nice AST splicing. Seems to basically work; handful of test failures that will need investigation. Also generated code not yet awesome.
 20:29
20:31 cognominal joined
dalek osystem: aceab63 | (Carl Mäsak)++ | META.list:
[META.list] added Adventure::Engine
 20:38
masak \o/ 20:39
this should make lue++ happy, too.
jnthn nqp: my $a; foo(:$a) 20:41
p6eval nqp: OUTPUT«Could not find sub foo␤current instr.: '_block1000' pc 55 ((file unknown):137888379) (/tmp/vpXdprrJ7x:1)␤»
moritz jnthn: the syntax is supported, yes 20:42
nqp: sub f(:$x) { say($x) }; my $x = 5; f(:$x)
p6eval nqp: OUTPUT«Assignment ("=") not supported in NQP, use ":=" instead at line 2, near " 5; f(:$x)"␤current instr.: 'nqp;HLL;Grammar;panic' pc 22407 (src/stage2/gen/NQPHLL.pir:8523) (src/stage2/gen/NQPHLL.pm:326)␤»
moritz nqp: sub f(:$x) { say($x) }; my $x := 5; f(:$x)
p6eval nqp: OUTPUT«5␤»
20:43 spider-mario joined
TimToady ooc, how does nqp store in natives without "="? 20:44
jnthn TimToady: It doesn't have a container model, so everything is really "shove X in memory location Y" 20:45
TimToady: Rakudo actually goes to some effort to hide the fact that = assigning to a native means there isn't a scalar container. 20:46
Under the hood, my int $x = 5; is compiled down to something more like my $x := 5; is.
20:47 cognominal left
jnthn The use of "=" is to keep the user comfy, not the compiler :) 20:47
20:47 cognominal joined
TimToady I guess that's an area where nqp isn't really a subset of p6 20:48
jnthn Yeah, that had bothered me a bit
TimToady since := in p6 is always indirect
jnthn otoh, the analysis to make = work for natives and only natives wouldn't be terribly hard.
TimToady I think of := as a = to a (native) pointer locati 20:50
*tion
jnthn Right
Which it basically is
dalek osystem: 93d6026 | (Carl Mäsak)++ | META.list:
[META.list] added Crypt::Game
 20:51
jnthn But that really means "to the body of the object itself", which is what you're doing with native types too
TimToady my hope is, of course, that we can keep the p6 semantics clean enough in that regard to make it a drop-in replacement for nqp someday :) 20:52
of course, I want everything else too :)
masak .oO( we're so fond of Perl 6 that we want even Not Quite Perl 6 to be Perl 6! )
jnthn Well, a worthy initial goal is that NQP code compiles with the Perl 6 compiler. 20:54
TimToady and I'm quite willing to throw a few pragmas at it when I want the Very Fast subset :)
jnthn Yeah, that'll be interesting when I get the NQP optimizer in place.
Oh argh. The Rakudo optimizer has just, er, thrown up some fun too...
It inlined a routine in wrap.t.
sorear jnthn: you're making me want to get a Rakudo CLA filed and abandon Niecza
jnthn And...yeah, you can guess... :) 20:55
TimToady sorear: otoh, rakudo could use a better parser :)
jnthn sorear: Which bit of what I said, ooc? :) 20:56
TimToady: What is it that you really want to see Rakudo do here? 20:57
20:57 sergot left
jnthn TimToady: Is it Rakudo not having STD's error model that's the main pain point? 20:57
TimToady not say "Confused" when niecza doesn't
sjn just read gabor's post 20:58
TimToady but also do a bit more in terms of grammar derivation, like support quote languages better
jnthn *nod*
Well, I did implement mixins in NQP like a week ago :)
sjn <- thinking about "what's a good idomatic way to name junction variables?" 20:59
lue hello world o/ 21:01
sjn first thought: my $some_weekday = any <Mon Tue Wed Thu Fri Sat Sun>; 21:02
lue goes to get Adventure::Engine
sjn or perhaps $a_weekday
masak lue! \o/
21:03 skids left
japhb masak, why Crypt::Game instead of Game::Crypt? The former seems namespace-reversed 21:03
TimToady n: my \any_weekday = any <Mon Tue Wed Thu Fri Sat Sun>; say 'Fri' ~~ any_weekday;
p6eval niecza v19-15-g051783d: OUTPUT«True␤»
lue (installing it with panda seems like a good idea...)
pmichaud yes, I wonder when/if nqp and Perl 6 / rakudo will converge someday. But at the moment having the extra layer of abstraction still seems very useful. 21:04
masak japhb: I kinda thought of Crypt as a top-level namespace for events, exceptions, and the game.
japhb: did the same with Hanoi.
japhb Ah
pmichaud I totally understand about Rakudo's grammar not being as sophisticated at error handling as it ought to be; patches welcome there.
we just have to be careful to not make the parser too slow in the process.
japhb masak, but when put into the ecosystem, it feels odd
pmichaud also, some things are now more possible with ltm that we couldn't really handle well before. 21:05
sorear jnthn: you've got a working fully integrated parser. I don't, and I'm losing hope that I ever will
21:05 kaare_ left
japhb pmichaud, I'd settle for just having Rakudo able to parse any NQP correctly. It bothers me when I can't make the same code work under both. 21:05
sjn TimToady: what did you just do there? 21:06
pmichaud japhb: what parts aren't parsing, ooc?
jnthn sorear: In the nobootstrap?
sorear jnthn: correct
masak japhb: because of "namespace squatting" issues?
japhb Actually, come to think of it, the last time I ran over it may not have been a parse error, but rather argument semantics for a builtin, lemme check ...
masak japhb: I don't think Perl 6 is mature enough for us to have to worry about that. I could rename it to Game::Crypt, but I'd rather do it for a deeper reason than "it feels odd". 21:07
sjn doesn't know the my bareword = any <foo bar>; syntax
sorear jnthn: rakudo is rapidly catching up in the places where I once had niecza working better (inlining!) and has always been ahead in other places
japhb masak: yes, because Crypt:: makes me think Cryptography
masak sjn: it's fairly new.
pmichaud ...and nyi in rakudo, iiuc 21:08
sjn what is it?
masak japhb: that's been the trouble with that name since it got it. :)
japhb Heh
masak sjn: assign-once variables.
japhb masak, but Game::Crypt gives it enough context to be clear. ;-)
sorear jnthn: it might just be the failed branch malaise, but I'm kinda feeling now like niecza has no future.
sjn masak: so... it's a kind of.. function? 21:09
japhb pmichaud, ah yes, it was bless.
sjn because "assign-once variable" really makes no sense
masak sjn: more like a kind of constant.
pmichaud I think of my \bareword = as being more of a way to have a sigiless noun.
*sigil-less noun 21:10
jnthn sorear: There's things Niecza has managed much better at than Rakudo too, as TimToady++ pointed out. If you'd find working on Rakudo more -Ofun, though, I'm all for that. :)
masak japhb: you're right.
21:10 birdwindupbird left
masak japhb: I'll see about renaming it. tomorrow. 21:10
sjn ok, so it's a sigil-less constant
masak rn: constant foo = 42; say foo
japhb masak, cool beans, thanks.
p6eval rakudo c1bfbb, niecza v19-15-g051783d: OUTPUT«42␤»
pmichaud japhb: yeah, nqp doesn't necessarily implement the full Perl 6 runtime.
and .bless is part of that runtime.
masak constants can be sigil-less, too :)
sorear japhb: generic crypto, ore
sjn so, what's the difference?
sorear japhb: generic crypto, or things that resemble crypt(3)?
japhb pmichaud, right, but that one felt like an arbitrary difference. 21:11
masak sjn: the difference between \foo and constant foo is that the former is assigned at runtime.
sjn ok
japhb Does NQP even recognize * at all? Meaning, can it be made to recognize bless(*,...) and just ignore that leading * silently? 21:12
Oh, another strange incompatibility: whether MAIN's args include the program name 21:13
pmichaud that's probably not an intentional incompatibility. I didn't even remember that NQP has a MAIN
jnthn It's decidedly not intentional. 21:15
japhb sorear, Generic crypto, for me. I don't even think about the 'o' because Perl 5 history has led me to think Crypto:: is stranger than Crypt:: 21:16
pmichaud ...but getting NQP's MAIN to work like Perl 6's MAIN is also likely to be... problemeatic.
japhb But I know that's just habit.
pmichaud I'm even wondering if NQP should have a MAIN (or needs one)
jnthn NQP's MAIN just really gets @*ARG passed, or thereabouts.
pmichaud: I found it helpful to have it when porting NQP to NQP.
japhb pmichaud, it doesn't have to work the same, just not be different in a way that I can't correct for easily.
pmichaud japhb: sure, but that starts to sound very much like a slippery-slope sort of thing 21:17
japhb When dealing with the MAIN args incompatibility, must first thought was "Is there any way to know that I'm in NQP or Rakudo without making the other one die?"
pmichaud seems like we could get $*VM to provide that. 21:18
japhb When all you want is just positional args, they are effectively the same.
pmichaud and I suspect we could get NQP's MAIN to work like Rakudo's MAIN. But more to the point, if you want something to work the same in NQP and Rakudo in that case, just don't use MAIN at all :-)
i.e., the question becomes "why use MAIN there?" 21:19
21:19 cognominal left
japhb So ... what would be the common way to get identical CLI args in both? 21:19
pmichaud I would think looking at @*ARGS ought to work in both. 21:20
(perhaps it doesn't, but it ought.)
japhb nqp: say @*ARGS;
p6eval nqp: OUTPUT«Confused at line 2, near "say @*ARGS"␤current instr.: 'nqp;HLL;Grammar;panic' pc 22407 (src/stage2/gen/NQPHLL.pir:8523) (src/stage2/gen/NQPHLL.pm:326)␤»
japhb nqp: say(@*ARGS);
p6eval nqp: OUTPUT«Contextual @*ARGS not found␤current instr.: '_block1000' pc 43 ((file unknown):42) (/tmp/psRw6cmqQM:1)␤»
pmichaud *that's* what nqp should likely be providing instead of MAIN. 21:21
japhb Fine by me. I just wanted *some* common way to get args in a way that didn't get off-by-one with one of them
21:21 moukeddar joined
pmichaud nqp likely needs to catch up with recent changes to rakudo / perl 6 there. 21:22
21:22 moukeddar left
pmichaud there have been some changes made to HLL::Compiler for Rakudo's benefit; I'm not sure that nQP was equally updated. 21:22
anyway, I have to run some errands :-/ bbiaw 21:23
21:26 atrodo left
dalek p: d8f20ed | jnthn++ | src/QAST/Compiler.nqp:
Provide a better, more uniform way to handle Want and coercion (nothing is switched to using it yet, though).
 21:26
p: 6485fa4 | jnthn++ | src/QAST/Compiler.nqp:
Switch handling of attribute bind values to the new Want-aware coercion scheme.
jnthn Hmm...seems none of my YAPC::EU talks got accepted this year :(
japhb That seems very odd. 21:27
jnthn Guess I may as well still go.
21:28 cognominal joined
jnthn Though not being a speaker makes it decidedly less attractive. It's not like crouched over a laptop is a comfortable way to hack. 21:28
*being
(For me, that is...)
masak jnthn: :( 21:29
dalek kudo/inliner: 0c46ef0 | jnthn++ | src/Perl6/Optimizer.pm:
D'oh, left behind debugging code.
 21:30
21:31 bbkr joined
masak I had completely forgotten this, but my blogging site just went down during the night. 21:32
21:32 cognominal_ joined
masak I guess I'll publish the blog post in a gist for the time being, and then actually publish it tomorrow morning. 21:32
bbkr hmm, crypt installed by panda says "Missing block at /Users/bbkr/.perl6/bin/crypt:326" 21:33
21:33 cognominal left
masak bbkr: what version of Rakudo is that? 21:34
bbkr Star 2012.07
masak huh.
This is perl6 version 2012.07-208-gb87659a built on parrot 4.5.0 revision RELEASE_4_5_0
and it runs fine here. 21:35
and I don't see anything odd on line 325.
er, 326.
anyway, here we go: "lol", I "blogged"! "\o/" gist.github.com/3220769
bbkr i just cloned repo and got the same result when trying to validate with -c flag.
masak bbkr: game starts normally here. no parse error. 21:36
bbkr have you tried on star release?
lue how do I get rakudo to install into ~/.perl6 so I can make panda work?
masak bbkr: nope.
tadzik lue: how are those two related?
bbkr lue: you don't have to. do you use Star or only compiler? 21:37
lue tadzik: because make install (no sudo) just installs stuff in ~/rakudo/install, which I don't like (in fact, I have no ~/.perl6 right now)
tadzik lue: why would you need rakudo in ~/.perl6?
panda installs stuff to ~/.perl6
~/.perl6 is automatically in @*INC, so you don't need to care about it 21:38
lue Hm, I appear not to be thinking correctly. If only I could sudo make install on feather...
tadzik lue: 1) install rakudo 2) bootstrap panda 3) everything works 21:39
you don't need to care about where stuff gets installed
21:40 cognominal__ joined
lue number 3 doesn't wfm. Hang on, lemme finish compiling rakudo and then I'll try panda again. 21:40
21:41 cognominal_ left
lue But just running "perl6 bootstrap.pl" throws an error, likely because feather's rakudo is very old. So I'd have to do some magic with $PATH to get it to work. 21:43
21:44 cognominal__ left 21:45 cognominal joined
lue masak++ 21:45
masak yes, feather's rakudo is very old.
you could install a newer one in your homedir, perhaps?
lue that's what I wanted to do :) right now it just goes in ~/rakudo/install, and I thought (for whatever reason) it was supposed to be in ~/.perl6 for a homedir install 21:46
tadzik ~/rakudo/install is fine 21:47
panda doesn't care where your rakudo is, as long as 'perl6' is in $PATH
tadzik sleeps &
lue I should probably put ~/rakudo/install/bin first in $PATH's list, right? 21:48
tadzik I think so
masak sounds about right.
lue
.oO(Is it bad that I'm so used to being able to install stuff in /usr/local that when I'm on feather and can't I've lost the ability to do things correctly? :) )
 21:49
masak consider it an opportunity to learn how to do without sudo ;) 21:52
bbkr masak: this is weird, syntax error about missing block disappeared when i commented line "when X::Adventure …" in CATCH, but at the same time Adventure::Engine syntax that provides this class is OK 21:54
masak bbkr: that does sound weird. 21:55
22:04 PacoAir left 22:09 spider-mario left
bbkr star: CATCH { when X::Y {} } 22:10
p6eval star 2012.07: OUTPUT«===SORRY!===␤Missing block␤at /tmp/4LoyksdaLg:1␤»
bbkr hm, does not make any sense 22:11
r: CATCH { when X::Y {} }
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤Missing block␤at /tmp/NTZDH8L6yj:1␤»
lue r: class X::Y{ }; CATCH when X::Y { } } 22:12
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤Missing block␤at /tmp/AG1zwwTQ08:1␤»
lue r: class X::Y{ }; CATCH { when X::Y { } }
p6eval rakudo c1bfbb: ( no output )
bbkr i know X::Y is not defined, but in this case error message is way LTA
lue agreed 22:13
star: class X::Y{ }; CATCH { when X::Y { } }
p6eval star 2012.07: ( no output )
bbkr reports 22:14
22:19 fridim_ left
bbkr that partially explains masak's crypt game issue - somehow it cannot see X::Adventure type in CATCH scope 22:19
22:20 benabik left 22:21 benabik joined
masak bbkr++ 22:23
lue can't believe he didn't set up panda sooner! 22:24
masak :) 22:25
lue Now I'm off to try using Adventure::Engine! 22:28
masak \o/ 22:30
bbkr digging further problem with missing block message disappeared (X::Adventure became visible inside CATCH scope) after I removed Hanoi/Game.pir. 22:31
can you compile this file to pir and check if crypt works? 22:32
maybe we have similiar issue that was once spotten on non-deterministic URI compilation 22:33
masak ugh. tomorrow. :/
bbkr spotted*
masak: no rush
masak I've shut down my brain for today, sorry.
22:34 skids joined
lue (ooh, I was about to ask about putting a subroutine on an exit, but then I spot on_try_exit in Crypt's BUILD!) 22:35
22:35 bluescreen10 left
masak lue: yes! that's why reading BUILD is instructive ;) 22:39
lue Since I've suddenly decided my real game (just writing a small demo to get used to A::E) takes place in the sky, there are a lot of directions where ***You have died***. 22:41
And in those directions, I'd like to ask "Are you sure?" first, so a sub is necessary.
masak well, hooks allow further events to be emitted, but not user interaction. 22:43
the .on_try_exit hook is the only one that conditionally cancels its parent action as well (with a Bool). 22:44
pmichaud jnthn: my talk wasn't accepted either, it would seem.
masak :(
that sucks.
pmichaud looks like two of Carl's talks were accepted... and that's it. 22:45
act.yapc.eu/ye2012/talks/tag/perl6
22:45 Coleoid joined, Chillance left 22:46 Chillance joined
pmichaud masak: I guess you're the voice of Perl 6 for YAPC::EU :-) :-) 22:46
masak oh crap :P
benabik Apparently the voice of Perl 6 is a potty mouth. ;-) 22:47
jnthn pmichaud: Aww. :( 22:48
jnthn always enjoys the Pm talks :) 22:49
masak too
and the jnthn talks as well.
jnthn masak: huh, you get to see jnthn talks more than most :P
Tene masak: I guess you'll just have to impersonate both jnthn and pmichaud during your talk
masak the results don't feel commensurate with the awesomenesses of the speakers, is all I'm saying.
22:50 lucas_ joined
masak Tene: somehow I'm not sure that would help anything, really... 22:50
Tene I think i've seen recorded jnthn and pm talks, but I don't think I've ever seen a masak talk. 22:51
masak there's a few of them out there. none of them are very notable.
jnthn Oh come on. The module cheese one... :P 22:52
Though dunno if that was videod. :)
masak don't think so. 22:53
well, I remember a camera from that one.
but I've never seen any video come out of it.
jnthn: it was actually still a Rakudo cheese at that point. the module cheese evolved out from that.
but yeah, that was a terribly successful talk. I have bugger all idea why, because it wasn't very good :P 22:54
www.perlworkshop.no/npw2009/talk/1734
pmichaud masak: you're way in the minority opinion there 22:55
I thought that talk was the best I saw at the workshop.
masak o.O
wow. thanks.
pmichaud (of course, I didn't see my talk, so I can't compare. I still thought yours was better than mine, but I'm a biased observer :)
masak anyway, tadzik++ has realized most of the ecosystem dreams of that talk. 22:56
plus Emmentaler, which I never put in a talk.
masak yawns 22:57
good night, #perl6. it was nice talking to you. you're awesome.
sorear goodbye masak *cries* 23:00
pmichaud oh, wait, looks like Moritz' talk was accepted also.
both of his talks! So that makes four.
act.yapc.eu/ye2012/talk/4055 23:01
act.yapc.eu/ye2012/talk/4202
sorear agh, I really need to get travel plans sorted ASAP
sorear makes panicky gestures
pmichaud okay, I feel better now, although I was really looking forward to jnthn++'s talk on roles.
that just seems sooooooo timely and useful for this conference 23:02
jnthn pmichaud: There is a roles talk though: act.yapc.eu/ye2012/talk/4150 23:03
pmichaud yeah, true.
jnthn pmichaud: I guess maybe my one may have clashed a bit wiht that. But I did submit another one too... :)
pmichaud so, it looks like they've accepted around 20 new talks since the "no perl 6 talks at yapc::eu" posting, and 4 of those are Perl 6 related. That's not too bad, considering where things started. 23:06
jnthn aye
pmichaud anyway, I guess I'll have to come up with a good lightning talk, although those slots are filling up also.
and I'm still fairly disappointed that none of my yapc::na talks have made it to youtube yet :-| 23:07
(at least, not officially)
looks like they got up to "M" (Mark Straman) and stopped.
(youtube.com/yapcna) 23:08
sorear what's the official explanation? 23:10
were they unable to get music distribution licenses for the other talks?
23:10 erkan joined, erkan left, erkan joined 23:15 plobsing joined
pmichaud there has been no explanation yet. 23:18
I'm certain they had license for distributing the other talks; reportedly it was one of the criteria for acceptance
but just looking through the talk page, it looks as though they've been posting talks in alphabetical order by first name.
thus they've only gotten as far as "Mark", and "Patrick" collates after that. 23:19
afk, dinner 23:20
sorear and they just stopped? how long has it been since they last posted a new talk? 23:21
pmichaud ten days. last post was July 21. 23:22
sorear wonders if they will ever get to pm 23:27
did they ever finish uploading ::NA 2011?
23:31 benabik left 23:37 benabik joined 23:40 thelazydeveloper left
skids r: sub a () returns Int { fail("foo") }; a().WHAT.say; # perlcabal.org/syn/S02.html#Undefined_types would suggest a Failure here, though not necessarily the "foo" Failure itself. 23:43
p6eval rakudo c1bfbb: OUTPUT«Type check failed for return value; expected 'Int' but got 'Failure'␤ in sub a at /tmp/eSqEPViPUG:1␤ in block at /tmp/eSqEPViPUG:1␤␤»
[Coke] does panda's bootstrap care which perl6 is in your path, or does it use the perl6 it was invoked with? 23:45
skids: huh. I wonder how you're you're supposed to fail out of a typed sub, ten. 23:48
*then.
23:49 plobsing left
benabik Either Failure should be ignored by the type checker, or the act of checking its type should cause it to throw. 23:53
23:58 cognominal_ joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!