The topic for #perl6 is: pugscode.org/ planetsix.perl.org/ | nopaste: sial.org/pbot/perl6 | pugs: [~] <m oo se> (or rakudo:, kp6:, elf: etc.) (or perl6: for all) | irclog: irc.pugscode.org/
Set by TimToady on 25 November 2008.
rakudo_svn r33212 | jonathan++ | [rakudo] Add another spectest. 00:00
jnthn Hmm. I guess you should be able to do things like "&my_sub does Foo;". 00:01
TimToady and rakudo will presumably allow us to attach info to methods better than p5 does
not sure if we can say "call this method, but not really, just return its property..." 00:02
jnthn Should do. We might have a little fun with constant vs non-constant PMCs along the way. :-|
TimToady in p5 I guess that's a ->can()
jnthn Oh yes, hmm. How to get a sub props... :-)
&foo.SUB.prop_name
TimToady well, can certainly fake that up in rakudo like we do for STD currently, if need be 00:03
jnthn Sure.
Shouldn't be too hard.
TimToady &foo.prop_name isn't the problem, it's finding the right &foo to call it on :) 00:04
jnthn Ahhh...yes.
OK.
TimToady my token's lexer is based on some other set of tokens, many of which might come from other grammars, treated as virtual methods 00:05
that's the thing that STD is very careful to do correctly 00:06
jnthn It's essentially just inheritance and composition, though?
With the overriding etc that comes with it?
TimToady yes, but the composition is hairy since it has to work the patterns in parallel to preserve LTM 00:07
RE::sequence::longest has to work rather hard at that
since it has to do cross product of what might match before and what might match after 00:08
jnthn Ouch.
I'm glad you worked out how to do this... :-) 00:09
TimToady it would be very easy to define a lexer with millions of tokens, so the grammar writer needs to be a bit careful
at the moment I've got STD limited to expecting 8299 tokens where a term is expected 00:10
but meta-operators tend to multiple that
jnthn That's why we're not letting those be nested recursively without predeclaring, right?
TimToady not letting them nest at all, at the moment 00:11
if you predeclare, it's just an ordinary operator
as far as the lexer is concerned
assuming you declare it infix or some such 00:12
jnthn So you can do [*] and >>*<< but not [>>*<<]?
If that last one even makes sense...
TimToady you can nest different types
jnthn Ah, OK. But not many of the same.
TimToady so [>>*<<] is allowed
00:12 ZuLuuuuuu joined
TimToady but not >>X>>*<<X<< 00:12
and, in fact, STD accepts [>>*<<] 00:13
jnthn I think I'm glad of that from a language design point of view too. :-)
TimToady well, LTM basically can't handle left recursion, is all
not if you're going to enumerate all possible token patterns 00:14
jnthn Sure. But I'm not sure there's so many people who could look at >>X>>*<<X<< and be like "oh yeah, it does..."
:-)
TimToady agree, it's probably a feature
jnthn So I don't think it's a big loss not being able to nest.
Thanks for all of the answers/explanations. I'm sure I'll be back for more once I start implementing... 00:16
TimToady it's enough that it allows [\>>minmax<<]<<
'k, I may be spottily connect over the next two weeks, given I'm going around the world
*connected 00:17
jnthn That sounds like the tiring kind of fun I like. :-)
Conferences, vacation, boht?
*both
TimToady Australia, then Amsterdam
jnthn That's a lot of flying.
TimToady conferences, bit of sightseeing in Australia
00:17 Exodist left
TimToady I'm in denial 00:17
and I'm trying to figure out if I lose a day or gain a day in my life that way... 00:18
00:18 vixey left
jnthn I would say just think, Australia is probably nice and warm right now. But you're not living in a cold place. :-) 00:19
TimToady it'll pretty close to the same
but probably a bit wetter there
jnthn Oh, I can never figure out timezone stuff too well.
When I travelled across Russia, all the train tickets have Moscow time on them. So I kept my watch on that the whole trip in an effort not to mess things up.
It was successful in that sense, but it menat explaining a few things to confused Russians who asked me the time. :-) 00:20
TimToady must be strange for the sun to rise at 1:00 am... 00:21
well, far enough north, it does that anyway :)
jnthn Ah, I was pretty far south at that point in the journey (Vladivostok).
TimToady but it might explain the cavalier attitude toward time zones
since no amount of daylight savings time can make up for that much swing 00:22
jnthn China is a curiosity, since they have the same time zone across the whole country.
TimToady China doesn't have that excuse though
jnthn All the way west in Tibet, it feels a bit odd. 00:23
TimToady "We don't care. We don't have to. We're the phone company."
takeoff on an old phone company ad we used to hear 00:24
jnthn Sounds like phone companies.
TimToady well, the original was "We care. We have to. We're the phone company."
00:24 pcbuetow joined
TimToady anyway, there's a kind of monopoly in China, you might say 00:25
jnthn If they're anything like most phone companies I've dealt with, it's a rather uncaring definition of caring...
TimToady hence the parody
jnthn Yes.
When the advertising you think will be good, backfires. :-) 00:26
TimToady it doesn't matter how hard you think about how to name your kids; junior highers will find a way to make fun of anything...
jnthn In the UK, people seemed to take to giving the train companies alternative names.
"Worst Late Western" 00:27
jnthn doesn't miss this aspect of UK life at all
TimToady yeah, we used to fly on Hughes Air Worst
I'm flying to Oz on Untied
jnthn I flew with them to the Parrot Summit. 00:28
00:28 ab5tract joined
jnthn I've had better, I've had worse. 00:28
TimToady they're a bit on the sardine side
00:29 km2 left
jnthn Is it direct flight down to Oz? 00:29
TimToady I believe so
jnthn Ah, not so bad. 00:30
Oz from here is a long, long way.
Can't do it direct. At least, not without a touchdown for refueling somewhere.
TimToady 14h 27m 00:32
jnthn Longer than any I've done yet.
TimToady I guess it's not that much worse than the 11h to Japan
jnthn I figure once it gets beyond 10 hours, it's just looooong.
TimToady I vote for suborbital. 00:34
jnthn Hey, supersonic wouldn't be so bad... 00:35
TimToady well, hypersonic maybe 00:36
jnthn That'd be awesome, but if we didn't make supersonic affordable...
TimToady slow enough to keep a bit of gravity
jnthn Or economical, anyways...
On Japan, I'm pondering trying to make YAPC::Asia next year, provided there is one.
TimToady but if the can fly it back and forth 5 times in the same time as the old plane flies once... 00:37
*they
jnthn Sure, but I think the problem was more maintenance costs/wear and tear etc...
jnthn has never been to Japan
00:37 pbuetow left
TimToady I guess I've been there five times now 00:37
00:38 araujo joined
jnthn Worth it? 00:38
I guess if you've been five times it can't be so bad... :-)
TimToady I've enjoyed it immensely, but then I'm a crazy japanophile... 00:40
jnthn My biggest fear, is that I don't eat fish/sea food.
Which is pretty much everywhre.
But I guess it's possible to get by without eating that.
TimToady well, neither do I, because I'm allergic to it
jnthn Ah, OK.
TimToady it's possible to get by 00:41
jnthn I'm not allergic, I've just never been able to find it likable.
I've tried a few times, in various places, and just can't enjoy it at all.
TimToady alas, I like it...
got any other allergies you'd like to trade?
jnthn I haven't got much in the way of allergies. Just various pigmentation problems from different sides of the family. 00:44
My eyesight problem is because I lack sufficient pigment in my retinas, meaning a bunch of the rod and cones are defective, which basically means I see in low resolution. Though in full color. 00:45
(Thus why the trains always running late in the UK drove me nuts. I didn't have much choice but to use them...)
TimToady huh. everything wrong with my eyes has been in front, and hence fixable, fortunately 00:46
jnthn Yeah. This one is unfixable.
However, thankfully, it's stable.
00:46 japhb__ left
TimToady well, I probably ought to pretend to work... 00:46
jnthn I wear glasses for a minor lens issue (they don't help with the retina one). But more so for the color-change lenses...
Yeah, I guess I ought to pretend to sleep. 00:47
jnthn hasn't slept normal hours since the Parrot Summit
TimToady night
jnthn night
00:51 japhb joined 00:52 DemoFreak left
wayland76 Btw, re: warmth in Australia, it's not cold any more, but it's not that warm either 01:06
...although I'm near Melbourne, and people talk about "Melbourne weather" :) 01:07
Re: junior higher's mocking names, I know a family who had division of labour in choosing names for their children; the mother would suggest names, and the father tried his best to mock them 01:10
When they came up with something less mockable, they used that
01:14 ZuLuuuuuu left 01:20 smallfish joined 01:22 smg left 01:32 PZt left 01:34 kst` joined 01:47 mbtrimpe joined, mtrimpe left 01:50 mbtrimpe left 01:51 mtrimpe joined, FurnaceBoy joined 01:54 ab5tract left
pugs_svn r23093 | lwall++ | [t/spec] my() is a function call 02:02
02:14 ab5tract joined 02:16 PZt joined
pmichaud rakudo is treating ^@a as being prefix:<^> applied to a List. 02:25
rakudo: say ^(3,3); 02:26
p6eval rakudo 33212: OUTPUT[012012ā¤]
pmichaud (it's doing it wrongly, but that's what it's dong.) 02:27
*doing
so, how should we distinguish ^@a from ^(3,3) ?
jnthn @a is an Array, (3,3) is a List?
lambdabot Maybe you meant: activity activity-full admin all-dicts arr ask . ? @ v
pmichaud S03 doesn't make any explicit mention of ^@a 02:29
jnthn Ah.
I'm not sure what a sensible thing for that to do is anyway.
pmichaud it does discuss 0..^@a, and it says that ^$n is equivalent to 0..^$n 02:30
jnthn Yeah
But ..^ is a separate operator
pmichaud but it also says that ^(list) produces a multidimensional set of subscripts.
so there's some ambiguity about ^@array
jnthn Right
Because I'm sure sure that we can (by the spec) say that Array is a subclass of List 02:31
pmichaud I'm pretty certain that the ^@array form will end up meaning 0..^@array -- but I'm curious what the disambiguation will be.
jnthn If it were, we could go by implication...
02:31 mtrimpe left
jnthn hmm 02:31
I don't know.
I did have one question for you about something else... 02:32
On subid
I guess every PAST::Block will get a subid
pmichaud yes.
jnthn But will there be a way to say in the PAST tree "I want to get hold of that block by its subid"? 02:33
But without ending up duplicately having that block in the tree.
pmichaud yes, there will be a way to do it.
I don't know what it will be yet.
certainly we'll allow HLL compilers to assign subids that override the default generated one. 02:34
jnthn (Reason: subset types need to have their condition block live in the lexical scope they were declared in, but we want to create them earlier than that to set up the type in the namespace)
So I'd want to put the block in the right lexical scope (and thus in the right place in the PAST tree, but be able to look it up in init code.
So I can create the subtype referencing it. 02:35
pmichaud oh, the loadinit code will still have a 'block' register variable.
that will be initalized with the block's PMC
02:35 jferrero left
jnthn Oh, I can do it that way too I guess... 02:35
Ah, aye. As a loadinit on the block that is the condition.
Cunning.
Why don't I think of this stuff.
:-
)
pmichaud yes, the point is that these days are largely taken care of for you 02:36
s/these days/these types of recordkeeping/
(can't type .. a little tired tonight.)
jnthn I wish these days were taken care of for me. Then I could spend them relaxing.
It's tonight where you are too?
Oops.
pmichaud 20:36 localtime
jnthn I've slept at the wrong times since the Parrot summit. :-)
pmichaud I'm very familiar with that :-)
02:37 FurnaceBoy left
jnthn Partly because I don't care about sleeping at the "right times" 02:38
pmichaud me neither.
jnthn Only day I need to be up earlyish is Sunday for church. That's about it.
I'm Rakudo daying tomorrow. 02:39
pmichaud okay, excellent. I should be around much of the day.
jnthn Plan to mostly look at RT.
pmichaud However, kids are off from school, so I may get distracted.
yes, I'm wanting to look at RT as well. :-)
jnthn There's at least one patch from Chris Dolan I want to reivew.
pmichaud I'll probably do a few PCT things here and there, and maybe work on the HLLCompiler refactor so that we can start moving to .hll
jnthn Oh, yes, please.
pmichaud either that or I'll work on p6 builtins.
02:40 Limbic_Region left
jnthn As in, having a Perl 6 prelude? 02:40
pmichaud yes.
jnthn OK
One thing I really want to do
Refactor variable declarator.
Then I can do has $.answer = 42;
pmichaud don't we need to fix signature first for that? 02:41
jnthn Not really
I mean, yes, to get it *really* right we do.
But I'd rather do it a step at a time.
Any objections? 02:42
pmichaud depends on what you mean by "a step at a time"
jnthn Until I look at it, I won't know exactly.
Probably though, just fixing up the grammar to parse the initialization as in STD.pm 02:43
pmichaud ...that's kinda my point.
jnthn And then doing that initialization in variable_declarator
OK, so you'll maybe object after I've worked it out? ;-)
pmichaud I'm trying to find where the initialization occurs now in STD.pm 02:44
jnthn I *can* leave this on your plate if you prefer. But I am keen to get "has $.x = 42" style tings working.
oh my, don't say it's changed...
pmichaud yes, it has.
02:45 slavik joined
jnthn argh 02:45
What on earth is :dba 02:46
pmichaud "doing business as"
it says how '(' ~ ')' <EXPR> should report back if the match fails
jnthn Ah, OK 02:47
pmichaud anyway, I'm not sure if my $a = 3; parses as assignment or as something else.
jnthn Looking at this, now I'm really not sure how the initialization is parsed.
slavik who do I bother about not being able to do %hash = 1..5 | 6..10;
pmichaud slavik: what are you wanting that to do? 02:48
that looks like a junction of two ranges.
jnthn Hopefully give an error.
slavik rakudo: say (1..5 | 6..10).perl
p6eval rakudo 33212: OUTPUT[1..any(5, 6)..10ā¤]
slavik rakudo: say (1..5 Z 6..10).perl
p6eval rakudo 33212: OUTPUT[elements() not implemented in class 'Range'ā¤current instr.: 'infix:Z' pc 3915 (src/gen_builtins.pir:2519)ā¤]
slavik right, that
pmichaud oh, you want infix:<Z>. Send email to rakudobug.
[email@hidden.address]
slavik pmichaud: zip() doesn't work at all I think 02:49
pmichaud one of Jonathan, myself, or our other incredible team of coders will look at it.
slavik rakudo: say zip(1..5, 6..10)
p6eval rakudo 33212: OUTPUT[Could not find non-existent sub zipā¤current instr.: '_block11' pc 63 (EVAL_15:23)ā¤]
pmichaud We might not be able to do much with it because we don't have slices implemented yet.
or captures.
slavik pmichaud: I have a grammar that produces two arrays, one of keys and other of values ... so I want to stick them into a hash
jnthn slavik: Infix Z works, I believe. 02:50
pmichaud rakudo: say (3,3) Z (1,2);
p6eval rakudo 33212: OUTPUT[3132ā¤]
pmichaud that doesn't look right.
jnthn No?
pmichaud oh wait, yes it does. 02:51
jnthn Why not?
slavik hmm
pmichaud rakudo: say ((3,3) Z (1,2)).perl;
p6eval rakudo 33212: OUTPUT[[3, 1, 3, 2]ā¤]
slavik maybe I am doing something wrong
pmichaud rakudo: my %h = <1 2 3 4> Z <a b c d>; say %h.perl;
p6eval rakudo 33212: OUTPUT[{"1" => "a", "2" => "b", "3" => "c", "4" => "d"}ā¤]
pasteling "slavik" at 67.100.227.140 pasted "URL grammar" (37 lines, 1.3K) at sial.org/pbot/33362
pmichaud yummy.
02:52 smg joined 02:53 ihrd joined
pmichaud that's going to be a bit weird, since the components being zipped are Match objects 02:53
I'm not sure how the hash composer is going to deal with those.
slavik it doesn't
I get a {}
how can I coerce them into lists?
@() didn't work
lambdabot Maybe you meant: . ? @ bf ft id pl rc v wn
pmichaud they're already lists (of match objects)
slavik oh ... 02:54
so $? contains a tree of match objects, not scalars or hashes or lists?
pmichaud $/, yes
slavik right, $/
how can I make it a list of scalars though? 02:55
{ %get{$^a} = "$^b" } for $/<resource>[0]<getstr>[0][0] Z $/<resource>[0]<getstr>[0][1];
that works
pmichaud can also do prefix:<~> 02:56
{ %get{~$^a} = ~$^b } for ...
slavik pmichaud: I want strings there :)
pmichaud prefix:<~> makes strings.
slavik oh
right
ihrd perl6: sub it { for 1..5 { $_ }; }; say it(); 02:57
p6eval pugs: No output (you need to produce output to STDOUT)
..elf 23093: OUTPUT[Useless use of a variable in void context at (eval 123) line 4.ā¤ā¤]
..rakudo 33212: OUTPUT[-1ā¤]
jnthn ihrd: Trust you. :-P
ihrd hi
pmichaud rakudo doesn't know return values in for loops yet. 02:58
jnthn Maybe, the loop control exception stuff gets in the way?
ihrd but I understand right, it should return 5?
pmichaud I'd have to go back an look -- iirc, the return value of a for loop is a list of the results of its body
slavik pmichaud: like a map? 02:59
jnthn O RLY?
That's...nuts. And so Perl 6...
ihrd (1,2,3,4,5) in that case?
jnthn OH NOES! It's 4am...I should sleep!
slavik map works in this what ... 03:00
jnthn: lieas, 10PM
jnthn Where's lieas?
slavik bah
lies*
jnthn Oh
No, it feels like 4am too :-)
slavik has timemachine() been implemented yet?
jnthn Only on OSX. ;-) 03:01
pmichaud S04, "Loop statements": The value of a loop statement is the list of values from each iteration.
slavik jnthn: wrong time machine ;) 03:02
very limited ...
jnthn pmichaud: ...is that as ambiguous as I think it is? ;-)
pmichaud jnthn: not really.
jnthn The values could be the values produced...or the values iterated over...
pmichaud 'for' and 'map' are equivalent, essentially.
it's the value returned by each execution of the body
slavik jnthn: values produced, like map
jnthn I suspect that's the useful answer and what was meant, yes.
pmichaud not only that, but .next and .last can specify the value to be returned for the current iteration 03:03
jnthn Perl 6 is nuts. :-)
pmichaud &?BLOCK.next(3); # return a value of 3 for this iteration
thus
03:03 dukeleto_ joined
jnthn Thus why I love working on it. ;-) 03:03
slavik pmichaud: wha ... ? 03:04
pmichaud sub it { for 1..5 { &?BLOCK.next(3); } }; say it(); # "33333\n"
slavik I don't get the .last and .next think
jnthn Nice!
OK, I am now going to sleep.
slavik and if it was ,last?
jnthn Probably will be about midday european time. 03:05
pmichaud sub it { for 1..5 { &?BLOCK.last(3); } }; say it() # "3\n"
jnthn night all
slavik jnthn: night
ihrd good night jnthn 03:06
03:08 ab5tract left
slavik pmichaud: so, how can I coerce the list of matches into a list of scalars? 03:09
rakudo: for 0..7 say $_; 03:15
p6eval rakudo 33212: OUTPUT[Statement not terminated properly at line 1, near "say $_;"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)ā¤]
slavik rakudo: for 0..7 { say $_; }
p6eval rakudo 33212: OUTPUT[0ā¤1ā¤2ā¤3ā¤4ā¤5ā¤6ā¤7ā¤]
ihrd excuse me, but I still in some confusion, if I want return only last value from loop, how can I do that? 03:16
slavik ihrd: like if it generates a sequence based on previous things? 03:17
ihrd slavik: I just do some method in loop, and whant to now last returned value from it 03:18
slavik hmm 03:19
ihrd right now I use workaround, like
03:19 japhb left
slavik @(for 1..10 { $_ })[-1] 03:19
lambdabot Unknown command, try @list
slavik or something
rakudo: say @(for 1..10 { $_ })[-1];
p6eval rakudo 33212: OUTPUT[Statement not terminated properly at line 1, near "@(for 1..1"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)ā¤]
slavik rakudo: say @(for 1..10 { $_; })[-1];
p6eval rakudo 33212: OUTPUT[Statement not terminated properly at line 1, near "@(for 1..1"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)ā¤]
slavik rakudo: say for 1..10 { $_; };
p6eval rakudo 33212: OUTPUT[Statement not terminated properly at line 1, near "{ $_; };"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)ā¤] 03:20
slavik rakudo: say (for 1..10 { $_; });
p6eval rakudo 33212: OUTPUT[-1ā¤]
slavik :(
it doesn't work yet
03:20 jhorwitz left
ihrd perl6: my $r; sub some { $^d ** 2 }; sub it { for 1..5 { my $r =some($_) } }; say $r; 03:21
p6eval rakudo 33212: OUTPUT[ā¤]
..elf 23093: OUTPUT[Global symbol "$d" requires explicit package name at (eval 127) line 4.ā¤ at ./elf_f line 3861ā¤]
..pugs: No output (you need to produce output to STDOUT)
pmichaud do placeholder variables work in subs...?
ihrd perl6: my $r; sub some { $^d ** 2 }; sub it { for 1..5 { my $r =some($_) } }; it(); say $r; 03:22
p6eval rakudo 33212: OUTPUT[ā¤]
..pugs: No output (you need to produce output to STDOUT)
..elf 23093: OUTPUT[Global symbol "$d" requires explicit package name at (eval 129) line 4.ā¤ at ./elf_f line 3861ā¤]
ihrd It worked for my
pmichaud well.... we can't really tell what some() is returning there. 03:23
because the $r we're saying isn't the $r we're assigning.
ihrd perl6: my $r; sub some{ $^d ** 2 }; sub it { for 1..5 { $r =some($_) } }; it(); say $r; 03:24
p6eval pugs: No output (you need to produce output to STDOUT)
..elf 23093: OUTPUT[Global symbol "$d" requires explicit package name at (eval 129) line 4.ā¤ at ./elf_f line 3861ā¤]
..rakudo 33212: OUTPUT[25ā¤]
ihrd yes, excuse me
now right
so this is workaround, how can I do the same in perl6 without $r? 03:25
Can I see somehow the last iteration in loop? 03:27
something like: for 1..5 { do($_); last(1) if .it_is_last_iteration }? 03:28
pmichaud rakudo doesn't have a last() implemented yet. 03:30
we're waiting for a couple of fixes to parrot exception handling.
(and a refactor of control blocks in general)
slavik pmichaud: are exceptions the only way to return errors? 03:31
or can we return undef
pmichaud the common way to signal an error is to use fail() 03:32
that returns a "special undef" that has the exception attached 03:33
if anything tries to use that undef, the exception gets thrown
(where "use the undef" means "do something other than check it for truth or definedness")
slavik ie: sub f { fail(); } if(f == undef) { say "failed" }
or something 03:34
pmichaud if (!f) { say "failed" }
or
if (!f.defined) { say "failed" };
comparing with == undef is often wrong, because there's not just a single value of undef.
slavik so, try catch finally aren't necessary ...
pmichaud only if you want to try/catch exceptions :-) 03:35
you could do:
{ say f(); CATCH { say 'failed'; } }
and if f() returns a Failure object then the CATCH block gets invoked because we attemped to use the returned Failure 03:36
slavik I see
pmichaud (except that CATCH isn't working yet in Rakudo. See above note about needing to implement control blocks :-)
slavik but inside the if statement, it doesn't get used, just looked at, correct?
pmichaud testing for truth or definedness doesn't count as "used", no.
trying to get an int, num, or str does. 03:37
(I'm not sure what else does.)
slavik hmm
does fail work?
pmichaud in Rakudo? Yes.
at least I think it works. Need tests. :-) 03:38
'try' still needs a bit of work.
ihrd I know about last in Rakudo, my question about perl6
excuse me lag :) 03:39
slavik: [-1] do not work in perl6 as in perl5
slavik pmichaud: if I wanted to have pir bindings for C libraries ... what is the major difficulty in this endeavor ?
ihrd: huh? 03:40
ihrd slavik: in you exapmle should be [*-1]
slavik oh
ok
pmichaud oh. Normally we say "Perl 6" when asking about the language. But yes, last(1) would cause the for loop to put a 1 as the last value in its return list.
slavik: I don't know, I'm not too familiar with invoking C from PIR
slavik: I know it can be done and people are doing it, but I haven't done it in years and so I'm not familiar with the details. 03:41
ihrd pimchaud: ok, thank you! but hov can I see that is last iteration?
TimToady @tell jnthn STD currently just parses initialization as assignment, and delegates any change in meaning of = to the semantic analysis phase
lambdabot Consider it noted.
pmichaud TimToady: yay, excellent.
TimToady: that's actually what rakudo does now, so jnthn (and I) are happy to hear this.
03:41 Psyche^ joined
TimToady it just seems like the sane thing to do, though we'll have to be careful with (state $x) += 2 and such 03:42
pmichaud TimToady: did you see my comment about ^@a being treated as prefix:<^> on a List ?
TimToady yes
dunno what to think about that
pmichaud okay.
TimToady I want contradictory things, it seems. not the first time... 03:43
pmichaud I'm wondering if ^(3,3) is all that important or useful as a shortcut. Maybe a different shortcut.
sometimes I think (language) design is just about resolving our contradictory desires. :-)
TimToady ^Ā« maybe
rakudo: say ^<<(3,3) 03:44
p6eval rakudo 33212: OUTPUT[Statement not terminated properly at line 1, near "^<<(3,3)"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)ā¤]
pmichaud @tell jnthn (comment on TimToady's note) yay! So it is assignment, which rakudo already does, and you can start working on has $.x = 3; based on that :-) 03:45
lambdabot Consider it noted.
pmichaud yeah, rakudo doesn't understand the hypers yet.
03:45 wayland76 left
pmichaud silly meta_postfix_prefix_inafix_wontfix operator parsing. 03:46
TimToady there's only 8299 possible termish tokens currently... 03:47
03:48 km2 joined
TimToady oops, 8298 really 03:48
pmichaud "8299 tokens in term from the Wall... 8299 tokens in term... take one down, meta it around, 16598 tokens in term from the Wall..."
TimToady much much better
pmichaud :D
TimToady forgot the first line is the dba
03:52 ihrd left 03:56 wayland76 joined
pmichaud time for sleep or relaxation here -- see you all tomorrow 03:57
03:58 Patterner left, Psyche^ is now known as Patterner
TimToady does that mean you don't relax when you're asleep? 03:59
slavik rofl
so, TimToady ... cups
TimToady Common Unix Printing System 04:00
slavik not that one 04:01
the broken cups ... heh
there should be an easter egg relating to it 04:02
04:03 elmex_ joined 04:17 elmex left, elmex_ is now known as elmex 04:25 stephens left
pugs_svn r23094 | putter++ | [elf_h] Match.pm: Fix string literals with unquoted escapes. 04:25
r23094 | putter++ | [STD_blue] Triage remaining elf source files.
r23095 | hinrik++ | [util/perl6.vim] add support for directives (=use, =config, =encoding) and verbatim regions 04:27
04:47 ab5tract joined 04:48 ab5tract left, alc joined 04:51 revdiablo joined 04:56 wayland76 left 04:57 slavik left 05:04 xinming joined 05:11 xinming_ left 05:45 apeiron left 05:48 smg left 06:11 Bzek joined 06:45 kst` left, kst` joined 06:46 ashizawa joined 06:49 wayland76 joined 06:56 aindilis` left, aindilis` joined 07:05 DemoFreak joined 07:18 smg joined 07:34 Jedai left, Coleoid__ joined, Jedai joined 07:35 japhb joined 07:37 pnu left 07:38 pnu joined 07:39 Coleoid left, Coleoid__ is now known as Coleoid
moritz_ rakudo: say (^(2,3)).perl 07:48
p6eval rakudo 33212: OUTPUT[[0, 1, 0, 1, 2]ā¤]
moritz_ rakudo: say (^(3,3)).perl
p6eval rakudo 33212: OUTPUT[[0, 1, 2, 0, 1, 2]ā¤]
07:51 justatheory joined
wayland76 rakudo: say .WHAT() 07:51
p6eval rakudo 33212: OUTPUT[Failureā¤]
moritz_ rakudo: say (^(10,3)).perl
p6eval rakudo 33212: OUTPUT[[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 0, 1, 2]ā¤]
wayland76 Well, it was worth a go :)
moritz_ wayland76: it's not wrong, $_ is undef, which can be an instance of class Failure 07:52
07:58 pcbuetow left 08:00 justatheory left 08:01 lambdabot joined 08:05 bacek_ left
wayland76 Oh, I agree it's not wrong, I just wanted to see what it would do (I was being silly) 08:05
rakudo: y0! 08:06
p6eval rakudo 33212: OUTPUT[Statement not terminated properly at line 1, near "!"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)ā¤]
wayland76 rakudo: 0!
p6eval rakudo 33212: OUTPUT[Statement not terminated properly at line 1, near "!"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)ā¤]
wayland76 No zero-factorial :) 08:07
moritz_ postfix:<!> is not part of the language 08:10
(we had to leave at least one simple example for those who want to show off with defining operators :-)
08:30 iblechbot joined
wayland76 rakudo: sub postfix:<!> { say "Sending mail using postfix... :)"; } 08:33
p6eval rakudo 33212: OUTPUT[Statement not terminated properly at line 1, near ":<!> { say"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)ā¤]
wayland76 rakudo: sub postfix:<!>() { say "Sending mail using postfix... :)"; } 08:34
p6eval rakudo 33212: OUTPUT[Statement not terminated properly at line 1, near ":<!>() { s"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)ā¤]
wayland76 rakudo: sub postfix:<!>(String a, String b) { say "Sending mail using postfix... :)"; }
p6eval rakudo 33212: OUTPUT[Statement not terminated properly at line 1, near ":<!>(Strin"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)ā¤]
moritz_ wayland76: rakudo doesn't let you define operators yet
wayland76: and the signature would be ($a), not (String a, ...) or something 08:35
wayland76 Oh, yeah, forgot; I have a project or two that are waiting on operator overloading :)
(ie. so I should've known); thanks :) 08:36
moritz_ STD.pm does parse it correctly, and even parses the defined operator with correct precedence
std: 5!
wayland76 STD: sub postfix:<!>(String a, String b) { say "Sending mail using postfix... :)"; }
p6eval std 23095: OUTPUT[############# PARSE FAILED #############ā¤Syntax error at /tmp/JaFLgI1O8t line 1:ā¤------> 5!ā¤ expecting any of:ā¤ POSTā¤ infix or meta-infixā¤ infix stopperā¤ postfixā¤ postfix_prefix_meta_operatorā¤ standard stopperā¤ statement modifier loopā¤
..terminatorā¤ whitespaceā¤00:0...
wayland76 STD: sub postfix:<!>($a) { say "Sending mail using postfix... :)"; } 08:37
moritz_ wayland76: you still need sigils on your variables (and String isn't a builtin type)
wayland76 STD: sub postfix:<!>($a) { say "Sending mail using postfix... :)"; } 5!
moritz_ I think we borked evalbot :) 08:38
wayland76 Uh-oh -- now that I know how to do it... :)
08:39 p6eval joined
moritz_ std: sub postfix:<!>($a) { say "Sending mail using postfix... :)"; } 5! 08:39
p6eval std 23095: OUTPUT[############# PARSE FAILED #############ā¤Syntax error at /tmp/A43rmv7nyt line 1:ā¤------> y "Sending mail using postfix... :)"; } 5!ā¤ expecting any of:ā¤ infix or meta-infixā¤ infix stopperā¤ standard stopperā¤ statement modifier loopā¤ terminatorā¤00:12
moritz_ ah no, it's just case sensitive
p6eval ..156mā¤]
moritz_ std: sub postfix:<!>($a) { say "Sending mail using postfix... :)" }; 5!
wayland76 Ok :) 08:40
p6eval std 23095: OUTPUT[00:12 156mā¤]
moritz_ and there need to be a ; after } unless it's followed be a newline
wayland76 Ok :)
DemoFreak rafl: gm ;) 08:49
rafl DemoFreak: moin! 08:53
08:55 schmalbe joined 09:01 justatheory joined 09:10 ejs joined 09:11 revdiablo left 09:13 kane_ left 09:19 dukeleto_ left 09:21 justatheory left, kane_ joined 09:44 smallfish left 09:47 smg left 10:21 alc left 10:55 apeiron joined 10:59 drbean joined 11:17 smg joined 11:23 smg left 11:25 ashizawa left, ashizawa joined, cookys joined, lisppaste3 joined, rafl joined 11:51 kanru left 11:52 kanru joined 12:05 vixey joined 12:15 drbean_ joined 12:28 drbean left 12:33 meppl joined 12:34 dalek left, dalek joined, alech left
pugs_svn r23096 | jnthn++ | [t/spec] Unfudge complex autoviv test for Rakudo. 12:36
rakudo_svn r33214 | jonathan++ | [rakudo] Complex multis for various ops should coerce their maybe-not-Complex argument so we don't run into Parrot MMD errors. Do it by introducing .Complex method on Any and calling that. 12:40
12:54 smg joined, sri_kraih_ joined 12:55 Bzek left 13:03 Bzek joined 13:04 d4l3k_ joined, wolv joined, jnthn_ joined
pugs_svn r23097 | jnthn++ | [t/spec] Tests for declaring and doing roles in namespaces. 13:04
13:05 pmichaud_ joined, wolverian left, PerlPilot joined, wolv is now known as wolverian, dalek left, pmichaud left, PerlJam left, d4l3k_ is now known as dalek, jnthn left
rakudo_svn r33215 | jonathan++ | [rakudo] Fix declaration and doing of roles in multi-jointed namespaces. Patch partly courtesy of Chris Dolan. 13:10
r33216 | jonathan++ | [rakudo] Add namespaced roles test file.
13:12 sri_kraih left, jnthn_ left 13:26 StephenPollei joined 13:32 ruoso joined
ruoso Hello! 13:35
lambdabot ruoso: You have 1 new message. '/msg lambdabot @messages' to read it.
rakudo_svn r33217 | jonathan++ | [rakudo] Fix bug in enum code generation. When adding a vtable method, you need to mark it anon too, otherwise it'll also appear as a normal method. 13:40
ruoso @tell pmurias git clone [email@hidden.address] 13:49
lambdabot Consider it noted.
ruoso @tell pmurias but I didn't start anything yet there... I'm quite convinced we can do a lot of things without it, like most of the boostrapping of ClassHOW and Object... 13:51
lambdabot Consider it noted.
13:55 jonathan_ joined, jonathan_ is now known as jnthn
jnthn rakudo: eval { class A { has $.x } }; say A.new(x=>5).x 13:55
p6eval rakudo 33217: OUTPUT[5ā¤Null PMC access in find_method()ā¤current instr.: 'parrot;Perl6;Compiler;main' pc 138837 (src/gen_actions.pir:14613)ā¤]
jnthn Hmm. But it doesn't do that in the REPL...
But with -e it does. Oddness. 13:56
13:58 schmalbe left
rakudo_svn r33218 | jonathan++ | [rakudo] Add a missing null check in !clone_attr. 14:00
r33219 | bernhard++ | [codingstd] remove a trailing space
14:22 PerlPilot is now known as PerlJam 14:25 pmichaud_ left, pmichaud joined 14:27 PerlJam left, PerlJam joined 14:32 ejs left, ejs joined 14:38 masak joined 14:42 ashizawa left
pugs_svn r23098 | jnthn++ | [t/spec] Eliminate useless/wrong use of eval. 14:57
14:57 ejs left 14:58 smg left
rakudo_svn r33220 | jonathan++ | [rakudo] Should check the parameter passed to eval is a string, in line with S29. 15:00
15:09 ejs joined 15:11 adc_penner joined
pugs_svn r23099 | ruoso++ | [mildew] simplify lexical-sub test a little bit... 15:12
rakudo_svn r33221 | jonathan++ | [rakudo] Call panic on sub-nodes of EXPR nodes, which end up being PGE::Match rather than Perl6::Grammar. This means we show the intended, helpful error. 15:20
15:20 ejs left
rakudo_svn r33222 | jonathan++ | [rakudo] When we fail a type check on a parameter, report the name of the thing we were calling. 15:30
15:34 Lorn joined
pugs_svn r23100 | ruoso++ | [mildew] lexical subs working!!!! 15:39
15:46 jhorwitz joined 15:55 stephens joined
pugs_svn r23101 | ruoso++ | [smop] idconst supports .true, it is only false for 0-sized identifiers... 16:01
r23101 | ruoso++ | [mildew] a test showing that recursion on lexical subs work!
moritz_ ruoso: did you see the "man or boy" test? 16:04
ruoso moritz_, I don't think so
moritz_ ruoso: t/spec/integration/man-or-boy.t 16:05
ruoso: in that file there's a link to wikipedia with background
ruoso: it''s basically a heavy regression test for lexial vars and subs
ruoso cool 16:07
I'll take a look
moritz_ ruoso: you might need to desugar it a bit before it runs on mildew, but if lexical subs work than all the grunt work is already there 16:08
ruoso moritz_, yes... mildew still lacks "for", which is the next thing on list btw 16:09
and comparison tests also
but besides that, I kinda think it would work
16:10 pmurias joined
pugs_svn r23102 | ruoso++ | [mildew] small fix in the way to treat arguments to terms 16:10
ruoso moritz_, mildew is kinda coming the other way around if we compare to rakudo, for instance...
pmurias ruoso: hi 16:11
lambdabot pmurias: You have 2 new messages. '/msg lambdabot @messages' to read them.
moritz_ ruoso: yes, I noticed :-)
ruoso hi pmurias
pmurias
(touchpad)---
moritz_ pmurias++ # keep karma preservation law valid :-) 16:12
ruoso btw... mildew pass 30 tests now... of its own test suite, of course... 16:13
pmurias that's not much
ruoso well.. it's 100%
:D
that's a prettier number ;)
masak statistics++ 16:14
ruoso now we just need "for", "map" and "grep" to support ClassHOW.can 16:17
then we might have the proper ClassHOW.dispatch implemented
pmurias we could just use "map" instead of those 3 things
ruoso yeah... it's almost the same thing... 16:18
but I still would like to have some clarification on the iterator issues
before we implement it
but for some reason I think TimToady is expecting us to TITS and then he will realize if it's a good approach 16:19
pmurias ruoso: re smop5 it would be usefull if we wanted a proper compiletime/runtime mixing 16:20
rakudo_svn r33225 | jonathan++ | [rakudo] Refactor .= operator so that it now handles having a container looked up from a keyed access on the LHS.
ruoso pmurias, indeed... but I think we still have a lot of work to do before we can do that type of mixing
TimToady, do you think you would have clarifications on the iterator issues soon? or are you expecting implementations to experiment and do all the proofs of concept before we have a detailed spec? 16:23
pmurias ruoso: how much things do we need before we can bind a C++ class? 16:24
ruoso pmurias, it depends whether you want to callback smop or nto 16:25
s/nto/not/
16:25 sri_kraih joined
pmurias i just want to try binding the Ogre game engine ;) 16:25
ruoso using foreign objects should be fien 16:26
fine
you just need a typemap
pmurias ruoso: why is callbacking smop difficult
?
ruoso pmurias, because it would probably require longjmp hacks 16:27
16:29 smg joined, sri_kraih__ joined
ruoso pmurias, but maybe implementing a longjmp SMOP frame would be a good idea 16:29
there's probably a way to do it as a generic thing 16:30
probably something like setting up a continuation for a C call...
which is a SMOP interpreter 16:31
rephrasing...
a "C Coroutine SMOP Interpreter"
so it could interoperate with other interpreters
that's actually a freaking cool idea 16:33
and a very important step in the SMOPP5 project, to support XS-Compatibility
pugs_svn r23103 | jnthn++ | [rakudo] Some tests for .= operator. 16:34
pmurias ruoso: how is it supposed to work? 16:36
16:37 dukeleto joined
ruoso pmurias, the glue code would setup this "c coro interpreter" before calling the actual C code 16:37
in the middle of the C code, when calling back to SMOP
it would save the current C context inside a new "SMOP continuation", which would be set as the "current continuation" 16:38
and it would return to the main SMOP loop, using the context saved before calling the actual C code
16:39 Exodist joined
ruoso then the SMOP code would set a new continuation 16:39
continue the loop
eventually call setr
and goto
then it would call eval
and it would simply resume the C coroutine
probably returning the value sent in the setr 16:40
pmurias ruoso: you are aware we can only unwind the stack with setjmp/longjmp
ruoso hmmmm... 16:41
16:41 sri_kraih_ left
ruoso there should be a way 16:41
16:44 sri_kraih left
ruoso pmurias, I'm not sure that's entirely true 16:44
16:45 hercynium joined
ruoso pmurias, isn't it possible to save a state before calling a function, save another state inside the function, longjmp to the outer state and longjmp to the inner state again? 16:45
as long as the functions that call setjmp don't return 16:46
which is the case for the SMOP main loop
pmurias ruoso: not safely 16:47
ruoso why?
pmurias if you recurse in that function you overwrite the stack
ruoso hm?
pmurias setjmp doesn't save the stack it only saves the location of the stack top 16:49
ruoso right... that's why you can't return
(without invalidating the saved stack) 16:50
pmurias when i longjump to the outer state and call a function the new function uses the part of the stack used by the inner function
ruoso you mean that it can't have two stacks with the same root 16:52
well... I think there are coro implementations in C 16:53
they probably already solved all this issues
pmurias ruoso: it's possible if you reserve space on the stack 16:55
pmurias has to think it over 16:57
16:57 pmurias left, schmalbe joined
ruoso @tell pmurias www.softpanorama.org/Lang/Cilorama/...in_c.shtml 16:58
lambdabot Consider it noted.
ruoso @tell pmurias better yet: www.xmailserver.org/libpcl.html 17:04
lambdabot Consider it noted.
17:07 IRSeekBot left 17:13 renormalist left 17:15 masak left 17:18 d4l3k_ joined, dalek left, d4l3k_ is now known as dalek 17:28 [particle] joined 17:32 araujo left, dukeleto left 17:55 smg left
rakudo_svn r33228 | jonathan++ | [rakudo] Use the Class PMC's new add_vtable_override method. 18:20
18:27 bacek left 18:28 km3 joined
pugs_svn r23104 | moritz++ | [t/spec] simplify man-or-boy.t 18:29
18:40 [particle] left
rakudo_svn r33230 | jonathan++ | [rakudo] Make 'my &x = sub { ... }' store the sub lexically. 18:40
18:40 km2 left
TimToady what's with all the "sub {...}" when bare "{...}" should work just the same? 18:41
moritz_ paranoia ;)
18:42 bacek joined 18:46 [particle] joined 18:52 mtrimpe joined 19:08 tewk_ is now known as tewk, [particle] left 19:09 [particle] joined, [particle] left
rakudo_svn r33231 | jonathan++ | [rakudo] Bug fix for subs that take parameters with the & sigil; we need to register the symbol after we stripped. 19:10
19:11 mtrimpe left 19:13 ZuLuuuuuu joined, kst` is now known as kst 19:14 aindilis` left, aindilis joined
TimToady hmm, looks like commmitbit is down again :( 19:20
well, can't look at it now, gotta do a school run...
bbl &
19:21 ZuLuuuuuu left
moritz_ TimToady: it's up again 19:24
19:26 smg joined 19:33 araujo joined
pugs_svn r23105 | jnthn++ | [t/spec] Just compute the first 10 values of the boy or man test; it's enough to see it works and avoids hitting Parrot's default 1000 recursion levels limit. 19:46
r23106 | moritz++ | [t/spec] more comments to man-or-boy.t 19:48
19:49 pbuetow joined
rakudo_svn r33233 | jonathan++ | [rakudo] A crack at getting is copy to do the right thing. We may need to revisit this again later, as I've still not convinced myself it's going to do the right thing in every case (but didn't work out one where it won't yet). It passes all spectests that already passed and 19:50
..makes the man or boy example's use of is copy work, though.
r33234 | jonathan++ | [rakudo] Add the man or boy test to spectest.
19:52 mncharity joined
pmichaud TimToady (and others): where do we want the synopsis documents to live in pugs/ ? 19:54
19:56 pbuetow left 20:07 PZt left 20:09 PZt joined
moritz_ what's wrong with the current location, docs/Perl6/Spec/ ? 20:12
pmichaud nothing, just making sure that's where we want them. 20:21
moritz_ some brave soul needs to fix smartlinks.pl after that ;) 20:22
TimToady we probably want to move to names like S16-io.pod or some such 20:23
pmichaud I'll get the S##.pod files into docs/Perl6/Spec
then they can be easily renamed from there.
(or moved to other directories)
TimToady thanks
pugs_svn r23107 | putter++ | [misc/elfish/STD_blue] Continued adaptation to current STD.pm/gimme5. 20:25
r23107 | putter++ | Tweaks for: some hash literal support; foofix:<x> names; slurpy params; classname literals; empty multisigs. One more elf source file works.
pmichaud ...temporarily disabling pugs-commit mailing list so we don't end up with 100+ commit messages 20:27
pugs_svn r23108 | jnthn++ | [t/spec] Couple of tests for lexicals with & sigil. 20:28
mncharity TimToady: It looks like I'm getting down to two issues. (a) ./viv ../../misc/elf/elf_h_src/PrimitivesP5.pm burns >3GB on the 150 line file before I have to kill it. (b) The previously mentioned $.m(3) arguments showing up in an unexpected place. (b) I can work around, though it would be nifty if it "just worked". (a) is more painful. Any thoughts? 20:29
20:29 Exodist left
mncharity bbiab 20:30
20:30 a-lex joined, a-lex left
pugs_svn r23109 | pmichaud++ | Creating dirs for synopsis load from perl.org 20:32
pmichaud prepare for possible flood of pugs_svn commits 20:33
unless someone knows how to turn it off briefly
moritz_ there's nothing wrong with a brief flood ,)
jnthn Provided we get a nice rainbow afterwards. ;-) 20:34
20:34 [particle] joined
pmichaud crosses fingers, hope he doesn't screw this up. 20:34
pugs_svn r23115 | allison++ | Adding the first two retroactive synopses. 20:35
r23116 | allison++ | draft of syntax synopsis, after review by p6l
r23117 | larry++ | 2004 edition of rules. :-)
r23118 | larry++ | utf8ified, random cleanup
r23119 | larry++ | Structure is good.
r23120 | larry++ | Added precedence table, random cleanup.
r23121 | larry++ | Forgot utf8ification.
r23122 | larry++ | Utf8ified.
r23123 | larry++ | Random cleanup.
r23124 | larry++ | Documented "=" package.
r23125 | larry++ | Added notes about optimizing closures to mere blocks.
r23126 | larry++ | Incomplete utf8ification of E<> forms.
r23127 | larry++ | Utf8ification.
r23128 | luqui++ | Added the =~ to ~~ operator renaming.
r23129 | larry++ | Changed $=line et al. to $?line.
r23130 | larry++ | Uppercased the compiler variables.
r23131 | larry++ | Added examples of $? variables.
r23132 | larry++ | Recent tweaks to &?SUB and such.
r23133 | larry++ | Forgot to change date and version.
r23134 | larry++ | Clarified whitespace proscription following sigils.
r23135 | larry++ | Changed symref syntax back to ::($str) to not be ambiguous with %foo::{$key}.
r23136 | larry++ | Added ::?CLASS and ::?ROLE forms.
r23137 | larry++ | Changed main default to strict.
r23138 | larry++ | Simplified operator quoting rules by unification with hash subscript syntax.
20:35 pugs_svn left
moritz_ ok, enough flood for now ;) 20:36
literal aww
pmichaud yes, I'm up to revision 23500 on the import already
so we would've been here a while.
TimToady yowser
pmichaud I really hope I disabled the mail send properly. :-) 20:37
pmichaud checks
looks like "yes." 20:38
we're now up to 24057. 20:39
literal hm, what's the difference between doc/trunk/design/syn/S*.pod and vim docs/Perl6/Spec/*.pod ?
pmichaud do we just want them in Spec/, or should we have a special dir for the synopses?
literal: doc/trunk/design/syn/S* is where they were in the original repository
literal: I'll just move them now to their new location.
literal I see 20:40
pmichaud and then drop doc/trunk/design/syn
okay to restart pugs_svn 20:42
20:43 pugs_svn joined
pugs_svn r24058 | pmichaud++ | Move synopses to their new home. 20:43
r24059 | pmichaud++ | remove obsolete doc/* dirs from synopsis migration. 20:44
pmichaud restored email to pugs-commits 20:46
TimToady: do you want me to re-apply the updates since last Wednesday for you...?
TimToady if you've got 'em handy
moritz_ pmichaud: are all synopsis in the pugs repo now?
pmichaud I can get them handy. :-)
moritz: yes.
literal where were they before? 20:48
ruoso entering annoying mode...
TimToady, have you seen my earlier question?
moritz_ literal: in the svn.perl.org repository
literal ok 20:49
TimToady ruoso: yes, I think there are 4 or 5 levels of laziness/eagerness we need to look at supporting
ruoso looks complicated 20:50
TimToady on one end there's strictly eager that returns you the entire list
pmichaud TimToady: okay if they go in as one commit instead of three?
TimToady at the other is strictly lazy that only does the minimal work to return 1 value
pmichaud: sure
but we also want to support mostly-eager, that does everything not known to be infinite 20:51
and mostly-lazy, that returns a list of what we know right now without doing any more (significant) work 20:52
where 'significant' might also have more than one meaning
pugs_svn r24060 | pmichaud++ | (re-)apply larry++ commits since repository was archived.
TimToady my assumption is mostly-lazy will be more efficient than strictly lazy
pmichaud okay, the synopses are migrated. 20:53
TimToady in any case, we need an item context iterate, and more than one list-context iterate 20:54
which might be a single function with a laziness knob
20:55 masak joined
TimToady beyond that, I suspect which approaches are most efficient and dwimmy may be platform dependent 20:55
so I'm inclined to bind to mostly lazy and mostly eager by default, with pragmatic control in a lexical scope for other defaults
but I'm also interested in seeing the various platforms participate in the design of this, which is another reason the new S7 is out in the pugs repo 20:56
20:57 araujo left
pmichaud excellent, hadn't seen that yet. 20:58
ruoso TimToady, so I think we do need an Iterator type
TimToady oh, and there's hyper in there somewhere as well
pmichaud ...where is the S7?
TimToady ain't one yet
20:58 araujo joined
pmichaud ohhhhhh 20:58
TimToady it's one of the numbers I stole on the spreadsheet the other day
pmichaud I misunderstood "is out in the pugs repo" 20:59
we should put an index into Spec/
call it S00.pod :-P
TimToady copy from spreadsheets.google.com/ccc?key=p8p...v7Fsq9u4eA maybe 21:00
pmichaud commits to the synopses will now go to pugs-commits list. Should commits to docs/Perl6/Spec/ also go to p6l ?
moritz_ pmichaud: yes, IMHO
TimToady yes, IMNSHO 21:01
pmichaud 1+infinity == the 'ayes' have it.
ruoso TimToady, what do you think about the approach in www.perlfoundation.org/perl6/index....p_operator
TimToady it's surreal...
lambdabot Title: SMOP map operator / Perl 6
TimToady so we need a surreal iterator as well, that goes beyond infinity
ruoso TimToady, I'm mostly trying to figure out how that is represented in terms of an API 21:02
considering I'd expect it to be easily re-implemented by the user 21:03
21:03 alech joined
TimToady looks relatively sane to have different actual types that depend on the context; at bind time you could call the iterator factory to make the right subtype 21:05
ruoso right... that was my plan 21:06
so we have two parts on that issue
1) the iterator api
which is always lazy
2) the eagerness triggers
TimToady if we have list iterators that can return multiple values, then for might prefer an iterator that returns N values without having to call item = multiple times
ruoso we do have list iterators that can return multiple values, aka, map 21:07
TimToady how you structure it will depend on how efficient/inefficient the lazy item = is
I think defining everything in terms of item = is likely to be a performance mistake 21:08
but maybe not
ruoso I'm considering that optimizations can be done
if you know then actual implementation type
TimToady okay, as long as the API is rich enough
pmichaud at the same time, rich api should not be too constraining 21:09
TimToady we can probably afford a fairly rich API since most of this is hidden from the user
ruoso hidden from "most" users
but available to some
that's my point
and in SMOP, it will be implemented in terms of the API at start...
only optimizers will get it faster, later... 21:10
TimToady sure, but even the API might have some implementation dependencies...
unless we're sure everything can be expressed in terms of everything else.
well, probably too early to worry about that 21:11
ruoso that's an assumption I'm making
SMOP actually needs it
because that's how SMOP is bootstrapping its type system
pmichaud I think "everything can be expressed in terms of everything else" might limit implementations a bit. But I'll wait for a spec.
TimToady other platforms will provide other primitives, likely
ruoso pmichaud, if the interpreter know all the participants, it can optimize away the API use
TimToady and if a platform doesn't provide a particular degree of laziness, it's just have to collapse one or more levels 21:12
pmichaud "interpreter knows all the participants" I find to be unrealistic in many cases.
for on, it sounds like we're assuming an interpreter. :-)
*one
ruoso s/interpreter/runtime/
pmichaud, the thing is that, in Perl 6, you can't really know from start if the type is the type 21:13
you only know that at very late run time
pmichaud ruoso: I agree.
ruoso and that applies to very basic language operations
because there are no basic language operations...
everything is dispatched as high-level
so foreign objects will be potentially used by otherwise low-level operators 21:14
but most of the time,
you'll be dealing with native types
er...
native meaning things in your code base
i.e. p6opaque
pmichaud as I said, I'll wait for a spec and comment then. 21:15
ruoso or a compact int array
pmichaud okay, pugs commits to docs/Perl6/Spec are now being sent to p6l.
TimToady I think the best way to develop this api is by bouncing a straw man back and forth a few times
ruoso the thing is that in SMOP, I start assuming everything is foreign 21:16
pmichaud Someone probably needs to subscribe or otherwise whitelist messages coming from [email@hidden.address]
I'm not sure how to do that.
ruoso so I can have dumb types in SMOP that play the role of high-level types
i.e. "constant identifiers"
pmichaud I'll bbl. 21:17
ruoso that's why an API is so important to me
so I can replace the dumb implementations with the complete ones as the runtime evolves
TimToady, but anyway... is it ok to assume that List assignment should replace its internal storage, leaving the references to the old storage valid? 21:18
21:20 PtZ joined
ruoso that question basically means: Is list assignment implemented by the lvalue list itself? or does the runtime replaces the array? 21:20
ruoso have to leave now... 21:21
ruoso will backlog
21:22 ruoso left
TimToady the goal for assignment in general is for it to work as much as possible like Perl 5 does it, within reason 21:24
there isn't really a "list assignment" exactly. There's assignment to an array, and assignment to a slice. 21:25
a slice assignment just sends as many values (copy semantics) to each element that wants them, left to right 21:26
if any element of the slice is an entire array or hash, then the rest of the list is treated as a array/hash assignment
with array/hash assignment, the container empties itself, then copies everything in
pmichaud back (finished putting thanksgiving groceries away) 21:28
21:28 PZt left
pmichaud TimToady: what you describe sounds like the way rakudo is going. 21:29
(or, perhaps I should say that my plans for rakudo match what you just said.)
TimToady note that p5 goes to extra lengths to make sure that ($a,$b) = ($b,$a) works without clobbering the rhs value of $a 21:31
21:34 schmalbe left
TimToady and, in fact, a strictly lazy implementation cannot do ($a,$b) = ($b,$a) properly at all 21:37
pmichaud on this topic, S02-literals/autoref.t has stuff like:
@other[1] = @array;
TimToady so I think the moslty-eager is the right default for assignment
pmichaud where it expects @array to become an arrayref. I think that's a list assignment.
TimToady yeah 21:38
pmichaud to do the arrayref, would @other[1] = $@array; work?
TimToady or item @array
pmichaud okay
TimToady or [@array], of course
pmichaud I'll change the tests, then.
that answers one of my (as yet unanswered) questions on p6l 21:39
TimToady hmm, don't recall seeing that one
pmichaud I think you might've been shouldering other issues then.
21:40 hercynium left
pmichaud original at groups.google.com/group/perl.perl6....1381d4a2b1 21:41
TimToady always seem to get ADD when I don't need it, and vice versa...
lambdabot Title: More about arrayref/hashref in spectest suite - perl.perl6.language | Google Gro ..., tinyurl.com/6cmvb9
moritz_ rakudo: say [].min 21:43
p6eval rakudo 33241: OUTPUT[Use of uninitialized valueā¤ā¤]
masak bug? :) 21:44
21:44 Lorn left
pmichaud min of an empty list should be...? 21:44
masak Inf?
TimToady +Inf 21:45
moritz_ NaN? undef? FATAL_WARNING?
masak files rakudobug
TimToady according to S03, if you consider it equiv to [min]()
pmichaud undef is what it returns now. 21:46
and .min isn't constrained to be numeric, is it?
TimToady no, but +Inf isn't constrained to be numeric either, necessarily 21:47
pmichaud ah. :-)
Tene p6eval: [].min
pmichaud Tene: try 'rakudo' or 'perl6' 21:48
literal perl6: [].min
p6eval pugs: No output (you need to produce output to STDOUT)
..elf 24060: RESULT[undefā¤]
..rakudo 33243: RESULT[undef]
moritz_ pugs is broken in p6eval atm
pmichaud it might be nice if p6eval responded to "p6eval:" 21:49
moritz_ pmichaud: with a usage message? or with the same as 'perl6:'? 21:51
pmichaud I was thinking same as 'perl6' 21:52
21:52 smg left
pmichaud but usage message could work. 21:52
21:52 pbuetow joined
pugs_svn r24061 | moritz++ | [evalbot] implement usage message as suggested by pmichaud++ 21:56
21:57 p6eval left, p6eval joined
moritz_ p6eval: say "hi" 21:57
rakudo: say "alive" 21:58
p6eval rakudo 33243: OUTPUT[aliveā¤]
moritz_ wonders why it didn't work... 22:05
mncharity back 22:07
TimToady: where are we at? 22:09
TimToady mncharity: re (a), currently STD generates a new lexer for every macro, which seems to chew up memory quickly 22:12
I need to find some way of cutting down unnecessary lexer generation
mncharity macro? you mean multi infix:<blah> decls? 22:14
moritz_ if there'a already a proto for that another multi shouldn't require a new lexer. 22:15
22:18 Bzek left
mncharity there are 50 multis. but little overlap - almost all are distinct... (don't know what our vocabulary is)... names. 22:18
wayland76 Are macros generally better than eval?
moritz_ is fish better than bycicles? ;-)
masak yes! 22:19
(at least in water)
mncharity So 50 lexers at 3GB is order 100MB per lexer... does that sound plausible?
wayland76 And certainly tastier :)
Maybe a different question; do macros have less chance of causing runtime errors? 22:21
22:21 adc_penner left
moritz_ do fishes contain more plutonium than bycicles? ;-) 22:23
wayland76: a randomly written macro will likely cause as many errors as feeding random data to eval 22:24
mncharity wayland76: re fish riding bicycles, fish get compiled away at compile time. So you can have abstractions which don't have a runtime performance cost. The compiler can see and optimize.
masak moritz_: as far as I know, fishes contain no bicycles at all... 22:25
moritz_ masak: at least not the small fishes ;) 22:26
masak and the large ones are mammals :)
pugs_svn r24062 | lwall++ | [Spec] get started on the renaming 22:29
22:29 dukeleto joined
wayland76 Well, I think that pretty much covers everything. Thanks :) 22:29
Incidentally, on the topic of Iterators, I had a go at implementing tree iterators for Perl6 using the SMOP map interface 22:30
mncharity TimToady: ok, thanks. regards (b), any guidance on whether I should expect to see it fixed in current weeks, or should proceed with working around it? 22:31
pugs_svn r24063 | lwall++ | [Spec] more name stuff
wayland76 Apart from the fact that I needed to pass in a bunch of parameters to determine the iterator type (self-or-descendant, child, following, whatever), it was sufficient to the task
mncharity fyi, with the 51 multis pruned to 25, it looks like memory usage tops off at 2GB. 22:32
pugs_svn r24064 | lwall++ | [Spec] bleah, svn getting confused, and so am I 22:33
TimToady mncharity: re (b), .() is parsed as any other postfix, so notionaly is a wrapper around the "name 22:34
"
masak TimToady: can I assign the value *-4 to a variable? 22:38
(I'm trying to figure out how magical Whatever stars in indexes really are) 22:39
mncharity And 12 multis is 1.2GB. So two datapoints and an inequality suggest ~100MB per lexer. so 100MB per 'multi foofix:blah (...){...}'. ah well. I can split the file in half as a temporary hack.
pugs_svn r24065 | lwall++ | [Spec] more twiddling
r24066 | lwall++ | [Spec] more renaming 22:44
22:47 masak left 22:49 mncharity1 joined
mncharity1 sorry about that. non-graceful system go boom. backlogging. 22:50
TimToady basically I need to figure out when it's okay to clobber the current grammar in place instead of deriving a new one 22:51
or find a way of doing lightweight derived languages
22:51 Psyche^ joined 22:54 ruoso joined
ruoso Hello! 22:55
mncharity1 TimToady: re lexers, it looks like I'm all set. For elf bootstrap work, I can get by with a halved prelude.
TimToady okay, though that doesn't help with the (lack of) speed
mncharity1 hi ruoso :) 22:56
ruoso hi mncharity1
TimToady, so is it sane to think that @a = @b implies creating a third list that is then stored in VAR(@a) ?
mncharity1 TimToady: re speed, for elf bootstrap work, I can cache. hmm, though... can viv parse STD.pm itself? That's my next step. And it has a bunch of foofix:bar decls I think... checking... 22:57
TimToady p5 only creates the 3rd list if there is a shared name, but in the abstract, yes
sure, viv parses STD
22:57 Patterner left, Psyche^ is now known as Patterner
TimToady and caches lex/STD.pm.store for everything else to use 22:58
mncharity1: no, it has foofix:sym<> declarations really, which are a lot cheaper 22:59
because they don't modify the current grammar on the fly
mncharity1 mncharity1: re (b) and "parsed as any other postfix", so is token variable's #= methcall branch, with its postcircumfix field, now unused?
ruoso TimToady, considering how deep the lazyness can go, I'd say it's pretty hard to figure out if you can avoid creating a third list (unless it's the only assignment to that list)
TimToady but assignment is now defined as eager, I believe 23:00
or mostly-eager
ruoso sure...
TimToady but yes, you'd still have to glare at it pretty hard to determine non-overlap 23:01
mncharity1 re foofix:sym<> are a lot cheaper... foofix:sym<xx> is different than foofix:<xx> ?
TimToady yes
we use different names for talking about the rules that parse an operator than the name of the operator itself
so grammars can distinguish their own operators from what they're parsing 23:02
a grammar can parse prefix:sym<++> without changing the meaning of its own prefix:<++>
23:03 Psyche^ joined
TimToady and the foofix:bar shortcut is probably a bit confusing that way, since it's really short for foofix:sym<bar> 23:03
23:03 mncharity left
ruoso TimToady, so it is sane to think that all the eagerness is evaluated by the interpreter? not by the list object itself and its fellow iterators... 23:04
(by "interpreter" here I refer to all the infra-structure that is external to the object)
TimToady I think it's too sane 23:05
ruoso is that bad?
mncharity1 hmm.... so it sounds like svn.pugscode.org/pugs/misc/elf/elf_...tivesP5.pm 's multi prefix:<++> (...) etc can and should be prefix:sym<++> instead, and that will make the performance problem go away?
TimToady I don't think it will perform well if iterators don't know how to return a "batch" when that's appropriate
ruoso TimToady, again... I assume that's a job for the optimizers 23:06
TimToady mncharity1: only if you write it as a grammar rather than a prelude
and then you'd have to run it through gimme5 currently
if the prelude wants to discuss its own operators then it needs to use the non-:sym form 23:07
now, it's certainly *possible* to express a lot of prelude in grammar, after all, half of what's in STD should be in the prelude, conceptually 23:08
mncharity1 re "only if you write it as a grammar"... ah. hmm. or I simply declare it a temporary fudge. gimme5 isn't even being given the elf preludes, it's just getting by on its own.
ok, so that leaves (b). is token variable's methcall branch depreciated? 23:10
ruoso TimToady, but is it sane to assume that the iterators should provide at least the full-blown API? while it's implementation-dependent when that API is actually used? 23:11
TimToady no, I think it's probably just broken 23:13
there should probably be a .[0] in there since $<twigil> is optional
and probably a .text too 23:14
ruoso: I don't know which one you mean by "full-blown" 23:15
ruoso TimToady, the API that won't probably perform well...
TimToady I'd call that one the minimal one, not the full-blown one...
ruoso ok... agreed.. 23:16
mncharity1 ok. my thanks. I'm all set. I'll kludge (a), and deal with (b) as a temporary workaround if I confront it before it gets fixed.
ruoso TimToady, and the runtime should use the minimal API with every custom iterator
TimToady certainly if item context is pulling items out one by one, you have to have some mechanism for storing the rest of any that you already know
ruoso that's what I've been calling "iterator in item context" 23:17
23:17 Patterner left, Psyche^ is now known as Patterner
ruoso it would store the capture and iterate in its items, as well as consume several times while an empty capture is returned 23:18
23:22 smg joined
ruoso that's how map { }, grep { }, map { } can work 23:23
23:23 pbuetow left
TimToady assuming an empty capture is distinguishable from out-of-data 23:23
ruoso sure... out-of-data is an exception, as you said other day 23:24
TimToady now considering what happens when someone wants to generate a list of exceptions... 23:27
ruoso TimToady, a list of exceptions is still returned inside a capture
TimToady trew nuff 23:29
23:29 bacek_ joined
TimToady it's a good thing one of us is smart 23:29
mncharity1: I believe I have a fix for $.m(), testing... 23:30
23:30 dukeleto left
pugs_svn r24067 | lwall++ | [STD] fix parsing of $.meth() to attach postcircumfix as special form 23:32
23:32 iblechbot left
ruoso TimToady, so I'll proceed with the API in the wiki 23:33
mncharity1 TimToady: oh, yay! hmm... now just how late can I be for dinner... 23:36
TimToady ruoso: feel free to transcribe that into S07-iterators.pod if you like too 23:39
and if you're feeling particularly energetic you can write S08-captures.pod while you're there too :) 23:41
meppl good night 23:47
23:49 meppl left
wayland76 S07++ and S08++ :) 23:51
23:54 hercynium joined, vixey left
pugs_svn r24068 | jnthn++ | [t/spec] Unfudge a couple of array passing tests for Rakudo. 23:54
mncharity1 almost there... 23:57
good night all! &
pugs_svn r24069 | putter++ | [STD_blue] $.meth() works. TimToady++
r24069 | putter++ | 3 more elf source files work.
23:57 mncharity1 left

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!