% iblechbot has joined #parrot % cosimo has joined #parrot % cosimo has left cosimo!~cosimo@pat-tdc.opera.com % ruoso has joined #parrot % jisom has left jisom!~jisom@74-134-230-123.dhcp.insightbb.com % alvar has joined #parrot % Ademan_ has joined #parrot % Ademan_ has left Ademan_!~dan@h-69-3-235-2.snfccasy.dynamic.covad.net % kj has joined #parrot % IllvilJa has left IllvilJa!~jilves@emea-netcache1.oracle.co.uk r25579 | kjs++ | trunk: : [docs] add a bit to the pct_optable_guide.pod diff: http://perlsix.org/svn/parrot/revision/?rev=25579 % IllvilJa has joined #parrot r25580 | kjs++ | trunk: : [docs] more updates to pct/pct_optable_guide.pod diff: http://perlsix.org/svn/parrot/revision/?rev=25580 r25581 | kjs++ | trunk: : [docs] add a see-also, faq, and some more sections that need explanation to pct/pct_optable_guide.pod diff: http://perlsix.org/svn/parrot/revision/?rev=25581 % cognominal_ has left cognominal_!~cognomina@82.67.232.89 r25582 | kjs++ | trunk: : [docs] past_building_blocks.pod: fix a spelling error and explain term "rakudo". diff: http://perlsix.org/svn/parrot/revision/?rev=25582 % cognominal_ has joined #parrot Aside question, did you fix the build for parrot yet? Can you do a make install to somewhere and detach it from the svn checkout? Or is that still buggy % contingencyplan has joined #parrot % HG` has joined #parrot % DarkWolf84 has joined #parrot % cosimo has joined #parrot % dwave has left dwave!~asksolem@pat-tdc.opera.com % dwave has joined #parrot % dwave has left dwave!~asksolem@pat-tdc.opera.com % dwave has joined #parrot % Partizx has left Partizx!~Partizx@220.110.142.38 AndyAway: aren't the shootout examples supposed to be examples of specific shootout tests? (wondering if reducing 9 to 7 there is a bad thing long term) Alias_: (fix the build) which platform? % Andy has joined #parrot % Andy has left Andy!~Andy@64.81.227.163 % DarkWolf84 has left DarkWolf84!~dwolf@89.215.234.147 % GeJ has left GeJ!~geraud@edna.nealab.net % GeJ_ has joined #parrot % skv_ has joined #parrot % skv has left skv!~skv@87.242.97.68 % skv_ is now known as skv % AndyAway is now known as Andy Coke: It might be, but at this point it is effectively a stop to my running "make test" % alvar has left alvar!~alvar@sub87-230-120-57.he-dsl2plus.de Coke: so call it a shootout-related test instead % rdice has joined #parrot % silug has joined #parrot % peeps[work] has joined #parrot Coke: note at head of examples/shootout/ack.pir says 'ackermann - ack(3, 9) is default / shootout runs ack(3, 11)', so 9 seems already reduced well, that's good then. it's more ok that andy changed it % mj41_ has joined #parrot % uniejo has left uniejo!~uniejo@langebro.adapt.dk % dwave has left dwave!~asksolem@pat-tdc.opera.com % mj41 has left mj41!chatzilla@pc-jurosz.ro.vutbr.cz % dwave has joined #parrot % mj41_ is now known as mj41 (also: 'time for ack(3,11): 0.8s (AMD X2@2000)', so we've quite a way to go) YEAH :w ack(3,11) should be 2^(11+3)-3, i think: 16381 a bit of memoizing would make it run _real_ fast! Or just make it a print. oh, no! calculate, just don't calculate anything twice. % rdice has left rdice!~richarddi@67.99.242.2 Yes, I understand. I was making a funny. Building on your funny. * spinclad ENOHUMOR today :( so you're suggesting pre-memoizing... ikindigit s/g/gg/ % gryphon has joined #parrot % contingencyplan has left contingencyplan!~contingen@cpe-76-186-27-146.tx.res.rr.com % contingencyplan has joined #parrot r25583 | kjs++ | trunk: : [docs] add section about pasttype and pirop traits in optable guide. diff: http://perlsix.org/svn/parrot/revision/?rev=25583 % gryphon has left gryphon!~gryphon@dsl-209-221-185-54.zipcon.net % gryphon has joined #parrot % cosimo has left cosimo!~cosimo@pat-tdc.opera.com % sjansen has joined #parrot anybody care to do a first review on docs/pct_optable_guide.pod? It's kidna complete, except some loose ends that I need to check, but don't really feel like doing right now. % DarkWolf84 has joined #parrot r25584 | kjs++ | trunk: : [docs] pct fix more todos in documentation. diff: http://perlsix.org/svn/parrot/revision/?rev=25584 kj: http://perlpunks.de/paste/show/47ab3276.688b.17d typo + one additional paragraph moritz: thanks! What does "bloat" mean? r25585 | kjs++ | trunk: : [docs] pct/pct_optable_guide.pod: add note about left being default assoc and 'none' as an option. don't know exactly what that implies. diff: http://perlsix.org/svn/parrot/revision/?rev=25585 (maybe it should read "can't bloat" .. kj: increase size unneccesarly as in "blow up" ah i see your grammar explodes by adding new precedence levels yes right. good point. that's perhaps easier to understand if you put it that way "Perl 6 rules are run as ordinary subroutines" - aren't they methods, technically? yeah, but for parrot there's not really a diff you think it adds to clarity? not clarity, but it's a bit more correct ;) (btw, I wrote all this from the top of my head; not sure if everything makes sense) (i've never been a math guy) i even just figured out what this associaty works (why is it called left and right :-P ) there is a difference between sub and method (self) ok will fix it afk for a bit % Ademan has left Ademan!~dan@h-69-3-235-2.snfccasy.dynamic.covad.net kj: the first example to "is assoc" is wrong - the (1 + (2 ^ (3 * 4)) kj: ... parse tree contradicts the precdence rules you defined earlier r25586 | kjs++ | trunk: : [docs] tell about rules being methods, not ordinary subs. suggested by moritz++. diff: http://perlsix.org/svn/parrot/revision/?rev=25586 kj: more doc nits: http://perlpunks.de/paste/show/47ab3845.2b0e.26 Coke: As I was told, all of them? In order to get a working perl6, you had to leave it inside the build You couldn't make install it % rdice has joined #parrot % rdice has left rdice!~richarddi@67.99.242.2 % parrot-poke has joined #parrot % alvar has joined #parrot moritz: thnx. but I don't quite understand kj: which part? the IMPORTANT part! wit the power op (1+(2^ .. ^ has higher precedence than *, right? yea ... mmm ... seen pmichaud pmichaud was last seen on #parrot 12 hours and 45 minutes ago, saying: iiuc, CPS is the underlying implementation for Parrot's call/return scheme so 2^(3*4) can't be the outcome of parsing 2^2*4 kj: better just stick with infix:- or infix:/ examples heh yeah :-) everybody understands those I suck at math mmm guess I better not make such remarks; future employers might check them out :-P just don't tell them you're 'kj' ;) ha ha. the example with - is good, right? right (as far as I understand - I'm better with math, but I have next to no experience with parsers) hrmm... do i need to specify .params if i'm going to do a tailcall? particle: .return the_thing_to_tailcall(x, y, z) on the calling side...and on the callee side, the param declarations are just normal. Or am I missing something? i'm thinking about implementing 'is export' in rakudo Ah. I was going to try and avoid doing a tailcall... in src/classes/List.pir we have .sub keys :method and .sub keys .sub keys is simply a tailcall to the method .sub keys :multi('List') .param pmc list .return list.'keys'() .end And just make a MultiSub PMC and push the method onto it...maybe. :-) Or if there already is a multi-sub PMC for the name, it's easy. Alias_: ... ah. install ne build, hence my confusion. Just find it and push it on. * Coke waves from home with strep throat. Thing is, is export is really a role, but I'm not sure MMD cares about "does" as well as "isa"... * jonathan had been thinking about is export too :-) well, it's a funny role i consider 'is export' to take effect during composition time, not at runtime but i suppose it could be either, now that you mention it is SomeClassHere for inheritance is blurry like that. :-) @keys = keys @array; so, at runtime, lookup is done for a function named 'keys' Yup. if we're using mmd lookup, it sees there's no invocant and notices the first arg is type List Right. previously, 'is export' has registered a multisub with that type ok, that works. however, it's the registration bit i'm fuzzy on. moritz: thanks for the review. particle: Yes, that is...fun. :-) kj: you're welcome If you have a class, then it has a block. The same block we stick the class declaration in. if you don't need to know the params, it's pretty straightforward huh? it == past? Yup Somewhere, you need to stick the code that applies the trait to the sub. r25587 | kjs++ | trunk: : [docs] fix a bit based on comments by moritz++ diff: http://perlsix.org/svn/parrot/revision/?rev=25587 By calling trait_auxiliary:is sure, fine. I *think* that pushing it onto $?PACKAGE wil usually work out OK. here's what i'm fuzzy on: i want to say: i mean, i *have* to say: .sub delete :multi('List') .param pmc list .param pmc indices :slurpy .return list.'delete'(indices :flat) .end that is, i need to know the params in order to call the method but i'd like to grab the invocant and call the method on that, without processing params Important question before I say what I'm about to...one second... * jonathan has an idea but wants to see if it'll work :-) On what you just said though, you can have non-multi params too, right? Oh, wait, you did what I was going to say. :-) :) yes, slurpy/flat helps but i don't think a nullary method will work well with slurpy/flat i think parrot will err on params... but that's just my in-head compiler particle: OK, here's a stream of stuff I've been thinking about. I think that we are going, for Perl 6, to actually generate *every* sub as a :multi, where it's marked with multi or not. I think we're going to have to, I mean. unless it's marked 'only' Not even that, though. Because you need to be able to install extra multis to handle junction auto-threading. iirc 'proto' is default At a Perl 6 level, sure, it's an error to re-declare a non-multi even with different arguments. if the sub doesn't accept junctions, it'll work with 'only' (at the parrot level) i see what you're saying. The way we need to implement auto-threading is 1) Try and dispatch. Every argument is Any by default, not object. 2) If it fails because we have a Junction in there, which is not a subtype of Any, we catch that error. 3) We look at see if the reason we failed was because one of our parameters was a Junction. If not, we show an error or call the AUTOLOAD or whatever Perl 6 calls it. yep. % DarkWolf84 has left DarkWolf84!~dwolf@89.215.234.147 associecity => associativity? we need our own mmd on top of parrots If so, then we generate a wrapper that does the auto-threading on that argument or those arguments, and install it as a multi. Then, next time, it's "free" because multi-dispatch does the Right Thing for us. yes, I'd do it as a failover the first time, so you get the benefits of direct dispatch if you never feed it a Junction For this to work at all, we need the call to fail when you pass a Junction to a non-Junction or Object parametter. yep binding should fail correctly so we need to do :multi(Any) on everything particle: Right, that's my thought. you mean to an Any parameter, and Object should accept Junction TimToady: Yes, that's what I mean. :-) non-(Junction or Object) :-) to many irc channels, not enough time... this will make the PAST uglier :( particle: So going back to your is export... Consider we write this class: .namespace [ 'Foo' ] .sub init :init :load $P0 = newclass 'Foo' .end .sub test :method :multi('Foo', int) .param int i say i .end gah, lack of line breaks. i can read it OK Then: Method call: $P0 = new 'Foo' $P0.test(1) Or sub call: $P1 = find_global [ 'Foo' ], 'test' $P1($P0, 1) given that Junctional processing is going to be generally pretty heavy, it might not be noticable to make it failover every time rather than just the first TimToady: that would incur the penalty for anything derived from Object and not Any esp if the Junction is the first thing you're trying to bind (which is currently just Junction) it's not a penalty unless binding fails right. you have to do something if binding fails in any case it's simpler particle: So here's how we avoid creating a wrapper at all. $P2 = new 'MultiSub' $P1 = $P1[0] # this gets the method itself, in fact we need an iterator here for all of 'em, I guess push $P2, $P1 set_global 'test', $P2 # stick it in the namespace test($P0, 42) # works! One Sub PMC to rule them all. :-) $P1 = $P1[0] # what's $P1? $P1 = find_global [ 'Foo' ], 'test' It's a continuation of my earlier example code ;-) hey! Duh! Oh no, not duh. :-) But anyway, it's not quite that simple. You need to find_global 'test' first, to see if there already is a multi. right. But hopefully I've explained the idea well enough. and what if the sub is defined first does 'only' on a sub affect a method in the same namespace? I guess if you export something and there is a sub marked only in the namespace you export it into, then you're asking for trouble. ;-) I think we're going to need to subclass the Sub PMC to give ourselves places to stick flags like "is_only". module Foo { only sub bar(Foo $x) {1} }; class Foo { method bar {1} } Ouch. I'd not throught of that. % DarkWolf84 has joined #parrot I'd not even thought of using module and class on the same namespace. Is that legal? I'd normally say "I hope not", but this is Perl 6. :-) ...enough rope... (subclass Sub) ... Tcl does that! (though there is definitely room for more stuff to go in Sub that we could share.) yes, implement things that Tcl will need. :) moritz: btw, I never talk about higher/lower precedence, because people can't keep those straight; instead I always use tighter/looser particle: So, you're going to implement is export now? ;-) % slightlyoff has joined #parrot sure, matt. just for you. .sub 'trait_auxiliary:is' :multi('export', 'Sub') "is export" is a trait, so theoretically has an effect immediately on whatever is being declared; of course, that effect might just be to mark it for later processing in this case You'll need to hack the name "export" into existence somehow though. TimToady: might be better, yes ;) it doesn't help that half the precedence tables in existence are printed with the high precedence operators on the bottom :/ we even made that mistake in the first Camel ;) maybe it goes with trees that grow downward... but then surely it should be recursive ascent parsing your roots are showing at least they're not square well, half of them are square; only every other generation has a square number * jonathan needs to go shop, cook, eat, etc. Back later. * TimToady waves damn. there's a lot more of perl 6 to implement. is there much left to spec? if we merely implement what's specced so far, we'll have a darn good language :) % Theory has joined #parrot heh % kj has left kj!~IceChat7@ip565fd420.direct-adsl.nl % slightlyoff has left slightlyoff!~slightlyo@216.239.45.19 % Ademan has joined #parrot % Davemon has joined #parrot % Dave has left Dave!~dave@pool-141-150-16-151.mad.east.verizon.net % rdice has joined #parrot % rdice has left rdice!~richarddi@67.99.242.2 % cj has left cj!~cjcollier@c-71-231-92-35.hsd1.or.comcast.net % Davemon is now known as Dave % DarkWolf84 has left DarkWolf84!~dwolf@89.215.234.147 % Theory has left Theory!~Theory@dsl093-038-250.pdx1.dsl.speakeasy.net % pjcj has left pjcj!~pjcj@84-73-177-217.dclient.hispeed.ch % pjcj has joined #parrot % Andy is now known as AndyAway % IllvilJa has left #parrot r25588 | jonathan++ | trunk: : [core] Tidy up my messy C. diff: http://perlsix.org/svn/parrot/revision/?rev=25588 % Limbic_Region has joined #parrot % rdice has joined #parrot r25589 | particle++ | trunk: : [config] great, i finally find a useful warning for 64bit conversion on windows, and it gets deprecated! diff: http://perlsix.org/svn/parrot/revision/?rev=25589 heh heh...gotta love MS compilers. I found a (minor) bug in the C# 3.0 one within a week of it's release. is it fixed yet? Dutch Perl Workshop 2008-02-29 www.perlworkshop.nl Tene: Don't think SP1 for VS 2008 is about yet. It's not a bug in the code it generates. It's a warning bug. Emits a warning about a field in a structure never being assigned to if the only place you assign to it is inside a Linq query. So minor, but annoying. Juerd: I'll be there! % iblechbot has left iblechbot!~iblechbot@ppp-62-216-216-9.dynamic.mnet-online.de w % sjansen has left sjansen!~sjansen@hq-nat2.gurulabs.com jonathan: Great :) Juerd: Last stop on my February conference visiting. I hope you will still have some energy left :) % rdice has left rdice!~richarddi@67.99.242.2 % wknight8111 has joined #parrot particle: Decided to do a workaround for the OO issues. Took all of 30 minutes. Going to blog about it now... yay blog! r25590 | jonathan++ | trunk: : [rakudo] Work around the class/role ordering issue for now, so people can play with OO in Rakudo until we work out a Real Solution. diff: http://perlsix.org/svn/parrot/revision/?rev=25590 % Theory has joined #parrot just noticed the following when doing mingw32-make perl6.exe - gcc: unrecognized option `-Foperl6.o' is that expected? % slightlyoff has joined #parrot % slightlyoff has left slightlyoff!~slightlyo@216.239.45.19 % slightlyoff has joined #parrot % peeps[work] has left peeps[work]!~peepsalot@bwext.kpimdp.com jonathan: ping % gryphon has left gryphon!~gryphon@dsl-209-221-185-54.zipcon.net http://use.perl.org/~JonathanWorthington/journal/35610 pmichaud: pong pmichaud: How's things? the OO code you've been adding seems to have a lot of inline PIR... is it really necessary to do that? Couldn't we just create a built-in function or method to handle it? (things are relatively good here... we have one more hospital night to go) pmichaud: Perhaps, though you need to generate bits to apply the traits, bits for the "does", etc. But yes, agree, would be tidier to factor out what we can. what does 'apply the traits' mean in this context? Call trait_auxiliary:is I haven't looked at the code in great detail... I just know that it seems to be an awful lot of inline PIR, where I was expecting to have some builtins to handle it Built-ins at what level? PCT, or just that code factored out to PIR subs? code factored out to subs i.e., creating a new subclass can be done via a method call to a helper sub that does it, instead of writing inline PIR to do it Right, agree. it was just an observation. Maybe I'll have some time to play with it a bit later tonight. I'll happily admint I've been going for getting it working and getting the right semantics over beautiful implementation. It was eaiser to see what was going on for me just having it all there in one place. well, I think that "getting the right semantics" is probably method-based anyway, also. :-) Aye, true. I don't disagree it wants some refactoring and clean-up anyway. It's a first cut at it. sounds good I'll see if I can do some refactoring on it Sure, well, I hope it's a good start if nothing else. oh, it's at least that. :-) I just did a journal post showing what people can do with it now. I want to try and start getting the spec tests in shape for this stuff. that would be a very good thing to do. It's certaily time to, now the very basic bits are in place. What do you suggest in terms of directories to break it up? Is S12-class and S12-role not granular enough? could start that way it's subversion, so it's easy enough to change :-) True. Also thinking maybe it'd be good to support writing bits of the compiler in Perl 6 now. Especially since there's enough support to start writing some simple classes. sure I'm having trouble finding good bits to write in Perl 6 though :-) Hmmm...in that case maybe it's less of a priority and there's better things I can do? Or Just Do It anyway for the bits we can do that way? definitely do it for the bits we can do that way OK, sure. I don't expect it to be so hard. ...famous last words... :-) Are parsing rules and grammars on your todo list? I'd love to be able to show those working in Ukraine... you mean to be added to rakudo? sure. I don't know if I'll have those within a week, but it might not be too difficult. Ukraine is two weeks. ;-) I'll have lots of time sat in German language talks next week that I can't really follow, so expect to have time to hack a decent bit on Rakudo. jonathan, was is the url of your site, I forgot it what.... you spoke of german... oops, that was in the backlog cognominal_: http://www.jnthn.net/ - really need to update the upcoming talks page there... Would be nice if jonathan's use.perl blog was syndicated on planet.parrotcode.org yea # jonathan++ anyway Maybe Coke can make that happen. * jonathan updates his forthcoming talks section on his site. jonathan: I can make that happen. plz. send me a reminder email. You can haz reminder email. I am reading jquery in action and see how much javascript is a rough semantical subset iof Perl6 except the jquery implementers do the multi dispatch manually % DarkWolf84 has joined #parrot * Coke guesses the reminder email was silly, since he just did it. ok. jonathan's new feed is now in planet parrot; no clue how long it will take to refresh and show his entries. w00t Coke++ Coke++ % Andy has joined #parrot sleep & % Ademan has left Ademan!~dan@h-69-3-235-2.snfccasy.dynamic.covad.net % slightlyoff_ has joined #parrot % slightlyoff has left slightlyoff!~slightlyo@216.239.45.19 r25591 | chromatic++ | trunk: : [pmc2c] Prettify generated PMCs slightly. Debugging these guys is less than : fun. diff: http://perlsix.org/svn/parrot/revision/?rev=25591 evening % Ademan has joined #parrot % AndyA has left AndyA!~andy@82.152.157.85 % AndyA has joined #parrot % alvar has left alvar!~alvar@sub87-230-120-57.he-dsl2plus.de % chromatic has joined #parrot r25592 | chromatic++ | trunk: : [src] Fixed a minor typo. diff: http://perlsix.org/svn/parrot/revision/?rev=25592 Is anyone having trouble building NQP? * particle tries like, gc trouble? No. My fault. For fun, ask yourself "Self, why are there three different versions of registering a type in three different places?" % Theory has left Theory!~Theory@dsl093-038-250.pdx1.dsl.speakeasy.net hey, c, anything we can pair on tonite? take schwern's 2038 code to parrot. Want to know some really grotty code that is a speed hit? sets.c in compilers/imcc/ is a huge timesink for Perl 6 and PGE. I'm not sure that's entirely the problem (but it's not great code). I think the real problem is that compute_dominators() is lousy. Cleaning that up (even making it so that it's remotely readable) would be an improvement. An easier task (if Amy makes you watch the O! channel tonight) is slapping Parrot_oo_ on all non-static functions in src/oo.c. A supposedly fun thing that I'll never do again (thank you, David Foster Wallace) is to figure out what's barfy about t/library/pg.t (probably our PostgreSQL NCI bindings are out of date) and make it not barfy. Finally, someone needs to poke Coke and ask him if the Tcl patch I sent him a few days ago helps. Who holds on to the commit bits? Could ajr get one, rather than feeding thru me? % HG` has left HG`!~wells@user-5446064f.lns1-c12.dsl.pol.co.uk chromatic: Actually, Amy is at school show, being a Teacher Presecnce or something. woohoo, it's the Hallmark channel for you tonight then! Right now watching Dirty Jobs w/Quinn later, I'll watch some Modern Marvels as background That Dirty Jobs show is a horrible invention. % Ademan has left Ademan!~dan@h-69-3-235-2.snfccasy.dynamic.covad.net % Ademan has joined #parrot Why do you say? Regarding ajr, the committers just have to look over the number and quality of patches he's submitted, then hold him down, make him sign the contributor agreement, put a sack over his head, and re-enact The Crossing of the Desert and The Unblinking Eye and he's in. compilers/imcc/sets.c doesnt look too messy to me, and I don't see the function "compute_dominators()" in there I went out to dinner with a very lovely young lady a while back, and she mentioned the show, and then we had a long discussion of animal husbandry. Quinn & I discuss much science watching it. it's in reg_alloc.c, wknight8111. okay, thanks It's in compilers/imcc/cfg.c Er, yes. Sorry. r25593 | petdance++ | trunk: : ajr documentation, plus some consting r25594 | chromatic++ | trunk: : [oo] We can get by with only one version of this function (though its cousin in : src/pmc.c looks awfully familiar). register_type() is now What bothers me about sets.c is that a struct is unnecessary (it's just a bitmap), and that working eight bits at a time seems a little silly to me. If we can rely on a 32-bit type, for example (and we can know this at compile time), we can chunk along four times faster. I don't even know what it's doing, but that doesn't stop me from refactoring. Which, sets or dominators? Yup. This is where I pull out my computer science education and tell you that Mozart was in his early 30s when he died without completing his Requiem, Handel was the red-haired priest, and Wagner hated Mendelssohn.