Pugs 6.2.9 released | pugscode.org | pugs.kwiki.org | paste: sial.org/pbot/perl6 | www.geeksunite.net
Set by autrijus on 4 August 2005.
dduncan fyi, Darren_Duncan is now dduncan by default, to be nicer to those on 80 character IRC clients 00:05
hobbs nicks over 9 chars long are heretical! ;) 00:06
dduncan well then, consider me saved!
seen the light, etc ... 00:07
autrijus, is it practical or possible to change my Pugs committer name likewise? 00:08
or don't worry about it
stevan dduncan: the pugs commiter name is the same name as the openfoundry user. You could try to change that in the user admin, however, the solution might be to get a new user account on openfoundry. However the problem with that is that open foundry keys user accounts by email address, so you would also need to change that. In other words, it may not be a trivial task. 00:33
dduncan okay then, no need to change anything 00:34
stevan dudley: if you want to work on the Java version go right ahead 00:45
stevan is backlogging sorry :)
dudley: I would use the Javascript version as a guide, I tried to mirror them both
dudley: there is also an ANT build.xml file which runs the JUnit test suite and can do coverage with EMMA
however, you need to have both the Junit and EMMA jars in your path (obvviously) 00:46
if you want I can email the jars to you, just email me at stevan at iinteractive.com and I will send them on over
dudley stevan: What's EMMA 00:56
?
dduncan: autrijus had me sign up as a new user, then just gave that new user the committer bit. So I guess I have two openfoundry names now. 00:57
dduncan I C 00:58
dudley Three, if you count my daughter, who got the committer bit for her 0th birthday ;-)
dduncan how'd you swing that?
dudley I announced here that my wife was going into labor, then when we got back, someone asked if the new baby was a committer yet... 00:59
ahh, code coverage. I'll check that out. 01:01
stevan: I'm going to port the javascript straight across at first. It makes for some ugly Java code, but I'll straighten that out eventually. 01:02
biab 01:03
iblech stevan: Hm, will Perl6.MetaModel support multisubs eventually? 01:14
Or are multisubs considered not to be part of the metamodel?
stevan dudley: Java is always a little ugly :) you cannot avoid it 01:46
iblech: re: multisubs and the metamodel
1) I dont really have enough knowledge of the inner workings of subroutines and MMD in particular to really do justice to the idea 01:47
2) I assumed that multi-methods would be built upon multi-subs, which I don't see as the responsibilty of the metamodel
and since p5 and JS did not have multi-subs, or formal parameters for that matter I punted 01:48
iblech 1) Neither do I, that's the problem :)
stevan LOL
iblech Idea...: 01:49
stevan iblech: Punt and let autijust do it
iblech Someone writes the MMD dispatching code in Perl 6
And we can compile it to JS and P5 then :)
Right :)
stevan nice
iblech: have you seen luqui's Class::MultiMethods::Pure?
iblech I've read the docs, but not more
stevan 3) they havent settled on the dispatch semantics of MMD anyway 01:50
iblech: personally I like the CLOS generic function idea
it is very clean
but unfortunately it is not appropriate to Perl 6 OO
iblech: have you implemented multi-subs? 01:51
in PIL2JS?
iblech No. Additionally, -CPIL doesn't give my any sub definitions for multi subs 01:52
stevan ah,.. then it's autrijus's fault anyway :)
iblech Good :) 01:53
stevan iblech: how has the integration of methods gone?
how are you handling that?
iblech method foo (Foo $self: ...) is compiled to sub foo (Foo $self, ...); PIL2JS.addmethod(::Foo, &foo) 01:54
Then the metamodel knows about the "foo" method of Foo 01:55
And calls are dispatched accordingly then
stevan nice
so are you using the Perl6.Method class still?
iblech Yep. 01:56
stevan hmm
iblech But I could do without it
stevan then we could possible hack together multi-methods if you want
iblech oooh
stevan here is my thought
iblech listens 01:57
stevan 1) we alter the metamodel such that if it gets a method of the same name, it checks to see if that method is a multi-method
because currently it just allows it to overwrite
2) we create a multi-method class
and put all the glue in
then as far as the dispatch is concerned 01:58
we punt on the details
and just hack together something simple for now
as long as the dispatch code is in it's own method, it should be able to be replaced easily 01:59
I can try to hack this into the JS metamodel if you want to try?
iblech Excellent.
:) 02:00
stevan it wont be pretty :)
iblech But we won't be able to test it from PIL2JS, because of the -CPIL problem
Sure :)
stevan iblech: so -CPIL does not handle it at all?
iblech Reusing the code for multisubs would be simple, too 02:01
Right, try pugs -CPIL -we 'multi foo (Str $a) {}; multi foo (Num $a) {}'
(No "foo" in the output) 02:02
But that's probably easy to fix in Pugs.Compile
stevan iblech: all I got was 02:03
PIL_Environment
{ pilMain = ((PStmts PNoop PNil))
, pilGlob = ([(PSub "&*END" SubPrim [] PNil )])
}
iblech Right.
No &foo.
stevan yeah, what happened?
iblech dunno, ask autrijus :)
stevan did it croak on the duplicate name? 02:04
ah
I see, multi is not understood maybe?
iblech It seems so. Probably (but I haven't looked at the source), &foo is internally a list of subs, and -CPIL doesn't know how to compile a list of subs 02:05
stevan iblech: yes,... hmmm 02:06
where is that autrijus when you need him :P
iblech :)
stevan iblech: well I will hack together a skeleton for mutlimethods anyway 02:07
however, I will caveat it with the fact I have no real idea what I am doing :)
iblech Thanks :) 02:08
We'll just ask Damian or Luke to give us appropriate JS code :D
stevan :)
whoever gives us the code first, wins 02:09
iblech Right :)
stevan if you look at t/80_Code.t in the p5 mm there is some basic multi-sub code
but it only dispatches on the number of params 02:10
iblech Hm, this is enough to get Test::skip working :) 02:11
stevan ok 02:12
I should have a rough sketch of the code in a few minutes
iblech :) 02:13
stevan++ # fast :)
stevan not really fast, just porting the perl 5 code :)
iblech :) cheater :D
stevan uh oh 02:14
I need to be able to check the num of params in the function signature 02:15
can I do that with PILJS?
where should I look for that
iblech You can (&foo.arity), but the problem is:
The metamodel doesn't really get the actual boxed Function object, but only methods which, when invoked, return that object 02:16
But I could simply give you the arity as a parameter
(That's because you're able to rebind &subs, and it'll be even more important when I've finished my full-continuations-in-JS patch...) 02:18
stevan so something like:
Perl6.MultiMethod.prototype.call = function (arity, inv, args)
iblech Will work
stevan wait, that wont work with the code I am writing though 02:19
I have a list of method objects
and I need to check the arity on them
iblech Oh, right.
stevan and match it against the length of args
iblech Then I'll pass the arity to new Perl6.Method(...)? 02:20
stevan ok
brentdax Continuations in Javascript? 02:21
iblech Right. Instead of emitting say(foo(), bar()), I emit foo(-> $a { bar(-> $b { say $a, $b }) }), basically 02:22
call/cc is then trivial
stevan so iblech will you pass the arity to the Perl6.Method constructor?
iblech And it'll allow me to add support for coroutines, gather/take, and &sleep
brentdax Awesome. Scary, but awesome. 02:23
iblech stevan: new Perl6.Method(associated_cls, function (...) {...}, arity)?
stevan iblech: nice, that way I can test it too :)
ok, give me a little more than 5 minutes on this one then :P 02:24
iblech stevan: Great :)
:D
stevan I have to write new code
iblech: is the inv included in the arity? 02:27
iblech It is. Should I -1?
stevan no
iblech ok
stevan I will include it in the calculations :) 02:28
ok, questions.
should methods be declared multi?
or should trying to add a method of the same name automagically make it multi 02:29
iblech It's probably to cleaner to demand a multi declaration
stevan I am leaning towards the first one right now
ok 02:30
well actually the metaclass wont care
polymorphism++
Perl6.Class is the only place I have to address it 02:31
iblech Hmm... Seems "use YAML; warn Dump($self->[PIL::CC])" modifies $self 02:36
(WTF?) 02:37
Seems a bare "$self;" does too 02:38
Apparently a bug in perl5
stevan iblech: hmmm, it seems arity is already a property of function objects, so i am renaming for now 02:47
oh wait
sorry my bad,.. no need
iblech FWIW, all PIL2JS subs have a JS-arity of 1 (always "args", like Perl 5's @_) 02:48
stevan ok, well we are future proofing it :) 02:49
for when PIL handles them correctly too
autrijus greetings. 02:58
autrijus backlogeth 03:00
mugwump hi there ... wow, my work queue is empty at new jobs already
stevan hola autrijus
iblech Hi :)
mugwump s/new jobs/new job/
autrijus hey stevan, iblech.
I dreamed of TSa last night.
more precisely, at his seemingly completely insane reply to my container thread. 03:01
and truth emerges: he is actually making sense.
I wonder if it means I can read deeper than the surface, _or_ I'm insane.
so I'd like to check with you a bit :) 03:02
stevan I am not sure if iblech and myself are the high watermarks of sanity
mugwump either :)
but go ahead and try 03:03
autrijus consider %h.kv and $s.method
intuitively there's a difference
.kv is acting on the container, and .method is acting on the object obtained by .FETCH on the container.
so far so good?
stevan yes
iblech In PIL2JS, .kv is actually acting on the .FETCH of the container, too (a PIL2JS.Hash object)
wolverian oh, why can't everything just be objects! or lists! :)
mugwump nods hesitantly
autrijus iblech: aha. how about @ary.end ? 03:04
is that acting on a FETCH'ed list also?
iblech Yep.
autrijus can %h contain anything else than the same Hash object? 03:05
iblech what same Hash object? 03:06
autrijus also, what's stopping $s from storing a Hash object, then?
er, I mean
iblech PIL2JS.Box.prototype.STORE is auto-reffing such cases
autrijus my %h = (1=>2); %h = ();
is %h storing the same Hash object, only with different content? 03:07
iblech Ah. No, in current PIL2JS, %h stores a new PIL2JS.Hash object then
autrijus you know... I like your treatment better :) 03:08
iblech++ # sanity
iblech :)
autrijus the model I was about to say
is that "Scalar of Any" has an AUTHMETH that redispatches all unhandled methods to its FETCH.
whilst Array and Hash does not do such thing. 03:09
so for $s, it handles .tie, .tied, := etc
but .method gets redispatched.
mugwump I'm really confused here. I know I've been out of the loop for a bit, but I don't understand why tie can't be implemented in terms of Proxy objects...
autrijus mugwump: tie %h, 'DB_File'; 03:10
mugwump: try rewrite it using proxy object?
iblech: also, how do you handle (%h := %g) then
iblech: does that mean they share a Box? 03:11
iblech %h := %g is exactly like $a := $b
mugwump can %h not get reblessed into the DB_File class?
iblech I.e. .STORE, .FETCH, and .BINDTO are copied
mugwump hmms 03:12
iblech autrijus: Why should Array and Hash not also redispatch?
autrijus iblech: @ary.foomethod
iblech: if Array does not handle foomethod
what should it redispatch to? 03:13
iblech Ah, of course. Was mixing up Perl 6/PIL2JS
autrijus mugwump: also note that when you tie(%h), then untie(%h) , original content of %h needs to prevail.
mugwump but wasn't "tie" always a hack, anyway? I thought that the only required goal of tie-like functions was to be able to make objects that behave like instrinsic types 03:14
autrijus mugwump: yeah, you're talking about killing tie, and I'm okay with that proposal
mugwump: but the spec says tie stays, so I need to find a way to implement it. 03:15
mugwump: the "tie should die" proposal belongs elsewhere :)
iblech: so, look at pugscode.org/images/container.png for a second 03:16
mugwump sends his comment "We're supposed to be leaving legacy behind in Perl 6" elsewhere
wolverian mugwump, that would be perl6-language
I'll post my "rename 'grep' to something even vaguely resembling English" post too, then :)
autrijus iblech: my initial thought is that $s.foo is dispatched first to the Cell
iblech: which is exposed as surface class Scalar and Scalar::Const 03:17
iblech: then if it does not handle .foo, its AUTOMETH redispatch it to the "Scalar Value" slot.
iblech: and in array's case, exactly the same thing applies, except there is no "array value" to dispatch to -- they are separate containers. 03:18
"they" == "the items"
iblech: your PIL2JS Box 's fetch looks like the Container-Cell fetch
iblech Sounds sane.
mugwump "Perl 6 variables are tied directly at declaration time, and for performance reasons may not be tied with a run-time tie statement unless the variable is explicitly declared with an implementation type that does the Tieable role.
autrijus not the Cell-Value fetch I was talking aboutl
mugwump - S06
iblech Yep.
autrijus mugwump: right, that. 03:19
iblech: good. so if we hav
Array does Cell; Scalar does Cell; Hash does Cell; 03:20
(the "cell" name is throwaway temporary hypothetical device)
iblech: then we can rule that Ref always hold a Cell
and enreference will promost anything not a cell into an appropriate cell -- that is, plural value into Array and other values into Scalar. 03:21
iblech What about \%hash? 03:22
autrijus iblech: so if you reopen the Scalar class to add new methods, it effectively intercepts its content's method. also "implementation class" are merely new Scalar Cell types beyond the default mutable and immutable cells.
iblech: \%hash takes a Hash, Hash does Cell, it holds a Cell, no autoenreference happens
iblech Ok.
autrijus as a direct result, what we used to call "lvalue" now actually means "Cell" 03:23
and the reason that
iblech Right, makes sense.
autrijus 4 = 5
does not work, is because 4 does not typecheck.
and I think that's what TSa was getting at. 03:24
stevan iblech: your JS multi-methods are in r6125 03:25
svnbot6 r6125 | Stevan++ | Perl6.MetaModel - (JS)
r6125 | Stevan++ | * adding very very primative MultiMethods
r6125 | Stevan++ | - added tests file for them
r6125 | Stevan++ | * added multi-method support to Perl6.Class
r6125 | Stevan++ | - added a test for this as well
stevan as promised they are not pretty :)
iblech oooh :)
stevan++
autrijus oh, what about PIL not dumping multimethods?
stevan iblech: it basically takes an array of Perl6.Method objects
hobbs so you guys are responsible for the massively inflated karma on perlbot
stevan and based on their arity, it dispatches
hobbs++ # yes :) 03:26
spinclad well, svnbot does most of the dirty work, but yeah
autrijus iblech: Compile.hs line 152, extend it to handle non-singular list contents, that's all
iblech: should be straightforward as it already allows multiple returns 03:27
stevan oh autrijus, I was talking to matts about doing a port of DBIx::Class to perl 6
and generating the classes with the metamodel
iblech autrijus: Simple :) Will do that tomorrow then
autrijus ooh. bridging it on perl5:DBI ?
iblech++ # cool 03:28
stevan autrijus: probably yes
I wont be able to start on it until mid-sept
iblech But I *really* need to sleep now. (05:28 AM here)
autrijus stevan: cool
iblech Night all :)
autrijus iblech: go sleep you crazy diamond
iblech :D
stevan nite iblech, I hope you dream of container types too
autrijus *wave*
stevan autrijus: I figure too that you will have the metamodel in by then 03:29
autrijus stevan: yeah, that sounds likely
I'll go out and grab some food. be back in a bit
stevan I am also working on documenting the meta object protocol (not sure if you saw the backlog)
autrijus no, will backlog further
I look forward to more Visio time :)
stevan see docs/S12.5.pod in the metamodel 03:30
it is just a start
however, I need to sleep soon as well, nite &
spinclad it sounds like you're extending the MM to meta-calling (MMD and such) 03:31
which i think is excellent and needed. 03:32
dudley tries to resist the urge to give up on java and learn scala.... 03:34
autrijus dudley: that will do you good :) 03:35
Scala and Nice are both very nice.
SamB it would be a shame if nice was not nice
dudley yeah, but I'm working on the java port of the metamodel. Changing horses in midstream and all that.
spinclad stevan++; iblech++ # CPS in JS
spinclad puts down pompoms for the night 03:38
SamB riding horses through rivers sounds odd enough to start with! 03:40
Khisanth eh? 03:42
oh dudley's statement :)
dudley Ok, that's enough for one day. Goodnight all. 03:45
sleep &
dduncan for those of you interested in Perl 5 stuff that will shortly be translated ... I would like to announce that svn.utsl.gen.nz/trunk/Rosetta/ is now fully up to date with Rosetta's current state of being as of r724; however, that is a snapshot and fails a test; r724 is good to study, but r723 passes all tests ... and so, any future commits have brand new stuff 05:49
svnbot6 r6126 | Stevan++ | Perl6::MetaModel - (p5)
r6126 | Stevan++ | * added ::get_P6opaque_instance_data to centralize p6opaque information
r6126 | Stevan++ | - changed all files which accessed data directly to use this function
r6126 | Stevan++ | * moved some things around in Perl6::Class to facilitate the name lookup
r6126 | Stevan++ | refactoring. It is getting there, but not quite done yet.
geoffb Wheeee 10:12
geoffb feels like he got run over by a conference or something
. . . but I just finished the last of my OSCON blogs, so I'm happy.
And maybe tomorrow I can come back to the fun here in Pugsland. 10:13
Mighty quiet around these parts. 10:16
Ah well, I'll come back and visit sometime when the sun is on my side of the planet again. :-)
g'night, all.
castaway_ night geoff :) 10:17
autrijus purrs 10:44
I did a lot of PIL2 design on paper today
autrijus finally got network back
will start hacking :)
castaway_ grats autrijus ! 10:53
10:54 castaway_ is now known as castaway
kgftr|konobi castaway: lo... how's it going? 10:54
jdv79 is the "paper" available?
castaway hey 11:03
not bad, coming together, I think :)
autrijus jdv79: available as dead trees 11:04
I'm coding them down
jdv79 oh, you meant _real_ paper:) my mistake. 11:05
autrijus :)
I'm targetting a reduced set of perl6 currently 11:06
just for experimenting with the new type-directed compilation and PIL2
castaway
.oO( people still write on paper? what a concept .)
 11:08
autrijus I went out of battery ;)
ok, I'll go offline now... can't code with IRC window on 11:09
autrijus waves &
castaway later a* 11:10
svnbot6 r6127 | autrijus++ | * beginning of a separate PIL2 repl loop, in PIL.hs. 11:39
r6128 | autrijus++ | * oops, it now compiles. next up: "print 1" in the grand
r6128 | autrijus++ | Punie tradition!
coral Punie?
or dvorak 11:40
coral &
integral coral: punie == perl1 on parrot 11:45
QtPlatypus Punie is Perl1 on Parrot. The first test was "print 1"
autrijus right, and the only running test. 11:50
"print 10" would fail, for example :) 11:51
Whitesocks I want to send the command "ls -l" to the perl open command from a apache web server... but when i type "ls+-l" in the browser, the "+" doesnt get exchanged for a space..... why?
integral Whitesocks: the perl6 open command?
autrijus Whitesocks: I suggest #perl
Whitesocks Ok thanks. :) 11:52
autrijus :)
autrijus mumbles something about rm+-rf.
nothingmuch rm -rf is the roxor 11:53
QtPlatypus Perhaps we should put something in the topic about it not being a channel for perl5 questions. 11:54
nothingmuch it fixes all those stupid alias rm="rm -i" entries
integral thought the 6 in the name was a giveaway
nothingmuch QtPlatypus: i don't think it's worth that much... we don't get them too often
QtPlatypus I've seen about one question per week. 11:55
Your right that isn't two much.
svnbot6 r6129 | autrijus++ | * the runM monad for generic monad translation
11:58 xinming_xchat is now known as xinming
autrijus hm. 12:01
class Foo { method f (Str $x --> Int) {} )
class Bar is Foo { method f (Int $x, Int $y --> Str) {} }
should such things be tolerated?
also:
my &f := sub (Str $x --> Int) {}; &f := sub (Int $x, Int $y --> Str) {}; 12:02
also:
my Int $f; $f := "string!";
they are similar but not exactly the same. thinking about how far we should go on those.
QtPlatypus First shouldn't be permitted unless multimethods 12:04
THe frist shouldn be permitted unless there muliti methods?
On the other hand you might wish to overwrite methods, which should also be permitted. 12:05
autrijus the thing is that each of those disables a form of early binding.
I'm pretty sure you can't say you "does" a role but does not have a compatible interface. 12:06
that's the thing that makes
role Coercable { method coerce:<as> (::t --> t) { ... } }
work.
QtPlatypus Isn't that why we have "is final"? 12:07
autrijus there is no "is final".
check again?
(we have "use optimize :classes<final>")
but that's not the same.
autrijus ponders some more. 12:08
masak are tests marked "todo" given as much attention as failing tests? 12:22
autrijus no.
I look at :todo<bug> from time to time 12:23
but most :todo<feature> will probably only be fulfilled in the new runcore.
masak it was mostly the :todo<bug> ones i meant
autrijus oh. then, still not 12:24
masak that's too bad
autrijus :todo<bug> means they are confirmed bugs
failing tests are untriaged
and triaging needs to happen before bugfix
so it makes sense to pay more attention to them :)
masak i suddenly realised that i expect bugs to be fixed by themselves 12:25
that's an unreasonable expectation
autrijus indeed :)
I mean, eventually, sure 12:26
masak i just wish there was more i could do to fix *my* bugs :)
autrijus but if you put any time frame on it, then it's less certain.
soon, soon. :)
(assuming you feel more proficient hacking perl6 than haskell)
masak i think i do
autrijus good then.
masak neither very much, but still, perl6 i sorta understand 12:27
autrijus *nod* 12:29
scook0 bah, in one day everything I think I know about the metamodel gets turned upside-down :( 12:31
autrijus scook0: hmm? 12:32
scook0 stevan's comments on p6l
nothingmuch g/qui 12:33
scook0 I thought I understood it
then he says something like "the Class object will be invocant of all the class methods", and I realise I must be totally off 12:34
perhaps someone needs to convince stevan to draw some pretty diagrams :)
autrijus or I should draw one for you
I have half-finished ones.
but I'm compelled to make "print 1" work now :) 12:35
scook0 yeah, I heard--I shall gleefully examine your latest PIL work...
autrijus scook0: it's based on type-directed translation 12:36
for once
for example, when it sees
print 1
then it compiles to
&print(&coerce:<as>(1 as Int, Str as Type) as Str) as Bool
jdv79 i thought what is now called Object was going to be called MetaMetaClass
autrijus scook0: and 12:37
print $_
&print(&coerce:<as>(&Scalar::FETCH($_ as Scalar of Any) as Any) as Str) as Bool
the FETCH is inserted by coercing (Scalar of Any) to (Any) via a compile time binding
but maybeit should be overridable as a coerce:<as> form as well. 12:38
dependent types are so much fun.
masak that's something you don't here every day 12:39
scook0 autrijus: I'm not sure I understood any of what you said, so I'll just peer at the code for now
masak s"here"hear"
autrijus scook0: ok :)
scook0 hmm, is there any way to tell GHC to use /less/ memory? 12:40
even 'make unoptimised' locks up my system with thrashing :(
autrijus allocate less heap?
scook0: oh, you can do the easy thing
edit config.yml
turn off precompile prelude 12:41
scook0 done (thanks)
not that I really run pugs much
I mostly just recompile to make sure Haddock doesn't screw up 12:42
Daniel_Nee Hi, Autrijus: Could you come in tomorrow afternoon around 1:00 PM and we'll go to nanshan to continue the UAT ?
autrijus Daniel_Nee: hey, I already replied your mail
Daniel_Nee: also if you look at the bottom of your chatzilla screen there should be a red "autrijus" button -- pressing it and you'll see a private message from me, and we can talk in that window in private, too :) 12:43
autrijus praises the amazing protocol that is irc. 12:44
Aankhen``: hey 12:47
Aankhen`` Hiya. 12:50
And bye-a. I can't stay. :-\
Will be back later.
autrijus ciao :)
bbiab too &
fglock ping 14:11
autrijus pong.
fglock ping 14:13
autrijus pong.
fglock hi! I'm reading the metamodel - I can't find 'List' there
maybe i'm in the wrong directory 14:14
autrijus hm?
List?
fglock There is an Array.pm - is there a List.pm? 14:15
under "Container"
autrijus I don't think so, currently
list is a value type
not a container type
and I don't think the p5 metamodel deals with value types yet
fglock i've got Lazy lists ready, i think 14:17
svnbot6 r6130 | autrijus++ | * layout for the new stuff -- "make pil" launches into ghci
r6130 | autrijus++ | for PIL.hs. This is strictly temporary.
r6131 | fglock++ | * Array::Lazy - code cleanup
r6132 | autrijus++ | * split into "make pil" and "pili"
r6133 | autrijus++ | * preliminary checkin for literal parsing.
r6134 | autrijus++ | * "print 1" is parsed... how very exciting :)
autrijus fglock: yay. porting it to p5 if you have the tuits? :D 14:19
I'll integrate it to the runtime when I get around to the Plural handling.
fglock do you mean - using coroutines? 14:20
autrijus fglock: yes. 14:23
you can fake coros with an extra callback sub param
the same way as ruby does
or you can ues the more expensive Coro.pm
fglock I started implementing using callbacks - then I changed to coro. but I think it can work both ways. 14:25
even a Sub with a state variable would do
how about "Perl6::List" 14:26
autrijus I'd prefer Perl6::Runtime::List or Perl6::Val::List 14:28
just so we don't conflict massively with other cpan things
or Perl6::Pugs::List
;)
I think Perl6::Value::List can work. 14:29
Arathorn crawls in from the cold 14:30
fglock ok - Perl6::Value::List. should it use 'shift' or 'SHIFT'? 14:31
autrijus you mean the method? call it shift, I don't think it'd hurt 14:32
fglock I'll work on it in the afternoon. Going for lunch now :) 14:33
autrijus fglock++ 14:35
stevan fglock: anything in the metamodel under ::Container should be probably ignored 14:40
svnbot6 r6135 | autrijus++ | * beginning of compilation
r6136 | autrijus++ | * rename PIL.Syn to PIL.Exp
r6137 | autrijus++ | * finishing up the rename.
stevan morning all :)
dudley Morning 14:41
stevan I am responding to TSa on p6l, I have a feeling the cyclical nature of the metamodel will not sit well in his clean Type DAG world
QtPlatypus Good evening
stevan morning dudley, evening QtPlatypus :)
s/Type/Type model/ 14:42
pasteling "Stevan" at 67.186.136.119 pasted "MetaModel cycles" (16 lines, 328B) at sial.org/pbot/12335 14:43
stevan I made him a pretty picture though :) 14:44
castaway pretty pictures++ 14:46
Arathorn hrm, my pine mutilates the pretty pictures :( 14:54
stevan Arathorn: see the nopaste above 14:58
Arathorn has found the H button to get pine to display rawtext sensibly :)
but thanks
15:03 autark is now known as jp-autark
QtPlatypus autrijus: Will PIL2 contain PPos infomation. And if so what do you expect it to look like? 15:07
autrijus QtPlatypus: question 1 is yes 15:13
I think the current line/col range is good enough 15:14
I'll probably put file into another form
and you can deduce your file and line/colrange by looking outward and find first one of each kind
does that sound ok?
yay, the skeleton is done
svnbot6 r6138 | autrijus++ | * rather improbably, "print 1" works, so does "say 123", as well 15:15
r6138 | autrijus++ | as "say say say 123".
autrijus I can focus on type directed translation now.
QtPlatypus nods
autrijus continues to sail on this Heart of Gold thing
QtPlatypus So I'm guessing its a psydo funtion wrapped around the code? 15:16
autrijus it's not pseudo function... 15:17
but rather each node can be annotated with one or more positions 15:18
something like
data PIL = Pos PIL | ...
or rather
data PIL = Pos Pos PIL | ...
QtPlatypus How are thouse annotations serilized?
autrijus and it has type a :: a
for example, say you have
Lit (Single (Int 3))
then the annotated form is 15:19
Pos somePos (Lit (Single (Int 3)))
where somePos is some position information.
I think I'll move the main part on the left
so it's
Pos (Lit (Single (Int 3))) somePos
i.e. exactly the same way current Exp and PIL handle things. 15:20
QtPlatypus nods
autrijus an alternative is have each node carry a position optionally
but then I can't easily "peel them" away
so probably not.
peek (Pos x _) = x
er
peel (Pos x _) = x
peel x = x 15:21
done
QtPlatypus thanks
autrijus and it fusions as well, so if all is in haskell, then those position slots are not filled in the first place
since their evaluation never triggers
so we don't get much speed penalty
QtPlatypus thought that PIL2 was going to look like flattened p6? 15:23
autrijus the prettified form, surely.
maybe it'll be littered with #line -- I mean #file -- directives if you want them. 15:24
QtPlatypus nods and can work with that
QtPlatypus beds & 15:26
autrijus purrs 15:27
fglock ping 15:37
autrijus pong
fglock hi autrijus - I need to define 'Inf' in perl5. Is Perl6::Value::Scalar a good place? 15:38
autrijus Perl6::Value::Num actually. 15:39
Scalar is container, Num is contained
but you can get inf in perl5 using 9**9**9
but otoh, you can't call .. on it.
so it's up to you :)
for the value type names, please follow the new PIL runcore
svn.openfoundry.org/pugs/src/PIL/Val.hs 15:40
Int Int | Num Num | Str Str | Ref Ref | Bit Bit
(they should inherit from the common abstract "Single" class.) 15:41
and List should inherif from "Plural", but you don't need to worry about that yet. 15:42
it'll all become Visiolized :)
visiolization++
fglock where can i get the operator names (the "List API") 15:43
like "..", comma, etc
i've got a solution for detecting "end of list", when the list contains "undef" 15:45
"List" can have the "elems" operator. It returns Inf until the list finishes, and then return "0" 15:46
autrijus cool.
that sounds nice 15:47
I don't think there is a canonical API.
you can grep List in src/Pugs/Prim.hs
to get some idea of that
if you come up with one and it works well 15:48
it will likely become the canonical api.
fglock when I cast a List to a Num - is it a $list.to_num() operation or a Num.from_list( $list ) ? I hope the question makes sense 15:52
autrijus perl6 uses coerce:<as>
so it is to, not from 15:53
otoh, you can define MMD forms from it
so it can potentially be both
but never only from.
also, you're doing the one-level-below runtime class 15:54
so you can use any API you like
it's the job of the inferencer to pick the correct coercion form, and the code generator to call your runtime class.
Aankhen`` Hi.
fglock Hi Aankhen
autrijus but to make things tidier, to_num sounds fine certainly, although you may or may not want use overload. 15:55
Aankhen`` Heya fglock. :-)
autrijus i.e. overload 0+ => \&to_num
(hopefully I'm still making sense)
Aankhen`` wonders how to create a class that functions as an array, with the right sigil, et al.
fglock sure - I'm trying to find out what is the minimum API. 15:56
Aankhen - me too :)
autrijus Aankhen``: class MyArray does Array { ... }; my @x is MyArray;
Aankhen``: will work, does not yet :)
Aankhen`` autrijus: What about if I want to be able to say: my @foo = MyArray.new();
Will that work too?
fglock Aankhen: I think it should work, but you are copying the data from your new array into @foo, and throwing it away, I think 16:00
autrijus Aankhen``: I don't think that works. := works.
Aankhen`` OK.
And if I just say `class MyArray does Array { ... }`, MyArray automatically knows to return an array? 16:01
autrijus er, that's what does means. 16:03
if you think about it a bit :)
liskov and all that.
Aankhen`` I try not to think about it; it hurts my head. 16:04
My understanding of `does` was always that it's simply a way to manage code reuse. Didn't think it would alter the `new` method.
autrijus er, "does" means 16:05
whatever you retur
n
you agree to conform to the Array API in full.
so it's "as good as" Array.
fglock can "does" be tested with "isa"? (I think I tried it before and it didn't work)
autrijus no. "isa" manages implementation reuse
"does" manages interface reuse.
(broadly speaking, that is.) 16:06
Aankhen`` Ah, I see.
"Does" that extend to other roles as well?
autrijus sure. 16:07
xinming is the the only difference between "say" and "print" is say will add a "\n" to the result. but not print? 16:08
autrijus xinming: yes. 16:10
in other languages it's be the different betweehn putStr and putStrLn, or print and println, or write and writeln.
but @Larry reasoned that the "ln" form is actually more common 16:11
and hence should be shorter.
it's called the principle of huffman encoding
er, Huffman coding.
Aankhen`` So are roles only supposed to be interfaces? 16:12
osfameron given how much we talk about Huffman coding, should we Huffman code it to somethign shorter?
autrijus osfameron: I don't think that comes often enough nowadays, and "huffman" would suffice :) 16:13
Aankhen``: they may contain implementation code as a "default case". but it's not reuse -- it's copied straight into the "does" consumer. 16:14
well, it's reuse, but not inheritance, sorry.
so you don't get to call SUPER.
Aankhen`` Ah.
autrijus there's no SUPER for roles.
Aankhen`` Well, thanks for the explanation(s). :-)
autrijus so you can't post/pre process the super methods.
if you'd like to do that, roles isn't for you.
np :)
Aankhen`` class MyClass does Hash does Array { ... }; MyClass.new() # what type of object? 16:15
autrijus MyClass of course.
Aankhen`` Er, right.
autrijus :)
Aankhen`` What sigil?
autrijus both.
my @x is MyClass; (fine)
my %x is MyClass; (fine too)
this is not entirely unlike in perl5 you can tie hash and array to the same class. 16:16
Aankhen`` And `my $x is MyClass` becomes a regular opaque object?
autrijus er, no, it will complain that MyClass doesn't do Scalar, or something like that.
my MyClass $x
is what you want.
Aankhen`` Ah, right.
autrijus shorthand of "my $x returns MyClass" or "my $x of MyClass".
Aankhen`` That's equivalent to `my $x of MyClass`, right? 16:17
OK.
autrijus :)
Aankhen`` boggles at the thought of an instant HashArray class.
autrijus notes you can do that in perl5 with three ways already.
Aankhen`` In one line
?
autrijus sure!
Aankhen`` That is to say, one semi-colon terminated statement? 16:18
autrijus sure!
Aankhen`` Wow.
I've never come across anything of the sort.
My life has been wasted.
:-\
autrijus use overload ('@{}' => sub { [] }, '%{}' => sub { {} }); 16:19
that's one way
*TIEARRAY = *TIEHASH = *FETCH = *STORE = ... = sub { ... };
that's another
[{}]
is the third (deprecated) way. 16:20
perl6 doesn't really add many runtime features
but it made it much more reasonable and cleaner.
Aankhen`` Looks like it.
autrijus (it does add a _lot_ of compile time features though.)
Aankhen`` my HashArray %hasharray = (a => 'b', c => 'd'); %hasharray[0] # == 'b'? 16:22
# given a `class HashArray does Hash does Array`, that is. 16:23
autrijus my %hasharray is HashArray
Aankhen`` Ah, right, my bad.
autrijus your form has it containing HashArray elements.
Aankhen`` That's going to come back to bite me.
Again and again and again.
autrijus :)
I blame it to @Larry.
different things made similar.
Aankhen`` Ah, I personally blame it on my sudden bursts of thick-skull-itis. 16:24
autrijus I contend that %hasharray[0] still won't work. :)
as the sigil rules over the interface
so your Array side of methods did not get recognized
by the inferencer.
but, if you get a p6l ruling
I can think of ways to implement that :)
Aankhen`` So if I want to provide both interfaces, do I use `my $x of HashArray`?
SamB you silly person! never blame thick-skull-itis! software should be designed to relieve it! 16:25
autrijus yes, that would work.
Aankhen`` Cool.
autrijus looks SamB strangely
Aankhen`` SamB: Oh yeah. Sorry. Must have been the thick-sk-- er, nevermind.
Anyway, time for din-din. 16:26
BBIAB.
&
autrijus :)
autrijus starts journaling
fglock I'm writing the file Perl6/Value/List.pm - is it ok to say 'package List' instead of 'package Perl6::Value::List' ?
autrijus fglock: well, I definitely prefer the later
fglock ok
autrijus you can set up local aliases like
my $List = __PACKAGE__; 16:27
or something like that :)
xinming Perl 6 is going to confuse more people then ever. :-)
autrijus xinming: that is just following the grand tradition of perl5.
PerlJam As long as perl6 doesn't cause too much confusion about basic things (or what programmers think of as basic (like the type system)) 16:30
autrijus indeed. 16:32
the type system is perhaps the most significant new semantics over perl5.
PerlJam new semantics == good 16:34
autrijus just like haskell is pronounced "world's finest imperative language" after it got monads, maybe perl6 can evolve to be the world's finest static language.
PerlJam new confusing semantics == bad
autrijus verily.
PerlJam I would be shocked if perl6 became the world's finest static language. 16:35
(but in a good way)
xinming I agree with autrijus, Perl 6 will be. :-) 16:36
autrijus it's pretty tough with grand prior arts like O'Caml, but one can certainly try :)
fglock is finding it hard to write Perl5 after a month doing only Perl6 16:40
PerlJam fglock: I've stopped writing perl6 for a while actually. It was affecting my ability to write perl5 for work. 16:41
fglock i'm too slow :)
PerlJam when I went to explain something to a co-worker and wrote a perl6 solution without thinking, I realized a few things: 1) perl6 will be great once it's ready for prime time 2) a perl6 solution in a house of perl5 isn't too productive and 3) more perl6 would only make me wish for it hard enough to hurt myself :) 16:43
Aankhen`` Heh. 16:44
I wub Perl 6.
Or rather, I wub that subset of it which I can understand. 16:45
Oh dear... the Empire Earth II demo is mad at me for not playing it since I installed it. I better go. 16:46
xinming while =$*IN -> $line {...} ,What does the equal sign mean here please?
brentdax Unary equals is the iteration operator.
(Like the old <blah>, but more general.)
fglock it is just like 'shift' 16:47
autrijus PerlJam: your experience is a prime example of why compiling perl6 to perl5 is The Way Forward :)
fglock: actually it's like "a lot of shift"
xinming: and that should read 16:48
for =$*IN -> $line { ... }
PerlJam autrijus: "a lot of lazy shift"
autrijus PerlJam: right.
"a lot of shift, lazily"
fglock "lazy list constructor"
autrijus decomposer, actually
the $*IN is the constructor 16:49
or is it called "generator / consumer"
xinming :-S
hard to understand for me. :'(
PerlJam xinming: why? What's confusing? 16:50
autrijus xinming: ok. $IN is this object that was called STDIN
xinming: and prefix = replaces <>
and for replaces while :)
xinming: in perl5 you could also say 16:51
for my $line (<STDIN>) {...}
except it will wait until EOF
but in perl6, it will eventually read one line at a time.
fglock is there a place under pugs/ for Perl6::* modules? or do they go straight to CPAN? 16:52
autrijus "eventually" = as soon as I port fglock's implentatino over.
fglock: yes, put them under perl5/
fglock: once they work coherently, they will go straight to CPAN as Perl6::Runtime or something like that.
before that, put it under its own perl5/ subproject.
autrijus can't wait to use Perl6::Runtime at $work. 16:53
xinming autrijus: go on please. 16:54
autrijus xinming: ok... the idea is that =$IN is like <$IN>
except that the list is not evaluated all at once
but rather only at the time it is bound to something, like $line 16:55
so the net effect is that it reads one extra line when the loop runs next
note that this does not describe the current pugs 16:56
which has no support for lazy arrays
but we're -- actually fglock is -- working on that :)
xinming Ok, Then, for Every Array, eval qq{ my $t = (1..8); for =$*t -> $_ { $_.say }; } means the value(not values) in the anonymous list is calculated every time reached for, right? 16:57
autrijus assuming that array does iterator 16:58
xinming Sorry for my poor English. :-)
autrijus and assuming you wrote q{} instead of qq{}
that's fine :) 16:59
and assuming you wrote $t instead of $*t
then yes, 2 is evaluated after 1 is printed.
so you could have written 1..Inf
and get a lot of strings on your screen
instead of like in perl5 where it plain dies.
but you do have to hit Ctrl-C eventually, before you run out of bits. 17:00
(but even 32 bits takes a lot of time to run out.)
xinming my $t = (1..Inf); for $t { $_.say }; 17:01
This will print nothing, right?
SamB what do you mean, run out of bits?
autrijus SamB: never mind... I forgot Int is bigint.
xinming If so, I think I understand what = exactly means here.
SamB I thought Int wrapped around? 17:02
autrijus ponders how array references should stringify.
SamB: nope. int may, but Int is big.
SamB: (note I don't like this.)
Khisanth autrijus: why not stringify to same thing as perl5? 17:03
autrijus Khisanth: ARRAY(0x1234) ?
Khisanth :)
autrijus Khisanth: I somehow doubt its usefulness.
xinming: assuming $t stringifies the array it points to, then yes, it will likely die of a out of memory.
Khisanth one of those things that should be a fatal error? 17:04
autrijus maybe, if we can detect it somehow. 17:05
but sometimes it's kind of hard.
PerlJam xinming: given $t = 1..Inf; for *$t { .say } will never output anything as it will halt executing trying to expand $t. 17:06
xinming: for **$t { .say } will do the same but halt compilation as it tries to expand $t
autrijus Khisanth: unless stringification is lazy also
Khisanth: but in p6, string ne list of char 17:07
so I doubt that can happen.
PerlJam xinming: for =$t { .say } will output a bunch of numbers.
fglock Perl6::Value::List is done - it's 30 lines
autrijus Khisanth: it's like
Data.PackedString.packString (show [1..])
which is also out of memory.
Khisanth autrijus: isn't stringification lazy in p5?
autrijus Khisanth: I mean the process itself 17:08
Khisanth oh
autrijus in haskell, show :: [Int] -> [Char]
and each cons cell is processed separately
so fusion happens and all's fine
but in perl6 String is not [Char]
so I doubt we can lazily print a stringified infinite array. 17:09
(I like that Str is not [Char], fwiw.)
svnbot6 r6139 | fglock++ | * Perl6::Value::List - implements "List" semantics in Perl5 17:15
fglock autrijus: you can print "start..end", but not the whole array
autrijus fglock: right
Limbic_Region IIRC, @larry - or perhaps even $larry has ruled that if you type the parameters of a sub and you don't pass args of that type - p6 will squawk. In what ways and conditions will p6 squawk for improper type use elsewhere? 17:17
autrijus actually, it will only squawk when it can't find the coerce:<as> form. 17:19
it will warn and insert coercion silently otherwise.
Limbic_Region: typechecks is only done on bindings. function application and return involves two implicit bindings. 17:20
svnbot6 r6140 | fglock++ | * List.pm - added reverse() method
Limbic_Region so is there going to be a pragma or something to really squawk?
autrijus Limbic_Region: sure, I think that'd be cool. 17:21
we're talking about
'1' - '2'
Limbic_Region right
and 1 - '2'
autrijus yup
but 17:22
Limbic_Region and '21asdf' - 4
autrijus 1 - ('2' as Int)
will always be fine.
(because coercion typechecks by definition)
Limbic_Region right - explicit type casts required if you say you don't want implicit
xinming PerlJam: So, for **$t {...}; will be a failure at compile time. And for *$t {...}; won't be a failure, But generating the list all the time. 17:23
autrijus Limbic_Region: use strict 'types';
Limbic_Region: I think that'd work.
in fact I'll implement it in pil2 :)
Limbic_Region autrijus - b4 you do 17:24
I found the thread I was talking about
autrijus although there is a very pedantic form of coercion
1 - $x
which I don't think I'll ever warn.
Limbic_Region it was luke - not larry - but he agreed there would be yelling, screaming, and gnashing of teeth
perhaps only in the case of no coercion though 17:25
autrijus my Int $x; # suppose that
Limbic_Region let me link
autrijus then I won't force user to write "1 - ($x as Int)"
(because you see, $x is of type "Scalar of Int")_
xinming thanks all for your patient explaination. :-) 17:26
autrijus to cast it into Int is to call FETCH to it.
ok, I need to run for food. bbl
Limbic_Region grrr - google groups is not playing nice 17:32
subject line was "Argument Type Checking" and the thread was started by me 17:33
svnbot6 r6141 | fglock++ | * List.pm - to_num(), to_str() 17:47
geoffb *yawn* 17:56
Hey, everyone . . . .
fglock hi geoffb
geoffb Two weeks without #perl6 -- I think I'm in withdrawal, despite the sleep deprivation 17:57
Limbic_Region is rather enjoying his respite from all things geek 17:58
geoffb starts the `svk pull` rounds . . . this is going to take a while
You take pretty a pretty soft break, Limbic_Region. :-)
Hey, iblech! 17:59
iblech Hi :)
geoffb How have you been, lo these many days?
iblech :) Implementing CPS in PIL2JS currently
geoffb ooh, very cool.
iblech Allowing me to implement coro, gather/take, &sleep, callcc, &?CALLER_CONTINUATION and many more things :) 18:00
geoffb How is PIL2JS doing otherwise? Is the OOMM integrated?
nice
iblech It is, but we've to wait for -CPIL2 to fully use it
geoffb (`svk pull` isn't finished, so I couldn't check for myself :-) 18:01
nodnod.
And how is -CPIL2 coming?
integral if PIL2JS can pass t/unspecced/cont.t that'll be one up on Eval.hs
iblech geoffb: It doesn't exist yet :)
geoffb wow, I thought for sure autrijus would be mostly there by now. $work must be really interfering. Damnable need to eat. 18:02
iblech But as the new PIL runcore (src/PIL*) will use PIL2, we'll have -CPIL2 soon, I think
geoffb fair enough 18:03
Oh, for those who didn't make it to OSCON in Portland: www.oreillynet.com/pub/au/2333 18:06
^ My two dozen or so OSCON posts
iblech ooh geoffb++ 18:10
geoffb Thanks. Ended up being a lot more work than I expected! 18:11
svnbot6 r6142 | fglock++ | * List.pm - added clone(), is_infinite(), to_bit(), to_ref(),
r6142 | fglock++ | from_single(), to_ref()
r6143 | fglock++ | * Lazy.pm - added concat_list(), from_range() 18:26
r6144 | fglock++ | (forgot to save last changes) 18:37
Arathorn big badda boom 18:48
Limbic_Region 5th element
great movie btw 18:49
wolverian agreed. 18:50
autrijus woot 18:57
role Coerceable[::to] { method coerce:<to> (--> ::to) { ... } } 18:58
multi *coerce:<as> ((Coerceable[T] $from: Type ::T) --> T) { $from.coerce:<to> }
Aankhen`` o_O
autrijus class Array does Coerceable[Int] { method coerce:<to> () { ./elems } } 18:59
very clean.
Aankhen`` Yup.
autrijus then, (@array as Int) follows directly from that
Aankhen`` Yay. 19:00
autrijus and hence ('foo' x @array) too.
now, the tricky part is to implement the dependent types ;)
(note how the ::T in coerce:<as> unifies with a _value_)
Aankhen`` What dependent types?
autrijus the (Type ::T) part.
Aankhen`` Ah, OK.
autrijus summons the power of F-omega-sub. 19:01
dependent types means you can write weird things like
($x as ref($y))
or even 19:02
($a as (rand(2)>1 ?? Int :: Str))
and expect it to work.
wolverian autrijus, wow, ::T can come after it's used in the signature?
Aankhen`` Coolness!
Right, I'm taking a break.
BBL, maybe. &
autrijus wolverian: yes, all :: bindings unify first into common supertype
regardless of position
brentdax Is ./ back again?
autrijus then is applied to restrict the non-:: form.
wolverian autrijus, right. greatness. autrijus++
autrijus wolverian: thank you, I do think that's kind of an innovation :) 19:03
svnbot6 r6145 | Aankhen++ | r6175@tef-base: Aankhen | 2005-08-10 00:25:07 +051800
r6145 | Aankhen++ | HTTP::Message:
r6145 | Aankhen++ | * added explicit invocant wherever appropriate.
r6145 | Aankhen++ | * changed `coerce:<as>` from method to sub.
r6146 | Aankhen++ | r6176@tef-base: Aankhen | 2005-08-10 00:26:43 +051800
r6146 | Aankhen++ | * changed `require` to `use` wherever classes were being imported in HTTP::Message, HTTP::Request, HTTP::Request::CGI and HTTP::Response.
autrijus brentdax: ./ is never left.
brentdax Ugh...never been a fan of it. 19:04
autrijus right.
wolverian hm, how do you declare coercion into multiple things in one class?
autrijus but it's still staying for now, as larry said it's okay to have implicit "use self './'" before we can implement self.pm 19:05
(which is kind of hard)
once that happens then you need that to use ./
good enough?
wolverian:
class MyClass does Coerce[Int] does Coerce[Str] {
multi method coerce:<to> (--> Int) { ./as_int } 19:06
multi method coerce:<to> (--> Str) { ./as_str }
}
wolverian hm, right. that's clean.
autrijus very.
and because of -->'s contravariance
if you can coerce to Int without defining coerce to Num
wolverian somehow the coerce:<to> looks like I should be writing coerce:<Int> etc
autrijus then pugs picks up Int anyway.
brentdax Yeah, that's certainly alright.
autrijus wolverian: yeah, but that'd be name, sadly 19:07
(picks up Int anyway when you demand Num that is)
integral will perl6 produce chains of coercions? $a as B =====> ($a as C) as B ?
brentdax Can you say `has Int multi method coerce:<to>() { ./as_int }`? Or is there no counterpart to my/our for methods? 19:08
wolverian autrijus, so why not just call it 'coerce'? what does the :<to> give us?
autrijus integral: I think that's generally a halting problem
integral autrijus: thank goodness :) that's one icky corner of c++
wolverian (I'm obviously a bit confused, sorry.)
autrijus integral: right. I don't think that's worth it.
wolverian: the coerce: is a grammatical category
(not my invention; see A12)
wolverian right, I'll reread it now. thanks. :) 19:09
autrijus things in that category may appear freely in infix position to add annotations.
so hm
wolverian oh, so you'd write '$foo to Int'?
autrijus no,
$foo as Int
so coerce:<to> is bogus
maybe just coerce_to then 19:10
wolverian now I'm even more confused. I'll just read A12 now. :)
autrijus wolverian: A12 just talk about that syntax.
I can explain :)
the &coerce:<as> function takes a thing and a type
and cast the thing to that type 19:11
provided that the thing does the Coerce role with that type as role parameter.
wolverian right.
I'm just confused about the <as> thing.
autrijus to implement that role, we need an actual method
oh, the <as>
wolverian i.e. why it's necessary.
autrijus it's just like
&infix:<and>
wolverian so we can write our own coerce operators?
what is the use in that? :)
autrijus actually a lot of. 19:12
I can think of "super" for one.
wolverian oh. well, that's why I'm confused, I can't think of any
autrijus return (f() super Int);
wolverian what does that do? :)
autrijus it ensures that f() returns something that can be used as Int or one of its super types
not terribly useful in that position, now I think about it. 19:13
wolverian whoa. method int () { .. } # what happens in integer context
autrijus yeah, we could have 19:14
subtype IntDuck can int(-->Int);
IntDuck does Coerce[Int] { method coerce () { ./int } } 19:15
wolverian A12 seems to say the int() thing is builtin
autrijus this is all hypothetical and noncanonical btw.
wolverian: sure, *int is builtin, but it better croak on things that can't be intted. 19:16
$ perl -e 'print int(qr//)'
134574816
very useful, that.
Limbic_Region autrijus - so in the example I tried to link to but google gropus can't seem to find - sub foo (Int $bar) { say $bar } foo('hello');
error or no error?
wolverian right, but A12 says $foo.int is called when $foo is used in any integer context, including implicit 19:17
autrijus Limbic_Region: depends on whether there is coerce defined bothways
wolverian unless I'm totally misreading what 'integer context' means
autrijus Limbic_Region: and whether coercion are implemented to raise warnings.
Limbic_Region and if error, would the answer change if it was foo('42answer');
ok - well, according to Luke in that thread - it is supposed to squawk
autrijus Limbic_Region: in hackathon lwall ruled that coercion will happen regardless, but '42' and 'Inf' won't raise warnings
while '42answer' will
Limbic_Region ahhh
wolverian yay. that's sane. 19:18
autrijus and those warnings can be made fatal.
and you can additionally make all such coercion fatal.
Limbic_Region hackathon trumps that thread
autrijus so '42' will also croak
but that's not by default.
yes.
wolverian: yeah, very sane and easy to implement also.
wolverian so, uh. if we have method int to define coercion into int, why do we have coerce:<as> at all?
autrijus (I should know because I pushed for the most implementable solution)
wolverian or is A12 just behind the times?
autrijus wolverian: because you still say
('asd' as Int)
Limbic_Region I still want to be able to have coercion be fatal or raise warnings if I have 1 or more pragmas enabled
autrijus that form must be legal 19:19
now it may call int() behind the scene
or other way around
the important thing is just the user needs to have access to both.
wolverian hm. right. so A12 is just unclear here, not strictly wrong
autrijus wolverian: the "IntDuck" above gives out coerce:<as> to anything that definds .int
Limbic_Region: right, you _will_ get those.
I want those for the perl6 compiler written in perl6.
wolverian has there been discussion about renaming grep and/or map? I seem to remember something like that on p6l 19:20
just wondering if I should bring it up (again)
autrijus I don't think I see those confirmed 19:21
I think map is fine
Limbic_Region I (incorrectly) thought originally, there would be 3 form of declarations for each type - int = looks like an int but will gladly autoconvert, Int = int and will autoconvert but will forget it ever happened (no caching of conversion), INT - int and if you try and use it as any other type I will die
wolverian right. I don't like grep.
autrijus Limbic_Region: the INT form is archaic.
and obsolete
int is unboxed (i.e. not an object, but will autobox)
Int is boxed (an object, can have other traits)
wolverian although even .map could be renamed to .filter or so.
autrijus that's all
wolverian I don't know if it'd be clearer though. 19:22
autrijus wolverian: .map is not .filter.
.grep is maybe .filter
but not .map.
wolverian .grep is more like .collect
I think.
autrijus but in haskell, filter means grep
I think python too
wolverian right. okay, map isn't filter then. :)
autrijus to have filter mean map will be weird.
now .for_each is fine ;)
or even .each
wolverian isn't that taken already? 19:23
autrijus eh?
wolverian oh, I was thinking of hashes
autrijus by ruby maybe
wolverian which makes me wonder if we have grep/map for hashes?
%foo.map(-> $k, $v { ... })
I guess it'd be %foo.=kv.map(...)? 19:24
does that chaining work? :)
autrijus yeah.
sure does.
wolverian yay!
btw - I don't like 'filter' because it's a noun as well as a verb
'collect' is a verb 19:25
but I don't know if that's much of a design rule in perl6. hm.
eric256 lol. i take a couple days off from pugscode.org to go buy a house and once agian there are like 8 posts to read. lol. the whole week before that only saw like 2 posts...guess i just have good timeing ;) 19:34
wolverian 8 isn't much at all. there have been 40 new messages in a few hours on p6l at times. 19:37
eric256 very true.
but this are pretty big ones ;)
renormalist Hi all! Does anybody know whether there is already a (maybe preliminary) perl6-mode for (X)Emacs? 19:54
Limbic_Region I know there is for vim 19:55
gaal_ there's a vim mode in util/perl6.vim
but i don't know of one for emacs
svnbot6 r6147 | iblech++ | * Usual svn props.
r6147 | iblech++ | * PIL2JS: Continuation Passing Style landed.
r6147 | iblech++ | * General idea: foo(3); bar(baz()) is rewritten as
r6147 | iblech++ | foo(3, -> () { baz(-> $arg0 { bar($arg0, {}) }) }).
r6147 | iblech++ | * Note: Most things (even &say) do not work yet again, will need work on the
r6147 | iblech++ | JS parts of Prelude::JS. PIL2JS.js is already converted to CPS.
r6147 | iblech++ | * return(...) respects scope now.
r6147 | iblech++ | * But &next, &last, etc. do not work yet again.
Limbic_Region autrijus - sorry, had to step afk
ingy___ hola
Limbic_Region the incorrect memory is just that - incorrect 19:56
OTOH, it would be nice to say int, Int, and INT each with different behaviors
renormalist k 19:57
svnbot6 r6148 | iblech++ | PIL2JS: 20:02
r6148 | iblech++ | * PIL: Removed dead code (&PIL::generic_catch).
r6148 | iblech++ | * Prelude::JS: Fixed some "calls" to &JS::inline so the resulting JS parses again.
autrijus Limbic_Region: then I can't name my class C3 20:04
stevan will be unhappy
and think about those chinese class names with no upper case forms! 20:05
:)
Limbic_Region err - I am not overly particular about the naming convention 20:06
autrijus iblech: whoa, CPS landed.
iblech: did you see my extra-sane coerce?
role Coerce[::to] {
method coerce (--> ::to) { ... }
}
multi *coerce:<as> ((
Coerceable[T] $from
Limbic_Region I just mean the ability to tell the compiler I mean int and if I ever use it as something other than that - hit me hard
autrijus : Type ::T
) --> T) { $from.coerce }
iblech autrijus: Yep, but absolutely nothing works, fixing :)
Indeed, very sane 20:07
Limbic_Region or, I mostly intend to use this as an int, but please forgive me if I don't - I understand I will pay a performance penalty when you have to bail me out - and that is ok
autrijus Limbic_Region: right, in both cases you use int
and choose the strictness level of coercion.
Limbic_Region and finally a mode to say - I don't know what I want or how I will use it so please bail me out and remember what you did so you can be faster about it next time
autrijus since int->Int is considered coercion.
or maybe we call it boxing 20:08
use strict 'boxing'
Limbic_Region autrijus - is the strictness level of coercion done on a per-variable declaration basis?
autrijus Limbic_Region: no, it's based on scope
Limbic_Region cause if not - I don't llike it
autrijus nod.
Limbic_Region so it is still all or none, but only for a given scope
*shrug* 20:09
autrijus you are certainly free to declare
Limbic_Region doesn't really have to like it
autrijus my int_rigid $x;
and implement int_ridig.pm by augmenting the int class
but disable its coercion bits.
which is not unlike Tie::VariousRestrictedThings in perl5. 20:10
except much faster.
Limbic_Region well - the only point in doing that is if it made things go faster - which is yet to be seen in p6
right
autrijus right, and if you use int then things will be faster whenever it can.
and even if you use Int
but without extra traits
Limbic_Region in p5 changing behavior outside of core meant performance penalties
autrijus it is possible to automagically unbox it too. 20:11
right, and in p6 the penalty is only paid in compile time.
Limbic_Region in p6 - time will tell
autrijus which hopefully isn't much.
yeah, time will tell :)
Limbic_Region you are more confident then I am on that
autrijus sure, because I'm the one writing the type directed compiler :)
Limbic_Region well - it isn't just types though 20:12
in this case it is
but in general, people who want to modify the behavior of how p5 works pay a penalty in performance
in general, p6 aims to allow morphing behavior AND still be fast
that's the part, IMO, that remains to be seen 20:13
autrijus sure, but currently 20:14
user-def operators, macros, itypes etc
are all compile time features
the runtime knows nothing about it
so it's hard to imagine it will make runtime slow.
journal finally up. 20:19
g'nite!
Limbic_Region autrijus - well, I am not trying to be a pessimist here
autrijus waves &
Limbic_Region TTFN
but
autrijus hm?
Limbic_Region history has shown that things shoehorned in after the fact lead to original designs being comprimised 20:20
autrijus right.
so, Better is Better, this time.
&
Limbic_Region stands behind his "time will tell" philosophy
stevan is actually enjoying responding to TSa, the man asks good questions :) 20:57
kolibrie stevan: you've been giving good answers 20:59
stevan kolibrie: thanks :) 21:00
this is my sanity break for the day, I have been on conf calls all day long for $work
metamodel stuff is a breath of fresh air :)
kolibrie I'm still trying to understand the whole thing
I've got a paper with lots of notes that I want to digitize somehow 21:01
and then ask you questions
stevan kolibrie: it took me about 2 months and about 5-10 prototypes (all false starts) to get it
kolibrie :0
I don't feel too bad, I've only been looking at it for a week or two
stevan but I didn't have "The Art of MOP" book either, which has been a big help 21:02
kolibrie don't have time for another book right now :(
later all 21:06
stevan later
fglock hi 21:10
eric256 hola
fglock hola eric256 21:11
eric256 tries to figure out the best way to see any new featuers in pugs that he will notice 21:12
ingy___ hola
eric256 seems like allot of thoeritcal work has been done. gonna have to check the svn log to see if any concrete everyday, things i'm going to notice, things have been done ;)
ingy___ has anyone here installed spidermonkey on osx?
obra hey ingy 21:13
ingy___ hi obramatic
obra how's stl?
ingy___ just fine :)
fglock eric256: i've worked on Array-Lazy, but it is not integrated into pugs - it's just a module now
eric256 hmmm lazy arrays...anything with objects been happing on the codeing level? i know alot has been brewing 21:14
fglock I think i've got to rename a few modules, like Array::Lazy to Perl6::Container::Array
eric256: it's already implemented, but integrating into the core is hard 21:15
eric256 implemented in pugs? 21:16
fglock yes, but it is just an object - you can't use "@lazy[2]" yet - only $lazy.FETCH(2) 21:17
eric256 i think we are talking about different things. i mean the object model stuff not the Lazy stuff 21:18
though i want to look into the lazy stuff too
fglock the object model is the priority right now - but i'm interested in the lazy stuff 21:19
so I started working on it 21:20
eric256 i noticed ;)
fglock :)
anyone knows how the file hierarchy in pugs works? i'm trying to find out where to place a file 21:22
how do I call a sub in the main namespace, like CORE:: in Perl5? 21:36
integral &*global? if by "main" you mean globals 21:37
obra fglock: is the current DateTime::Span on CPAN expected to pass all its tests?
fglock no - it depends on DateTime, which is not finished yet :( 21:38
oops - I thought you asked about perl6 version - yes, the CPAN version should pass 21:39
obra I ran into a circular dependency issue where installing DateTime tried to pull in DateTime::Span. which failed tests :/
ahh. ok.
I'm seeing failures in 0.22
t/15time_zone.t 5 1280 14 5 35.71% 7 10 12-14
fglock DateTime should not depend on DateTime::Span - strange
obra I bet it was some other module in a dependency chain. 21:40
What can I do to help debug those failures?
fglock i'm looking into it, just a moment
obra cool.
I don't mean to derail you from what you're working on, but I saw your nick right as I saw the failures ;) 21:41
fglock which t/time_zone.t? does it have a number on the name?
obra t/15time_zone.t 5 1280 14 5 35.71% 7 10 12-14 21:43
15
fglock testing... 21:45
obra - it passed all tests with my old DateTime.pm installation. It broke when I updated it. I'll check what's happening. 21:49
obra Ok. cool. I'm glad to hear that it's replicatable
svnbot6 r6149 | iblech++ | PIL2JS: 21:53
r6149 | iblech++ | * pil2js.pl: Load stevan++'s new Perl6.MultiMethod.
r6149 | iblech++ | * More steps towards a working PIL2JS.
fglock obra - It looks like the error is in the test itself - i think it depends on a timezone that has changed 21:55
obra *laugh* oy. 21:56
fglock and I can't login to sourceforge to fix it - it can't connect 21:57
it's in my list for tomorrow 21:58
21:58 wilx` is now known as wilx
obra thanks much. 21:59
If we're ever in the same place, I owe you a beer or equivalent. As much for the craziness that was reefknot as this
fglock you were in reefknot? 22:00
i guess i find craziness atractive 22:01
obra fglock: I came in relatively late. I spent time with srl working to untangle a lot of the code so mortals could hack on it. 22:02
fglock i tried to call global::push, but it didn't work - is there another syntax?
obra: it would be nice to have the calendar server working - a lot of people have asked for one 22:03
obra fglock: it sure would.
Right now, there are a lot of opensource efforts to create calendar servers.
Sadly, there are also like six new calendar protocols out there
fglock (people complain a lot about Set::Infinite code - the perl6 version is much more readable) 22:04
obra fglock: yeah. autrijus was showing me today.
Set::Infinite is deep magic for perl 5
and part of the problem I ran into at the time was trying to do infinite recurrences stored in the database and fast 22:05
fglock i was able to store recurrences in the database - but it was very slow
obra *nod*
fglock it can't optimize the queries - they get exponentially slow 22:06
obra yep.
doing infinite set algebra in perl/sql is...well, I haven't seen it done.
fglock take a look at ext/Recurrence - it is pretty short for what it does 22:07
i think it is doable in perl/sql - i was trying to do it in pure sql :) 22:08
obra heh 22:09
doing it fast in perl/sql?
fglock the problem i got with sql is that it can't do binary searches inside a select, and this means that if i had hour,minutes,seconds it would take 24*60*60 operations 22:10
obra I keep finding that I'm happier storing epoch time in the database 22:11
fglock the problem with perl/sql is that it would take cpu time on the database server - people don't want that 22:12
(i think)
mugwump I use the native date format, but never rely on server-side date calc
fglock i'm still trying to find out what's the namespace for 'push()' 22:14
brentdax I'm actually storing Perl 6's time() format in Kontent store databases--it's fairly handy. 22:16
buu brentdax: What is a 'kontent store database' ?
brentdax (But then, I'm also avoiding complicated queries.)
obra perl6 time is still seconds since y2k?
brentdax Floating-point seconds since Y2K, yes. 22:17
buu: Kontent is the Perl 6-based CMS I'm writing for Summer of Code. A store database is where it keeps its pages.
buu brentdax: Oh. Is it any good?
brentdax Too slow to be usable, but it just needs a faster interpreter. Once it's fast enough, I think it'll be pretty nice. 22:18
buu What are the advantages? 22:19
eric256 your getting paid to play with Perl6? man i missed that boat 22:21
hay perl6 time and perl5 time arn't going to play togther?
obra eric256: I'm sure there will be magic to make them deal well. and enough rope to hurt yourself too
brentdax It intrinsically supports multiple page types, so you can keep (say) a wiki and a message board on the same system. It can render the same page into multiple formats. Supports multiple stores (I'm writing a DBI one) and multiple web server APIs (writing a CGI one). And it's cleaner than comparable systems, like Everything and MediaWiki. 22:22
Khisanth obra: well Perl wouldn't be Perl if it didn't give you just enough rope to hang yourself :)
obra brentdax: how does it compare to bric? 22:23
brentdax Somewhat different emphasis--Kontent is largely intended for open-contribution sites (wikis, message boards, Slash/Scoop news sites, PerlMonks-ish sites), where it looks like Bricolage is mainly for a restricted group of admins. I think either could do anything the other can with some work, though. 22:28
svnbot6 r6150 | fglock++ | * Renamed Lazy::List to Perl6::Value::List 22:29
r6150 | fglock++ | * Renamed Array::Lazy to Perl6::Container::Array
brentdax I'm not sure if Bric creates static pages, but Kontent doesn't--it's all dynamic. (And no doubt I'll need to add in lots of caching.) 22:30
(Yeah, Bric creates static HTML.) 22:31
svnbot6 r6151 | iblech++ | PIL2JS: 22:32
r6151 | iblech++ | * PIL2JS.js, Prelude::JS{,::*}: Fixed hopefully all subs to call the cc.
r6151 | iblech++ | * Good news: "say "hi"" runs again!
r6151 | iblech++ | * Bad news: It produces garbled output. Fixing.
buu brentdax: So will you have to create your site in side your framework, or will it be just a couple of site types? 22:33
brentdax It pretty much assumes you're keeping your entire site in it. 22:34
buu brentdax: For example, how would you set up an Everything type system? 22:35
brentdax After installing the system (for the CGI supervisor, this involves saying "Alias / /path/to/supervisor/cgi.p6"), you would arrange for new pages to have the Kompose (owned page) class, probably by creating a /pages page with that class and having all of your pages under that. You would then grant the create permission on /users and /pages to everybody, customize a few settings like the... 22:41
...template and site name, and let 'em rip.
svnbot6 r6152 | fglock++ | * Array::Lazy TODO update
renormalist stupid question (my first experiments with p6regex): Is =~ still the matching operator? 22:44
brentdax Nope, ~~ now.
renormalist k, moment
ah, ok. ~~ works 22:45
(of course ) 22:46
eric256 is there a site for Kontent /
?
brentdax Not yet--just the 0.01 distribution on CPAN and the Subversion repository on my server. 22:47
iblech [PIL2JS] Yay, say "hi" works again :))
eric256 is it still mostly production? because i've been looking for something like that for our work portal, and i'll either have to get to work rewriting mine (written when i was too green to even salvage now) or find a goood one to use ;) 22:48
brentdax It's nowhere near production quality--the current SVN version even has a known remote-code-execution bug (which won't be closed until I implement users and security.) 22:49
And it's horribly slow because Pugs is unoptimized.
eric256 hehe.../me heads off to start his own design then. ;)
mine would need less bells and whistles anyway. 22:50
brentdax Probably. This thing has quite a few, including its own wikiish markup language.
svnbot6 r6153 | iblech++ | PIL2JS: Made say "hi" work again (and probably most other things, too). 23:03
r6153 | iblech++ | *Problem*: PIL2JS runs into "too much recursion" errors *really fast*, as
r6153 | iblech++ | PIL2JS never JS-return()s from a function, but instead simply calls the escape
r6153 | iblech++ | continuation. I.e. JS thinks we're still in 1000 functions and dies...
chip Has Larry posted about sets yet? 23:10
iblech No 23:16
renormalist Question about p6regexes/rules: Should "subpatterns" in Pugs already work? I don't get stuff like <$subrule> to work but I'm not sure whether they already should. 23:24
I try sth. like this: 23:25
# - Variant 1 --------------------------------------------------
my $PROGRAM = rx{^ \s* START \s* (affe|loewe|giraffe) \s* \. $};
# - Variant 2 --------------------------------------------------
#my $KEYWORD = rx{(affe|loewe|giraffe)};
#my $PROGRAM = rx{^ \s* START \s* <$keyword> \s* \. $};
# --------------------------------------------------------------
# - match ------------------------------------------------------
my $str = 'START loewe .';
say "ok" if $str ~~ $PROGRAM;
variant1 works, variant 2 not. 23:26
brentdax I don't think they do, but even if they did, they'd be case-sensitive. 23:28
renormalist ah, damn
hm, ok, but wasn't the bug
Khisanth hmm 23:51

Powered by Cro and the Raku® Programming Language.

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

Thank you!