Pugs t-shirts www.cafepress.com/pugscode | Pugs 6.2.9 released | pugscode.org | pugs.kwiki.org | paste: sial.org/pbot/perl6 | www.geeksunite.net
Set by stevan on 15 August 2005.
stevan autrijus: ping 01:02
autrijus: I assume you backlog,.. so I will tell you 01:06
I got singleton methods working
see t/36_singleton_methods.t
I found out how Ruby does it,.. which is insert an anon-class in between the regular class an the instance 01:07
so you have this [Object] -> [Foo] -> [anon class for $foo]-> $foo
then all the singleton methods get stashed into the anon class
since it is specific to $foo only,.. it works
I did this with a subclass of $::Class 01:08
:)
ok... time to watch some TV &
svnbot6 r7225 | stevan++ | Perl6::MetaModel 2.0 - 01:10
r7225 | stevan++ | * fixed the test to create *proper* singleton methods (ala Ruby)
r7225 | stevan++ | basically we just insert an instance specific anon-class into
r7225 | stevan++ | the mix at the right time, and whalla you have singleton methods.
r7225 | stevan++ | I implemented this by making a subclass of $::Class, however it
r7225 | stevan++ | might make sense to actually integrate this into the core of the
r7225 | stevan++ | metamodel itself. I would certainly make class methods easier, and
r7225 | stevan++ | allow for some very interesting assorted evil tricks :) 01:11
brentdax ?eval class X { submethod BEGIN { say "X" } } class Y is X { submethod BEGIN { say "Y" } } Y.new 06:19
evalbot_7225 Y.new();
brentdax ?eval class X { submethod BEGIN { say "X" } } class Y is X { submethod BEGIN { say "Y" } } Y.new()
evalbot_7225 Y.new();
brentdax ?eval class X { submethod BEGIN() { say "X" } }; class Y is X { submethod BEGIN() { say "Y" } }; Y.new() 06:20
evalbot_7225 Y.new();
Khisanth heh
brentdax evalbot is not being very helpful.
Khisanth ?eval class X { submethod BEGIN() { say "X" } } class Y is X { submethod BEGIN() { say "Y" } } Y.new() 06:21
evalbot_7225 Y.new();
Khisanth ?eval class X { submethod BEGIN() { say "X" } } class Y is X { submethod BEGIN() { say "Y" } } my $y = Y.new()
evalbot_7225 \Y.new();
Khisanth well ...
brentdax Even when I take X out of the equation, Y still doesn't print. This is odd. 06:22
...oh, wait, I'm an idiot.
?eval class X { submethod BUILD { say "X" } } class Y is X { submethod BUILD { say "Y" } } Y.new
evalbot_7225 X Y Y.new();
brentdax Okay. Doesn't explain why actions being taken by X's (or Cipher's) constructor aren't registering, but it's a start. 06:23
Er...except that I made the same mistake there, too. Heh. 06:24
nothingmuch glorf 06:42
svnbot6 r7226 | brentdax++ | Cipher API:
r7226 | brentdax++ | * Full testing of Cipher.pm (including new test cipher)
r7226 | brentdax++ | * Documentation for Cipher::Caesar
brentdax I actually have copies of Cipher::Stream and Cipher::Arcfour (RC4) locally, but I haven't tested either of them. 06:43
nothingmuch check in 06:46
we can submit some tests
the most basic test for all of them is round trip
and that the thing at the middle of the trip ne the thing before and after it 06:47
and that middle of trip is ne middle of trip when key is different
brentdax: ? 06:58
www.cypherspace.org/adam/rsa/rc4.html 07:04
brentdax I'd at least like to get some test vectors pulled together first. Testing it is pretty easy--I just have to do it. 07:12
nothingmuch i'm willing to help if you like 07:17
svnbot6 r7227 | nothingmuch++ | test for optional trailing semicolon on blockish exprs. 07:25
brentdax ?eval class X {} my %options = (:foo<bar>); X.new(*%options) 07:46
evalbot_7227 Error: Must only use named arguments to new() constructor
brentdax curses quite a bit. 07:47
?eval class X {} my %options = (:foo<bar>); X.new(%options.pairs)
evalbot_7227 Error: Must only use named arguments to new() constructor
brentdax ?eval class X {} my %options = (:foo<bar>); X.new(map { $_ } %options.pairs)
evalbot_7227 Error: Must only use named arguments to new() constructor
brentdax That makes it rather hard to pass options from the various wrapper routines into the constructor. 07:48
Guess I'll just test with the OO core. 07:51
Gah, I'll commit it anyway. Stupid OO. 07:58
nothingmuch hehe 07:59
nothingmuch wants luqui's tuples to make it in
svnbot6 r7228 | brentdax++ | Cipher API: 08:01
r7228 | brentdax++ | * Undocumented Cipher::Stream
r7228 | brentdax++ | * Undocumented Cipher::Arcfour (RC4)
r7228 | brentdax++ | * arcfour.t, which fails the second test for mysterious reasons
r7228 | brentdax++ | Ideas on what's causing arcfour.t to fail would be appreciated.
nothingmuch theorbtw1: long time no see! 08:14
theorbtw1 Allo, nothingmuch! 08:15
nothingmuch what's up? 08:16
theorbtw1 Been working on a contest entry, and ended up possibly missing the entry deadline by 40 minutes because the terms weren't clearly written. 08:19
08:20 theorbtw1 is now known as theorbtwo
kgftr|konobi theorbtwo: oioi 08:21
theorbtwo Yeah.
Well, it'll still get used, if only by Jess and I.
kgftr|konobi theorbtwo: jess on her way or still feeling under the weather?
theorbtwo Sorry I wasn't at the pub on Friday; Jess wasn't feeling well (and is staying home sick again today).
kgftr|konobi ah... hokey dokey... no probs! perhaps next time? 08:22
theorbtwo Next time indeed.
nothingmuch bah
nothingmuch backlogs
theorbtwo I should probably finish the "create a new user" page before I point you at it... 08:23
nothingmuch what is the contest?
theorbtwo: ? 08:31
kgftr|konobi theorbtwo: damn prvmsgs are blocked... can you ask jess to ring anthony to confirm sick, etc. 08:38
spinclad ?eval $?PUGS_VERSION 08:44
evalbot_7228 \'Perl6 User\'s Golfing System, version 6.2.9, August 3, 2005 (r7228)'
spinclad ?eval class X {} my %options = (:foo<bar>); X.new(*(map { $_ } %options.pairs))
evalbot_7228 Error: Must only use named arguments to new() constructor
spinclad ?eval class X {} X.new(:foo<bar>) 08:45
evalbot_7228 X.new(('foo' => 'bar'));
dada ?eval class X {} my %options = (:foo<bar>); X.new(%options) 08:48
evalbot_7228 Error: Must only use named arguments to new() constructor
dada mmm
that's unorthogonal
theorbtwo backstage.bbc.co.uk. 08:49
I created a TV schedule grid that shows customized ratings -- it tells you what's on that it thinks you'd like. 08:50
nothingmuch hola castaway_!
castaway_ kgftr|konobi: I already called HR, doesnt that cover it? Else, whats his number? 08:51
nothingmuch wow, attribute grammars are fun 08:53
now lets see if they actually work ;-)
castaway_ nudges Scott 08:54
svnbot6 r7229 | brentdax++ | Cipher API:
r7229 | brentdax++ | * Documentation for Cipher::Arcfour.
scook0 gah, goddamn tests!
I spend half an hour getting the damn test script to run 08:55
and then it pretends my test statements don't even exist!
(and the test takes about a minute to load before doing nothing!) 08:56
nothingmuch blah 08:58
scook0 well, hopefully it'll be worth all the trouble
! 08:59
svnbot6 r7230 | iblech++ | * Usual svn props. 09:00
r7230 | iblech++ | * System.FilePath: Added a not (null fname) guard (i.e., fixed a bug).
r7230 | iblech++ | We should probably propagate the fix upstream (at least when ok'd by a real
r7230 | iblech++ | lamda hacker).
r7230 | iblech++ | * Main: Added %?CONFIG<sourcedir> to the list of paths to search and
r7230 | iblech++ | added a tryIO call so the search doesn't die.
nothingmuch struggles with his pretty printer 09:01
kgftr|konobi castaway_: ah... that should be fine... will let anthony know 09:02
castaway_ ok good 09:03
castaway_ passes around the Sinupret
nothingmuch luqui: !!!!!!!!!!!! 09:04
i've been waiting for ever
luqui hi nothingmuch
nothingmuch fix your svn repo
luqui yeah, I've had my nose in school
it is fixed
castaway_ wanders again
nothingmuch also, i need to do some funky stuff with L::AG
luqui (it dies every once in a while ??)
nothingmuch i have a test for threaded attributes
luqui nothingmuch, like?
Oh, I figured out a new lazy algorithm that could be much better than the one I am using memory-wise
I have to do some analysis though
nothingmuch and also i have e.g. type::operator::arrow isa type::operator::binary 09:05
luqui hates threaded attributes
do it without punning if you can
type::operator::arrow?
svnbot6 r7231 | iblech++ | util/smokeserv/smokeserv-server.pl: Always set $*ENV<TZ> to "UTC" so the
r7231 | iblech++ | template ("Timezone is UTC") doesn't lie.
nothingmuch i would like 'to say operator::binary: layout($$) = { join(" ", layout($.left), $.symbol, layout($.right)) }'
and have that work on type::operator::arrow and type::operator::pair 09:06
luqui ahh yes
you mean it should respect inheritance?
pasteling "nothingmuch" at 82.81.247.180 pasted "an AG for pretty printing types" (40 lines, 1.2K) at sial.org/pbot/13424
nothingmuch also yes
if there is no attr defined on blessed($node) it should iterate Class::ISA::super_path
i guess that C3 order is a good choice 09:07
anyway, that's why i need threaded attributes
see free_in and count_in?
well, admittedly "need" is a strong word
luqui yeah, you can always make a threaded attribute with a inherited+synthesized attribute of different names 09:08
threaded is just a pun making two different attributes have the same name
nothingmuch but it's tres cool ;-)
anyway, that grammar is dying, and i can't figure out why
Nonlinear attribute: you have two or more ways to assign a value 09:09
to the attribute 'layout' near grammar line 11.
but i have no attrs that are nonlinear, IMHO
this is about variable::free: layout($$) = { ... }
anyway, new tests submitted to L::AG
you're welcome to mark t/threaded.t as TODO ;-)
luqui yeah, I saw that, thanks
nothingmuch wait... conflicts 09:10
luqui did a little renaming around, and then unrenaming, and yucks
I don't think I'm going to use Kasten's ordered approach. Lazy is just too much easier.
er, Kastens's
nothingmuch how do you get 3-way merge with svk? 09:11
luqui ??
nothingmuch i want the conflict to go to FileMerge.app
i don't know what command I want =)
luqui nothingmuch, this grammar is dying with the CPAN version of L::AG (ie. you haven't made any changes?)
nothingmuch this is with the svn version, but i have not edit the .pm file
s/have/did/
luqui hmm okay. That might be a bug. 09:12
let me look at it
nothingmuch okay
luqui even though I'm about to rewrite the .pm
oh, do you have a file with a test for this grammar? 09:13
nothingmuch not yet
i'm trying to get my earlier tests in
struggling with svk
luqui no, I mean, you said the grammar was dying
on what input tree?
pasteling "nothingmuch" at 82.81.247.180 pasted "the input tree" (10 lines, 430B) at sial.org/pbot/13425 09:14
nothingmuch oh shit, i think i get it 09:15
nothingmuch confused himself in the constructors
pasteling "nothingmuch" at 82.81.247.180 pasted "the correct input tree, generates same error" (9 lines, 398B) at sial.org/pbot/13426 09:16
"scook0" at 220.253.106.5 pasted "What the heck is wrong with my test?" (52 lines, 1.1K) at sial.org/pbot/13427
nothingmuch luqui: any idea what's going on? 09:22
luqui hmmm...
nothingmuch i'll post the full script
luqui well, it is installing layout of that thing twice
I'm trying to figure out why
pasteling "nothingmuch" at 82.81.247.180 pasted "type pretty printer & misc" (193 lines, 3.5K) at sial.org/pbot/13428 09:23
nothingmuch just run perl on that thing
luqui are you using "root" as the meta root, or as type::root? 09:24
nothingmuch no, type::root
luqui okay good 09:25
nothingmuch shouldn't that be ROOT: ?
(for pseudo-root)
luqui yeah it should
in other words, you did it right for what you were trying to do
luqui just realized that ROOT and -prefix don't mix
huh, the root is being scanned twice 09:28
interesting, that shouldn't happen
(and would explain the error)
nothingmuch hrmm
luqui is on the trail though 09:29
nothingmuch oh, also have a look at t/errors.t later
it tests for current behavior
i'm not 100% sure it's the right behavior
luqui got it 09:30
count_in is an inherited attribute
nothingmuch and also, in basic.t one test tries to assign the same attr to ROOT and Foo, where Foo is the root anyway
luqui but it doesn't have anybody to inherit it from
nothingmuch right now ROOT overrides
oh crap
=(
luqui (the new structure with declared unions ought to make finding errors like that easier) 09:31
nothingmuch root: count_in($.type) = { 0 } is not helping =(
luqui do you have a root node in your structure?
nothingmuch $pretty->apply(bless { type => $self }, "type::root")->layout; 09:32
}
luqui tries
uh, type => $self?
nothingmuch yep
luqui what's $self?
nothingmuch type::operator::arrow->new(type::variable::free->new, type::variable::free->new);
luqui oh, I see
nothingmuch pseudoroot is not good enough for root: layout($$), right? 09:33
luqui I think a real root was a good choice in this case
nothingmuch counting works 09:34
if i root: cont_out($$) = { count_out($.type) } then I get back 2
free_out doesn't work 09:35
i'll investigate
luqui yikes. need much more error checking code 09:36
free_in is an inherited attribute
nothingmuch right
luqui but operator::binary does not set free_in of its child
ren
nothingmuch right...
luqui but it should 09:37
nothingmuch fixing
luqui wants automatic attr generation now
I'll start the rewrite
nothingmuch yay! 09:38
luqui works?
nothingmuch luqui++
not yet, i'm rewriting the stuff
yay for the rewrite
luqui ah
luqui wishes he didn't have homework 09:39
nothingmuch woot! 09:40
# got: 'āˆ€Ļ‰. Ļ‰ ā†’ Ī±'
# expected: 'āˆ€Ī±. āˆ€Ī². Ī± ā†’ Ī²'
not *exactly* it, but the AG side is working =)
luqui oohh pretty 09:41
nothingmuch blammo!
ok 1
i have an elegance issue 09:42
dada smells APL
nothingmuch operator::binary: free_in($.left) = { Set::Object->new }
operator::binary: free_in($.right) = { Set::Object->new }
operator::binary: free_out($$) = { free_out($.left) + free_in($$) + free_out($.right) }
or
or thread the free stuff
so that free_in($.left) = { free_in($$) }
free_in($.right) = { free_out($.left) }
free_out($$) = { free_out($.right) }
which is cooler? 09:43
luqui well the latter doesn't do any set operations at all
nothingmuch right
luqui so I would probably like that better, as it's more general 09:44
without understanding your problem very well, of course
nothingmuch hmm
right
looking up all the free variables
luqui++; # thanks
pasteling "nothingmuch" at 82.81.247.180 pasted "working test script" (207 lines, 3.9K) at sial.org/pbot/13429 09:46
luqui ponders "multi attribute"
nothingmuch luqui: don't worry about backwards compatibility
luqui won't
nothingmuch i'll be glad to upgrade my code
luqui in particular, I'm going to start requiring data structure declarations, I think 09:47
nothingmuch but also try to make it at least semi efficient
since now that I've practiced a little with the pretty printing, i'm going to try to represent type inferrence in AGs
luqui interesting
pasteling "nothingmuch" at 82.81.247.180 pasted "working test script" (207 lines, 3.9K) at sial.org/pbot/13430
nothingmuch shit
sorry everyone
oi vey 09:50
{
my $type = type::operator::arrow->new(
my $a = type::variable::free->new,
my $b = type::variable::free->new,
);
$a->instantiate($b);
is($type->stringify, "āˆ€Ī±. Ī± ā†’ Ī±"); 09:51
}
this is problematic since $b exists both under $type and under $a now
Nonlinear attribute: you have two or more ways to assign a value
to the attribute 'free_in' near grammar line 27.
is there any way to deal with one thing appearing in several places in an AG? 09:52
luqui "one thing appearing in several places"?
what kind of thing
nothingmuch look at 'sub instantiate'
it assigns $b to $a.instance 09:53
and reblesses $a to type::variable::instantiated
luqui okay
... 09:54
nothingmuch the AG module tries to apply free_in($b) once from variable::instantiated and once from type::operator::binary 09:55
luqui oic... 09:56
uh, reblessing is not supported :-)
(however, one could fix it)
by instead of using overload::StrVal everywhere, use refaddr or whatever from Scalar::Util
nothingmuch luqui: reblessing happens wayb the AG gets to eat it 09:57
rafl_ luqui: I prepared a Debian package for Class::MultiMethods::Pure (it will be the first one that I'll upload with the newly created @debian.org account I'll get that week, btw. :-) But I still need a nice text for the description. You said you could provide one to me once..
luqui oh
nothingmuch forget the reblessing
it sets $a.instance to $b.
luqui rafl_, oh, yeah, but that requires work 09:58
luqui is a lazy bastard
bug me on email and I'll email you one back
nothingmuch asked about this on #haskell 09:59
and figured out that true lazy evaluation of AGs should eat this
luqui so what's going on? 10:00
nothingmuch what do you mean/
luqui I never understood your question
nothingmuch well 10:01
we're constructing a type, right?
in a lambda
luqui mkay
nothingmuch fun: type($$) = { type::operator::arrow->new( $a, $b ) }
$a and $b are two free type variables
in the env, the param is bound to $a
so whenever the param is used, $a is used 10:02
now, at some point a deeper operation is going to unify $a with something else
e.g. type::operator::named->new("int")
it does this by instantiating $a to be the new type
luqui by putting it in the env right?
nothingmuch in e.g. Ī»x. x 10:03
$a is unified to $b
to yield the type a -> a
forall a. a -> a
luqui alright
nothingmuch my $type = type::operator::arrow->new(
my $a = type::variable::free->new,
my $b = type::variable::free->new,
);
$a->instantiate($b);
this is the operation i'm trying to do
is($type->stringify, "āˆ€Ī±. Ī± ā†’ Ī±");
this is the result
variable::instantiated: free_in($.instance) = { free_in($$) } 10:04
that is where it blows up
since free_in($b) is set by that
but also by this:
operator::binary: free_in($.right) = { free_out($.left) }
luqui how is that possible?
is this a dag? 10:05
svnbot6 r7232 | iblech++ | * t/operators/operator_overloading.t: s/sub/multi/ in one case
r7232 | iblech++ | * t/packages/{symbols,reflection}.t: =begin obsolete'd the tests,
r7232 | iblech++ | see thread "Packages, Modules and Classes" on p6l started by Stevan:
r7232 | iblech++ | www.nntp.perl.org/group/perl.perl6....uage/23019
r7232 | iblech++ | * t/packages/{autoload,scope,import,require_and_use}.t:
r7232 | iblech++ | skip all tests under PIL2JS and PIL-Run, as the tests need eval().
luqui (me is still a little lost, so I'll ask a bunch of questions)
interesting grammar, luqui
nothingmuch you're going to ask a bunch of questions, and you want to have them answered immediately?
the reason it's possible: 10:06
$type = (a -> b)
$a.instantiate($b);
$type = ($a($b) -> $b)
$type = ($b -> $b)
so far it's a DAG
luqui that's why
dag not supported. If you can think of a way where an AG would make sense on a dag 10:07
I'll try to do it
but you have two ways to set an attribute of a single node. Which do you pick?
nothingmuch if it's evaluated lazliy, the attr of $b is once this, and once that
depending on it's caller
luqui you mean it takes on two values over the course of the evaluation? 10:08
nothingmuch hmm... yes
dynamically scoped
but that's hard
luqui that doesn't sound like a well-defined attribute to me
nothingmuch hmm 10:09
i think i'm slightly screwed
i'll seperate the phase so that the check for frees and allocation of symbols happens before the AG pass
luqui you're not mutating anything over the course of the evaluation though, are you?
nothingmuch no, i'm not 10:10
luqui hmm. yeah, it might be that what you're trying to do is not in AG's domain
nothingmuch *sob*
it was such a pretty grammar 10:11
luqui on the other hand, it could be, and your brain is being foggy about exactly what an attribute should be
but I don't really grok the problem... again.
nothingmuch do you know how milner type inferrence works? 10:12
luqui not in detail
nothingmuch okay
essentially going into the AST you allocate type variables, which are free
you bind type variables to identifiers
going out of the recursion you unify stuff 10:13
luqui by building a "substitution object", right? 10:14
nothingmuch "substitution object"?
luqui a map from variables to types
nothingmuch hmm
i went slightly more OO
yes
but the map is embedded in the vars
so each var is either free or instantiated 10:15
and it can be instantiated to point at another free var
luqui okay, fair enough 10:16
you want to do all this inside an AG?
nothingmuch this part can be in AG
because the AG that infers is not touching the types
i
the AG that comes later, to print the types cannot work with the DAG that is the output of the type inferring AG 10:17
luqui ahh
yeah, you'd have to convolve it into a tree
nothingmuch yep
i'll just do a single pass traversal first
cna you traverse a DAG with synthesized attrs only? 10:18
no, that doesn't make sense either
luqui well, it sorta does
but it would be tough not to use any inherited attributes anywhere 10:19
nothingmuch only for relative calls though
root: layout($$) = {
join(" ",
(map { "?${_}." } sort map { layout($_) } free_out($.type)->members ),
layout($.type),
);
}
that will be undefined
no, it won't, sorry
but anyway, i'll just make a higher order traversal cmd
svnbot6 r7233 | iblech++ | t/unspecced/chop.t: &chop should be non-destructive, exactly as &chomp is. 10:21
r7234 | iblech++ | PIL2JS:
r7234 | iblech++ | * Prelude::JS::Str: Implemented non-destructive chop.
r7234 | iblech++ | * Prelude::JS::Array: Fixed array autovivification ($undef[42] = 23) (two-char patch).
nothingmuch frank zappa is sick
www.lyricsfreak.com/f/frank-zappa/56813.html 10:22
luqui I hope he gets well soon
nothingmuch i think it's way too late
luqui is frightened by those lyric 10:23
s
I need to source filter attributes now 10:26
what would be a good syntax? 10:27
nothingmuch ?
luqui Node: SOMESYNTAX $.foo = { SOMESYNTAX $$ }
it needs to be recognizable, so I can turn it into, eg. $parent->{'child:attr'}->() 10:28
nothingmuch what's bad with the function call interface?
eep
luqui it doesn't work with the new algorithm
nothingmuch ah
i kinda wished for $$.attr
luqui yeah, and I guess it only makes sense anymore to do it on direct children
then we could compute deps to
*too
so we have $$.attr and $.left.attr
that should do it 10:29
nothingmuch yep
and anything that isn't an attr is a method call / hash key?
luqui error?
nothingmuch oh, wait
yes, i see
$$.attr vs $.key
luqui right
nothingmuch what about $.left.right.attr ?
luqui illegal 10:30
nothingmuch okay
svnbot6 r7235 | iblech++ | PIL2JS: Prelude::JS::Array: &push and &unshift autovivificate undefined vars 10:32
r7235 | iblech++ | now, too (this means "my $arrayref; push $arrayref, 42" works now and
r7235 | iblech++ | t/var/autovivification.t passes 19/20).
nothingmuch has a half picture of what AGs could look like as a true language feature in perl 6 10:35
luqui written or mental?
nothingmuch they are just methods
both
not syntax
but structuure
usually they are in a role tailored for a class
but that role is not integrated
instead you do a runtime mixin to get the effect of apply 10:36
that way you can use grammars with conflicting namespaces on the same object
but also use them statically to define a class
attribute resolution within the role uses itself in a fully qualified manner before method resolution 10:37
svnbot6 r7236 | iblech++ | Pugs.Prim: &chop is now non-destructive (&chomp is non-destructive, too -- consistency).
luqui you should probably write something up
nothingmuch so when it is applied to root object, it will apply itself as a role to all the children implicitly
(i'm still brainstorming)
luqui also see: svn.luqui.org/svn/misc/luke/work/co...grammar.pl
for a reasonable implementation strategy 10:38
nothingmuch i haven't read the guts yet
so i can't diff
please explain it a bit
luqui which one?
nothingmuch the reasonable impl strategy
visit creates all the possible attrs as thunks, right? 10:39
luqui you just do a one-pass traversal on the tree (using the "visit" method in this example)
nothingmuch, yeah, you got it. that's about it
nothingmuch what is t() constructing?
luqui that's just debug code
(defined at top) 10:40
nothingmuch ah
oh, i see
you should have used Devel::STDERR::Indent ;-)
luqui heh, yeah, I guess so :-)
hmm, how would you like to see attrs declared? 10:41
UUAG style (that always bothered me a bit)
nothingmuch same syntax as they're called
yes, me too
foo: $$.attr = { ... }
foo: $.child.attr = { } # inherited
luqui hmm.. I suppose we could just construct on the fly
no declaration then 10:42
svnbot6 r7237 | iblech++ | * t/general/basic.t: unEVAL, skipped the eval test under PIL2JS and PIL-Run,
r7237 | iblech++ | and moved the MMD test to the end of the file.
r7237 | iblech++ | * t/general/config.t: $*OS eq "browser" is ok.
nothingmuch ?
luqui defining it declares it, type-inference-style
nothingmuch oh, i see
luqui ("can't use attribute as both synthesized and inherited at grammar line...")
nothingmuch why do you hate threaded attrs?
luqui because it's a pun 10:43
you're really defining two different attributes
nothingmuch in that case, do you think a macro-ish approach to defining threaded attrs could work/
luqui I'd like to find a way to define a threaded attribute cleanly without punning
I mostly hate them because of how confused I was when I saw this in UUAG: 10:44
@lhs.number = @lhs.number + 1
nothingmuch $$.number = $$.number + 1 is a bit clearer
luqui uh, no 10:45
it's precisely as unclear. this is functional, damnit! $$.number is $$.number!
nothingmuch it feels like OO to me
okay okay =)
luqui :-)
nothingmuch the rule for this is pretty simple though, isn't it/ 10:46
?
the real .number is the synthesized one
and any calls to it from within it's body goes to the inherited one
luqui that's the rule?
nothingmuch and conflicting inherited ones get name-mangled
luqui yeah, I guess so.
I'd like to search for a better solution for a little while though 10:47
since it's macroey and doesn't really require any magic on the part of the core, I can put it off
nothingmuch what i would really like is to be able to be more generic
i wouldn't mount lack of threaded as a convenience if i could write 10:48
attr_out($$) = { attr_out($.child) }
and attr_in($.child) = { attr_in($$) }
only once
and have $.child be some sort of fmap
luqui elaborate
nothingmuch implemented with unordered (or order undefined) fold
so that i could either cons the thing up to get a map 10:49
or I could aggregate in the fold it to unify the attr into a scalar
luqui so the attr_in magic, afaict, is planned 10:50
you're saying that $.child represents more than one child?
nothingmuch no, that syntax is crazy
but I would like to be able to define aggregate attrs without enumerating all my children
especially for n-ary trees
since map { attr($_) } @{ children($$) } 10:51
whill no longer be alloweed
luqui oh. shit.
maybe allow @.children
as long as it is uniformly typed
hmm, maybe not though 10:52
I don't want to parse $.children[4].attr
luqui wonders what it would take to make it more dynamic 10:53
luqui stares at his attrgrammar.pl example
hmmm
It might be possible still to allow map { attr($_) } @{ children($$) } 10:54
nothingmuch btw, for all the haskell heads who don't appreciate foldr enough: let mymap f l = foldr ((:) . f) [] l
mymap (+ 1) [ 1 .. 3 ] 10:55
how is it possible? 10:56
and please allow reduce instead of map, it's more general
luqui wow
nothingmuch wow?
luqui the foldr thingy
nothingmuch ah
luqui by "It might be possible still..."
nothingmuch yes, I like it a lot, after figuring it out a few mins ago =) 10:57
luqui I meant that for arbitrary definitions of map
nothingmuch okay =)
how will that work?
luqui for parents that set inherited attributes
they just have a thunk hash key that looks like 'attr:Object=HASH(...)'
and then you look up yourself in your parent's hash 10:58
nothingmuch i don't understand 10:59
joao mymap seems obvious :) haskell is neat :)
nothingmuch who is yourself?
the user? the attr def?
luqui doesn't think mymap seems obvious. but it does seem cool
the node who is inheriting the attribute
joao luqui, do you understand foldr function?
nothingmuch attrs are not really inherited? 11:00
luqui no, they are looked up in the parent
nothingmuch oh
so the impl of inheritence is like OO inheritence
instead of putting the value in the child, it stays in the parent... right?
luqui joao, I understand it to the degree that I understand the compose operator: I know what it does, but I have to think about it for a while
nothingmuch, right 11:01
nothingmuch what is the compose operator?
luqui (.)
nothingmuch ah
luqui so then a definition that uses compose and foldr in the same statement takes a little more while
nothingmuch simpler form will be 'let my map f l = foldr (\x, xs -> (f x):xs) [] l
luqui right, that makes sense 11:02
broquaint That's a map() implementation, right, nothingmuch?
nothingmuch broquaint: yep
luqui that's one of my complaints with [+] and reduce in p6
joao i think the other is better :)
luqui there are two kinds of reduce, and we don't let the user pick 11:03
joao you can even forget that l :)
nothingmuch joao: ofcourse =)
broquaint I think, sometimes, Haskell is a little too terse.
nothingmuch you can also use autocurrying and flip to get rid of f too
let me try
luqui joao, the reason that people liked "the other better" is the reason that it took me two months to learn haskell instead of a week
joao luqui, :) 11:04
have you ever seen those catamorphisms diagrams? :)
luqui uh... catamorphism... rings a bell
joao after that, foldr (which is a catamorphism on lists) seems nicer
nothingmuch what's a catamorphism? 11:05
google has no defs
luqui @pl let mymap f l = foldr ((:) . f) [] l in mymap
flip foldr [] . ((:) .) 11:06
nothingmuch let mymap = ((flip foldr) []) . ((:) . )
luqui how very clear
nothingmuch right
oh, lambdabot does it? and i worked by hand? how lame!
luqui :-)
nothingmuch so what's a catamorphism? 11:07
joao well 11:08
a catamoorphism is a function
that operates on an indutive type
and builds some value
this is the most "general" definition I can give 11:09
I am searching a paper to show you
wiki.di.uminho.pt/wiki/pub/Educatio...calc_2.pdf
page 58 :)
nothingmuch "inductive type" is what I would call an aggregate type?
luqui recursive
like a list
nothingmuch ah
joao If an aggregate type is a recursive type, yes :)
nothingmuch really needs to take his dad's course 11:10
joao [] -> base ; (:) -> constructor
You can decompose [a] in ([] + a x [a])
meaning: the list is empty OR it is composed by an element (head) followed by another list (tail)
luqui ahh, x like cartesian product 11:11
joao yes
(a,[a]) :-)
Using diagrams to express catamorphisms is nice to explain recursive pointfree programming :) 11:12
nothingmuch define: decompose
be able to take it apart recursively?
joao yes, I think so :)
luqui he's just saying data List a = [] | a : [a] again
er, data [a]
nothingmuch and what is the + saying in that expr? 11:13
luqui union
nothingmuch tries to transpose what he knows about unions and products into this discussion
luqui thinks joao is using ocamlesque notation
nothingmuch so much to learn, so little brain 11:14
joao luqui, it's the notation I've learned to do pointfree proofs...
luqui, not sure if it derived from ocaml :P
luqui okay
nothingmuch what is pointfree? 11:15
luqui it's what you did to mymap
rep you mean pointless :P
luqui getting rid of all the names
nothingmuch ah
joao The nice thing about this stuff is that you capture the datatype recursion in the catamorphism definition (there is also the anamorphism notion, which is the opposite)
and then.. having a library defining the type and their cata and ana
luqui this is sounding like category theory 11:16
joao you just define the little pieces (called the gene) that make what you want :)
the recursion is captured in the libraries :)
you haskell code reduces a lot :)
luqui reminds me of the type analog of fix
nothingmuch thinks he wants to know that business =) 11:17
i have to quit my job and go study already
joao nothingmuch, I can give some examples someday (not now, sorry)
I'd be happy to show you how that stuff works :)
luqui joao, please do someday
nothingmuch joao: i'll try to remember to ask you... thanks =)
joao I will, thanks for letting me do it :)
nothingmuch luqui: whatever that stuff is, we want it in perl 6, right? 11:18
;-)
joao laughs
luqui of course 11:19
joao I'll eat something, be back later :)
nothingmuch ciao joao, and thanks again =)
luqui perl 6 is *the* fat language: everything that is in anything anywhere is in perl 6
nothingmuch s/is/will be/
luqui mathematically: perl6 = Union_(x in anything){ x }
nothingmuch where x is whitespace and anything is python? 11:20
luqui uh
perl6 = Union_(x is a feature of anything) { x }
joao but without cholesterol :)))
luqui and the stuff that is unique to perl 6 comes from the fact that perl6 (in) anything. 11:22
nothingmuch i think you want [ x | <- x in anything, f(x) > t ] where f x = (flip foldr [] . ((:) . )) (likes x) @larry
oh wait
no, you actaully want a real reduction, not a map ;-) 11:23
luqui haha
nothingmuch btw, wrt foldr - i learned to appreciate from the why ags matter article
luqui really?
nothingmuch so you should have done your homework better =)
yes, i thought that the def of diff *without* the AGs was really superb
not as maintainable or clear as an AG 11:24
but much more impressive ;-)
luqui in listing two?
nothingmuch yes
nothingmuch expanded it on paper
and then realized WTF was going on 11:25
luqui see, that's why I haven't switched to haskell yet. Programming is the middle ground between literature and mathematics
nothingmuch ?
luqui Haskell is much further on the right than most programming
you're supposed to be able to "read" code 11:26
nothingmuch yes, but the moment you bring yourself to that space, you can read almost as well
but write books that are black with ink
luqui is too tired to deal with metaphor right now 11:27
:-)
nothingmuch okay
basically: once you get used to haskell, you can still read code
less people in the world can
luqui that's sorta true
nothingmuch but the code you are now used to handling is much more dense
because most of the things that make your mental stack overflow become idiomatic the moment you understand them once. 11:28
luqui exactly, it's much further on the right, i.e. much more like mathematics
nothingmuch, good point
nothingmuch i don't want perl 6 to be this way though ;-)
luqui exactly
nothingmuch i think haskell, despite all it's advantages, is not accessible to enough people
luqui well, haskell forces you to rewire your brain 11:29
nothingmuch has a friend he would like to experiment on
the friend is just learning to program
luqui something that perl 6 is trying really hard not to do (but it's hard, because you'll have to rewire *more* to read code)
nothingmuch and the idea of teaching him haskell from day one crossed my mind 11:30
it will help me improve
luqui that would be a very interesting experiment
nothingmuch the problem is that his army duties have increased recently
so far we've gotten around to the meaning of structure and binding
he coded up a bottles of beer program
but we haven't done anything but get drunk together since 11:31
luqui haha
rep hehe 11:33
nothingmuch suddenly realized that there is more than met his brain, but seemed to have met your eyes in his last two sentances ;-) 11:34
masak :) 11:36
nothingmuch i need a break, i'm being silly instead of writing code
svnbot6 r7238 | iblech++ | * t/builtins/{arrays,hashes}/slice.t: Added tests for @array[1,2,3] := 11:38
r7238 | iblech++ | @array_containing_less_than_3_items;
r7238 | iblech++ | * PIL2JS: Prelude::JS::Array, Prelude::JS::Hash:
r7238 | iblech++ | Implemented binding of array and hash slices. This means
r7238 | iblech++ | t/builtins/{arrays,hashes}/slice.t pass 99% now.
r7238 | iblech++ | * t/builtins/undef.t: Added a try {...} around a (in PIL2JS) dieing test.
luqui nothingmuch, there was an AG discussion on PerlMonks 12:03
?
nothingmuch returns 12:04
where?
luqui r8709@syeeda: nothingmuch | 2005-10-01 14:52:23 +0200 renamed 'avg' to 'global_avg' to make things clearer (this confused Corion on perlmonks) 12:05
nothingmuch ah
i talked to Corion about this in pm chat
and he said 'i don't fully grok that yet'
so we went over it
and then he said 'well, i don't see how avg can be inherited, that just doesn't make sense'
luqui I see 12:06
nothingmuch so i said 'oh, that's because it's not really the average, it's the global average'
so yes, but not a posted discussion
arf! 12:08
i lost my copy of the hair soundtrack!
nothingmuch wonders how such idiocy happenned 12:09
12:28 Aankh|Clone is now known as Aankhen``
luqui sent off theory.pod to @Larry 12:30
nothingmuch woot 12:31
nothingmuch really hopes it'll be accepted
btw, i was meaning to implement the tuple type at some point
and type infer it
but i need to type infer regular pairs first ;-)
luqui oh you mean the arg tuple... 12:32
that could be a challenge
luqui hopes it isn't though
nothingmuch i don't think it's too hard
it's just a concrete data type that nests 12:33
err, abstract
but can be implemented concretely
and happens to be manipulated at compile time as well as runtime.
luqui well, I have to be off to school now 12:37
nothingmuch a.as-us.falkag.net/dat/bgf/200410/13/1085.gif (Add a FREE SpyWare program to your desktop)
luqui heh
& 12:38
nothingmuch ciao
svnbot6 r7239 | iblech++ | PIL2JS: Prelude::JS::Array: Sped up the new autovivification-capable &push and 13:20
r7239 | iblech++ | &unshift by 88% by porting the necessary if-clause to JS. As &push is used very
r7239 | iblech++ | often (for, map, etc. all use it), this should speed up make smoke by 10 min or so.
nothingmuch en.wikipedia.org/wiki/Wayne_McLaren 13:50
cute
joao nothingmuch, but people continue to smoke... 14:19
nothingmuch i smoke too, occasionally
i just think it's funny =)
not so much the actual cancer stuff
joao I don't mind if others want to destroy their health, actually. I'm just bothered when they don't respect mine.. 14:20
svnbot6 r7240 | iblech++ | * t/builtins/perl.t: Added tests for "\t".perl, "\n".perl, etc.
r7240 | iblech++ | * perl5/Perl6-Value/lib/Perl6/Value.pm: Fixed the stringification
r7240 | iblech++ | ("$key\t$value") and .perlification of pairs ("($key.perl() =>
r7240 | iblech++ | $value.perl())").
nothingmuch but the fact that an icon of maniliness (implying stamina, power, etc), that caused people to consume cigarettes so directly meets such an end
and i also find it ironic that while advertisement makes it easy for people to start smoking, you don't see it stopping anyone (my bet is the budget difference is the reason for this) 14:21
joao Probably 14:22
nothingmuch as for your health - smoking is like peeing in a swimming pool... I totally agree, and people should ask permission before smoking next to anyone
as for mine - i'm not too worried - i smoke around once a week 14:23
joao Well, I read somewhere that being a passive smoker gives you more 40% possibilities of having coronary issues
I'm not worried if someone smokes near me; I just think it's disrespect
nothingmuch it is 14:24
joao I'm from Portugal, and believe me, portuguese smokers don't care about others ;)
For instance, in my department I've seen people smoking below the "NO SMOKING" sign 14:25
Instead of removing the smokers... they removed the sign! :)
nothingmuch i thought the trend in europe in the past 20 years was against smoking
i remmeber that 10 years ago in austria you couldn't walk into a resteraunt without choking 14:26
but nowadays even my aunt's husband quit
joao It's funny, some days ago, I've seen some old men who want to sue tobacco companies because the companies made them believe that smoking was a "men thing" 14:29
Today they need special devices to emit some sounds... it's sad.
nothingmuch it is 14:30
i think you'll like this: www.adbusters.org/spoofads/alcohol/...impotence/
joao hehe :) 14:31
shakespeare was a bright fellow :)
nothingmuch yup 14:32
joao nothingmuch, where are the list operations defined? 14:36
nothingmuch src/Pugs/Prim.hs or src/perl6/Prelude.pm
svnbot6 r7241 | iblech++ | util/run-smoke.pl: "You may want to submit the report to the public
r7241 | iblech++ | smokeserver: [...]"
Aankhen`` I definitely am worried if people smoke near me. They are consciously choosing to adversely affect my health through secondhand smoke. "Disrespectful" is not the word. 14:37
joao I am trying to get into Pugs...
Aankhen``, I agree, byt my english level don't allow me to be more explicit :)) 14:38
Aankhen`` Heh.
joao :)
nothingmuch yowza! 15:33
type inferrence of non generic functions is working
now to do generics
svnbot6 r7242 | iblech++ | * STATUS: s/PIL2JS passes almost 90%/PIL2JS passes 90%/ :) 15:36
r7242 | iblech++ | * pugs::hack: s/a public server/the public smokeserver/
r7243 | iblech++ | * t/builtins/strings/substr.t: substr(..., $replacement) superceded by 16:47
r7243 | iblech++ | substr(...) = $replacement?
r7243 | iblech++ | * PIL2JS: Prelude::JS::Str: &substr:
r7243 | iblech++ | * Fixed &substr when $start_pos > 0 and $len < 0.
r7243 | iblech++ | * Made &substr return a readwrite proxy object:
r7243 | iblech++ | substr($str, $pos, $len) = $replacement; # works now
r7243 | iblech++ | * This means substr.t passes 30/31 now, with 6 unexpected successes. :)
joao what is the equivalent for "apt-get update; apt-get upgrade;" on freebsd? "freebsd-update fetch; freebsd-update install" ? 16:57
nothingmuch uh
autrijus: ping?
i guess you ought to try #freebsd
i don't know about anyone but autrijus who uses freebsd and is active on this channel
seen autrijus
jabbot nothingmuch: autrijus was seen 1 days 22 hours 38 minutes 38 seconds ago
joao nothingmuch, thanks :) 16:58
b6s /c/c
oops, sorry
joao let me see the handbook, first :D
nothingmuch b6s: that's really unacceptable
please take this behavior to another channel 16:59
joao grins
nothingmuch OOOF!!! Milner is melting my bRAIN 17:00
joao nothingmuch, Pugs should have a TODO list in its root :-)
nothingmuch errm. there's too much to do ;-)
i can spit out some random stuff
if you want to hack core - look at the smoke reports
if you want to hack perl 6 - make the smoke report have bigger numbers
or port/write a module 17:01
otherwise there are some side projects:
PIL2JS, PIL-RUN, the metamodel, Blondie
blondie has a side project that I'm working on right now - hindley-milner inferrencing in perl
with enough hooks to be able to generate type runtime type boxing and unboxing apps instead of type errors where applicable 17:02
when this is done it'll get backported to blondie
then that will hopefully give us knowlege on how to compile perl 6 statically
PIL2JS is a PIL compiler that emits JS written in perl 5 17:03
and PIL-RUN is a PIL interpreter written in perl 5
the metamodel is a spiritual project, and is implemented mostly in perl 5, but is supposed to be ported to haskell 17:04
i hope some of these interest yyou
and if they don't - you can add a TODO file ;-)
do you have commit access?
joao yes, I do
nothingmuch good
joao I just need to get into the structure, and then into *something* where I can be useful :) 17:05
nothingmuch would you like a brief explanation of how pugs works? 17:06
joao I don't want to bother or make you guys loose your time... but I'd appreciate if you really want to do it :)
nothingmuch thinks long and hard if he *really* wants to override isa
joao: well, in the long term I guess by helping you out I could be saving time ;-) 17:07
either way, it's worth the effort
even if only for the hype
so ask away
joao :)
well, what would you say to someone who's looking into the source for the first time? :) 17:08
nothingmuch ermm
pugs is an interpreter written in haskell 17:09
it parses (src/Pugs/Parser.hs) perl 6 with parsec
and constructs an AST
this AST is then reduced
joao nods and pays attention
nothingmuch the reduction is done along side an Env
this contains symbols and stuff
when you compile pugs an interesting thing happens: 17:10
src/perl6/Prelude.pm is compiled
into haskell
which is #included into Run.hs 17:11
that is +- the initial env
this is the big slow down in compiling Run.hs, btw
the AST itself can be serialized or evaluated by pugs
the serialization emits PIL (pugs intermediate language) 17:12
and the evaluation emits 'IO ()'
now your turn to ask more questions
*now it's
joao ok, first: you said the AST is reduced along with an Env. I didn't understand what you mean. (The reduction is optimizing code, right? Evaluating certain things to get a smaller AST, etc..) 17:14
nothingmuch no, the reduction is the actual evaluation
i kind of confused the order
we have no optimizer yet
joao oh
nothingmuch (though a peep hole optimizer for PIL could be a cool project)
joao ok, so..
the AST is evaluated using an Environment, is that it? 17:15
joao smiles
nothingmuch yes
the env keeps the symbols and stuff
i forgot the precise details
joao where is that env defined? :)
Run.hs ?
nothingmuch let me see
i doubt it
Run.hs is the command line arg processing stuff
and the code to take source from a file/string/interactive input, and compile/run it 17:16
AST.hs is everything AST
and AST/Internals.hs contians 'data Env'
joao ok, another question. Run.hs includes a PreludePC, is it the Prelude.pm compiled into haskell?
nothingmuch yes
joao ok, last question for now: when you say "AST serialization" what do you mean? 17:17
nothingmuch the AST can be dumped to a file/stdout
for example:
joao oh, ok
No need to explain 17:18
nothingmuch ./pugs -CPIL -e 'say 10'
joao :)
nothingmuch just FYI regarding the PIL: it's scheduled to be replaced by PIL2 soon
but it should be quite similar
joao hmm
PIL is also used for Parrot Intermediate Language, right?
nothingmuch as you can see the PIL output contains MkPos etc to give back nice runtime errors 17:19
joao (the acronym)
nothingmuch (no, that's PIR - which is like assembler, while PIL is like ANF)
(not that I know what ANF is - excpet the acronym ;-)
joao Good, PIL is defined as an Haskell DT
nothingmuch yes 17:20
rep DT? 17:21
joao What is the role of Prelude.pm in this figure? (sorry, I am just trying to put things together)
datatype
nothingmuch data type
Prelude.pm is written in perl
and it's just the basic functions
that can be implemented in perl
(src/Pugs/Prim.hs contains the table of perl stuff implemented in haskell) 17:22
it's precompiled for speed
joao Oh, but you could implement those in haskell too, right?
nothingmuch instead of being compiled as perl into every program
whatever's implemented in perl is not implemented in haskell
if I get my way (which there is no reason that won't happen, because no one said it's a stupid way yet) this distinction won't be there
the prelude will implement stubs for E.G. IO 17:23
and each runtime will implement whatever prelude functions it wants for efficiency/features/whatever
joao E.G. ?
nothingmuch err, e.g.
you should know, you speek a latin derived lang, don't you?
for example in english.. i forget the latin
joao ou 17:24
:)
nothingmuch i remember id es
joao I thought it was some acronym :)
nothingmuch exempli gratia
no, i just had a brain fizz and typed it uppercase by accident ;-)
joao :) 17:25
Ok...
Southen_, it would be nice for me to start by Parser.hs 17:26
And then take a look at Run.hs
s/Southen_/So,
nothingmuch tab completes words he shouldn't all the time too
joao :)
nothingmuch sure
you should know:
joao xchat, here..
nothingmuch in perl we have begin blocks
they are executed the moment they are done parsing 17:27
joao uh? you execute a block after parsing it?
nothingmuch ?eval BEGIN { say("foo") } oiu2165 {}{{{{{ syntax error
evalbot_7243 foo Error: unexpected "{" expecting block construct, ":", term, term postfix, operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
nothingmuch as you can see it emitted foo before even parsing the other stuff
that's the unsafe stsuff in the parser 17:28
joao hmm, what's the reason?
nothingmuch uh, perl just is that way =)
it's useful
joao I'm learning perl6 at the same time I look into Pugs. It will be nice ;) 17:29
nothingmuch =)
joao The problem is all the stuff I have to end this month
(work stuff)
nothingmuch work is always a problem ;-)
dev.perl.org/perl6/doc/synopsis.html
that link contains most of what you need to know about perl 6
joao yes... I was following the oreilly book 17:30
nothingmuch as a delta against perl 5
ah
nothingmuch bought it but hasn't read it yet
joao but I think some little things are already old
nothingmuch it's probably way out of date by now
so i'm undermotivated to start reading it now ;-)
joao :)
why 1-6, 9-13, 29 ? 17:31
nothingmuch 7 is obsolete, i think 17:32
8 is unwritten
joao oh , ok
nothingmuch 13-28 are unwritten (but mostly minor)
dev.perl.org/perl6/doc/apocalypse.html
these are much bigger, more detailed, and less processed 17:33
as you can see A7 is a placeholder
and E explains how that behavior will be implemented for backwards compatibility with perl 5 as a standard module
joao I think I won't read the book. Synopses look better :) 17:34
nothingmuch i think they're easier
since they encourage interactivity
joao :) thank you
nothingmuch you can find syopsis and tests that are interlinked here: nothingmuch.woobling.org/pugs_test_.../Synopsis/
for example operator overloading, in s06 is tested in two placed 17:35
denoted by the small 't' links next to the heading
the tests can also point to text within the section by matching a regex 17:36
that can help you see how these things look and run
joao thanks! very cool :) 17:37
nothingmuch will go biking soon 17:40
lets see if i can finish the generic function stuff first though 17:41
shit 17:44
joao what? 17:48
nothingmuch is not getting along with AGs at the moment
i have to keep track of which type variables are generic
and my code to unify the set with a new member is not being invoked at all =(
autrijus greetings from .fi! 18:00
nothingmuch hola autrijus!
what is fi?
finland? 18:01
joao finland?
:)
obra 'afternoon, autrijus
autrijus yup :)
autrijus gets ready for the weekly parrot meeting
nothingmuch autrijus: full milner type inferrencing is almost ready for blondie
autrijus nothingmuch: excellent. then we can talk about adding the annotation strata on it 18:02
nothingmuch define: strata
gah: No method found for args (type::operator::nullary=HASH(0x196c5b0) ide=HASH(0x191f440) val=HASH(0x196c550)) at (eval 138) line 20
i have no 'No method' or 'args' in the file
i wonder what bit of code is making that ;-)
autrijus nothingmuch: see Francois Pottier's paper... a sec 18:03
nothingmuch nono
no more papers today
my brain is at 98c
autrijus nothingmuch: search for "From ML Type Inference to Stratified Type Inference"
oh ok.
nothingmuch a little more stress and it'll melt
let me ask another question:
what is stratified type inference good for?
autrijus it's good for propagating user-annotated types first 18:04
nothingmuch ah
autrijus because you see, in H-M the user doesn't have to annotate any time.
nothingmuch right
autrijus but in richer type systems, H-M no longer suffice
nothingmuch yep
autrijus one well known example being higher ranked polymorphism
nothingmuch i learned some things about haskell that I didn't imagine really were
autrijus so there need to be a systematic way to codify the heuristic
of taking user annotations and propagate in "obvious" direction 18:05
befor you run the M-L
thus saving the user from having to type in every type.
nothingmuch woah
autrijus it's also known as "local type inference" and "wobbly types"
nothingmuch hmm
makes sense
autrijus but pottier gave it a consistent treatment.
nothingmuch will look at it tomorrow
nothingmuch 's code is actually running again 18:06
# got: 'āˆ€Ī±. Ī±'
# expected: 'int'
bah
now to find out where i forgot to unify now that we're dealing with generic types
autrijus is playing around with Visual Haskell 18:07
nothingmuch shyte
if i clone the a in a -> b and b is intantiated to a i need to keep them in synch
nothingmuch will do a deep clone of both
WOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOTTTTT 18:10
it actually work! 18:11
s
now to fix the pretty printer 18:13
what's the perl switch to make all handles utf8 by default? 18:14
Test::Builder is not listening to 'BEGIN { binmode STDOUT, ":utf8" }
autrijus nothingmuch: use encoding 'utf8' 18:19
nothingmuch doesn't help
the problem is sub _new_fh in T::B
perl -CDS is somehow not convincing *that* instance of open to use utf8 by default
oh well 18:20
whatever
now to splat a DAG into a tree
joao when I remove a file in a svn rep, doing svn up restores the file. svk pull shouldn't do the same? 18:23
autrijus joao: no, it's "svk revert" and for directories, "svk revert -R".
hey stevan! got questions for you!
(but in a meeting now)
joao autrijus, thanks
Khisanth nothingmuch: seems to work fine with the binmode 18:26
nothingmuch Khisanth: the problem is that T::B::_new_fh is opening local *FH
as a dup of STDOUT, i think
and somehow it's not applying to it
Khisanth hmm 18:27
nothingmuch Wide character in print at /System/Library/Perl/5.8.6/Test/Builder.pm line 1078.
asee also 'sub output' and 'sub _new_fh'
autrijus_tw gah. my laptop is drinkign tea again. 18:41
autrijus_tw hopelessly waits it to dry up
Khisanth oO 18:47
nothingmuch heh 18:52
WWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE
pasteling "nothingmuch" at 82.81.247.180 pasted "inferrence" (18 lines, 344B) at sial.org/pbot/13440 18:53
"nothingmuch" at 82.81.247.180 pasted "inferrence" (18 lines, 449B) at sial.org/pbot/13441 18:54
nothingmuch oops
nothingmuch whores lots of karma in this commit =) 19:01
we need more smokers 19:03
linux is not being done at all
oh, sorry 19:04
wrong group
svnbot6 r7244 | nothingmuch++ | Milner - Hindley type inferrencing for blondie 19:09
r7244 | nothingmuch++ | * implemented with Language::AttributeGrammar and Class::Multimethods::pure
r7244 | nothingmuch++ | (thanks to luqui++)
r7244 | nothingmuch++ | * new temporary AST for easier polymorphic thunks, will be backported
r7244 | nothingmuch++ | * type pretty printer with unicode
autrijus_tw smoke.pugscode.org is up as an easier-to-remmber facade 19:10
nothingmuch how do you svn prop?
yay! 19:11
svk
autrijus_tw nothingmuch++ for spreading the love to parrot :)
svk prop?
nothingmuch no, i mean the "default" ones we use
autrijus_tw svn:mime-type text/plain; charset=UTF-8 19:12
nothingmuch gaal told me about the script
autrijus_tw svn:eol-style native
nothingmuch feels like this could have been really beautiful code if it were perl 6 19:13
we need AGs for p6
autrijus_tw port L::AG over?
shouldn't be that hard :) 19:14
nothingmuch luqui is redoing it
so not now
and I think this can be implemented with a much better integrated interface with perl 6
svnbot6 r7245 | nothingmuch++ | r8815@syeeda: nothingmuch | 2005-10-03 21:12:32 +0200
r7245 | nothingmuch++ | propfix
nothingmuch eepp
nothingmuch always forgets to push --verbatim
vladtz I could do some linux smoking (however my box is a amd-64 and I use the 32 bit compilers) 19:15
nothingmuch autrijus_tw: thanks *yet again* for introducing me to AGs
vladtz: it still runs, doesn't it?
vladtz Did yesterday, lets svn and check, back in an hour,....
nothingmuch vladtz++ 19:16
autrijus_tw nothingmuch: looks like you are having a lot of fun with AG :) 19:20
nothingmuch autrijus_tw: lots
autrijus_tw wonders if nothingmuch will benefit from the EHC tutorials/code
nothingmuch i'll have a look tomorrow 19:21
autrijus_tw it's a surprisingly complete Haskell implementatino :) 19:22
nothingmuch what does it compile to?
or is it an interpreter?
autrijus_tw it compiles to Grin
nothingmuch cool
autrijus_tw from there to native, I think 19:23
there's some talk about LLVM but didn't transpire
P6 probably doesn't want grin, I think it wants ANF with some modifications.
nothingmuch is there any place I can learn about ANF that isn't the compiling with continuations paper? 19:24
it's a bit over my head
i'll go over it with dad tomorrow (he's a logician, so he can read this stuff) 19:25
autrijus_tw oh, ANF is actually very simple
nothingmuch what is diff(ANF, SSA) ? 19:26
Aankhen`` goes to sleep.
G'night.
joao wonders what ANF is 19:27
b6s A-something normal form?
autrijus_tw Administrative.
nothingmuch: www.cse.unsw.edu.au/~chak/papers/CKZ03.html
joao American Nurses Foundation? :)
b6s oh, restricts to lambda 19:31
vladtz pugs still build on linux/amd-64, cpan however is confused somehow and tries to run tests with pugs iso perl, anyone???
autrijus_tw vladtz: nopaste a log? 19:32
vladtz where is pastebot?
kolibrie perlbot nopaste 19:34
perlbot Paste your code here and #<channel> will be able to view it: sial.org/pbot/<channel>
pasteling "vladtz" at 80.127.84.50 pasted "CPAN confused into usign pugs" (13 lines, 841B) at sial.org/pbot/13443
vladtz kolibrie: thanks
kolibrie vladtz: no problem 19:35
vladtz time for a drink with the so, bye 19:39
IÄŗl check the logs for possible anwers.... 19:40
Limbic_Region nothingmuch - did you see luqui just journaled on AG again? 19:42
nothingmuch yep
he raised the topic earlier today
kolibrie nothingmuch: can you figure out what vladtz's paste means? Test-TAP-Model problem 19:43
nothingmuch phooey
i wonder what version of Module::Build it is 19:44
kolibrie who knows
nothingmuch T::TH has no relation what so ever to pugs
perhaps PERL_TEST_HARNESS leaked?
kolibrie that's what's so weird
nothingmuch it looks like it
autrijus_tw gives up on he h6-e 6f the 2eyb6ard behavng t nght 19:45
nothingmuch heh
i hope there was no sugar in the tea
autrijus_tw we'33 see if tomorrow wil3 mae it better
gah :(
nothingmuch go to sleep autrijus-san
you deserve it
autrijus_tw well at leat I'm bac to full wireless world 19:46
obra autrijus_tw: ow :/
autrijus_tw waves &
cognominal autrijus is cloning himself, for more productivity?
kolibrie autrijus_tw: good night, better laptop success tomorrow 19:47
Limbic_Region so without waking autrijus back up - how much has he gotten done with the Haskell MM implementation? 19:49
nothingmuch commented milner.pl 19:59
the algorithm is so much simpler than I had initially thought
Limbic_Region Just about everything, properly understood, is simple 20:01
svnbot6 r7246 | nothingmuch++ | r8817@syeeda: nothingmuch | 2005-10-03 21:57:57 +0200 20:02
r7246 | nothingmuch++ | Minor cleanup and lots of comments in the milner script
r7247 | nothingmuch++ | Also explain the unification multimethods 20:07
nothingmuch goes biking 20:32
svnbot6 r7248 | autrijus++ | * GHC 6.4.1 reportedly works without gcc_select, 20:38
r7248 | autrijus++ | so reflect this in Makefile.PL.
nothingmuch $30 for a fucking article?! 20:42
phooey!
this world needs to appreciate free knowlege better
ciao
vladtz found my CPAN problem: HARNESS_PERL was set (is that still needed?)
nothingmuch vladtz: that is not a good thing if you are installing perl modules instead of testing pugs 20:43
and for pugs it's only needed if you're running 'prove' instead of 'make test'
vladtz nothingmuch: it's seems to a bad thing while smoke testing as well...
nothingmuch could be... but i don't see why 20:44
vladtz nothingmuch: I'lll remove it from my profile, thanks. Smoke coming up soon (I hope).
nothingmuch yummy
nothingmuch leaves for the 2nd time ;-)
obra Which lambdacamels are blogging about perl6? 21:58
autrijus luqui in particular
stevan too
obra what's stevan's url? 21:59
Luke I've got
autrijus use.perl.org/~stevan/journal/
obra Last name is Little? 22:00
autrijus aye
obra anybody else who should be on planetsix? 22:02
autrijus not sure... nothingmuch didn't keep a journal iirc 22:04
gaal's got livejournal but not usually p6 related
obra nod 22:06
ok
joao autrijus, weren't you sleeping? :)) 22:30
autrijus ah, but my keyboard came back to live... 22:31
joao or you just need to sleep 2 or 3 hours? :)
autrijus nah, usually 10hrs. 22:32
joao really? wow
and still have time for everything you do? 22:33
joao smiles
autrijus the trick is to sleep 10 hours for every 25 or 26 hours :)
joao I thought you were like Paul Erdos, that needed 4/5 hours :)
hehe!
autrijus nah, I'm not currently powered by amphetamine :)
joao "currently" :) 22:34
rep i sleep 6 hours every 24 hours..
well
as in i go to bed at around 4am every day
and wake up at 10am 22:35
joao I can't do less than 8 hours. it's phisically impossible :PP
25 continuous hours looks a lot.. even with the 10h sleep :P 22:37
svnbot6 r7249 | autrijus++ | * finish removing IRC-OO.pm
clkao agent tang 23:07
svnbot6 r7250 | autrijus++ | * Cabalization for real, thanks to Visual Haskell's help.
r7250 | autrijus++ | * Also checks in the "solution" file for VS2003.
clkao src/Pugs/Compat.hs:80:5: 23:12
Duplicate instance declarations:
src/Pugs/Compat.hs:80:5: instance Typeable DirStream_
justatheory seen autrijus 23:46
jabbot justatheory: autrijus was seen 1 hours 12 minutes 48 seconds ago
buu How many freaking bots do we have any here? 23:53
s/any/in/
QtPlatypus None, all our buts are freek free.

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!