6.2.2 is airborne! | pugscode.org <Overview Journal Logs> | pugs.kwiki.org
Set by autrijus on 1 May 2005.
stevan mugwump: should I add visibility in MetaMethod as well? 00:02
mugwump sure, if you like. I'm not sure what will be "private" on the Class model yet, but no doubt there will be something eventually 00:03
stevan ok
and you have it with Assoc in Pugs/Class.hs 00:04
mugwump yep. they're all things tacked onto the class, which can be for private or public use 00:05
stevan should we default to Public? or is that not okay?
mugwump er, dunno. I'd say public by default 00:08
but that's a language design issue, we're writing a compiler not a language ;)
stevan ok
I am still barely on the edge of "getting this" :) 00:10
Odin- mugwump: That depends on various things. Like, for instance, sunspot conditions. 00:17
mugwump well, the position of the moon is more significant. but yes sunspots affect it. 00:18
it's the third of the month! still a good time to be planting new ideas 00:19
stevan mugwump: BTW - does it look like I got (? MetaClass A, B | A.clsSuper = B ? A ? B.clsSubClasses) right? 00:21
mugwump ok, well that mutator can't remove the superclass entirely 00:23
stevan mugwump: doh! i didnt even think about removing 00:24
mugwump and neither can the subClasses one remove subclasses etc.
stevan so if I change the superclass i need to remove it from the list of subclasses
mugwump yes. In general, when you change the members in an association that has a "companion"... 00:25
you need to send a message to the objects removed to remove self, and a message to the new objects to add self
stevan ok, actually I think once I refactor subclasses to be a hash (and get our "set") that might not be too bad
mugwump Having a "Set" class as a primitive makes this a lot easier than doing set operations with hashes 00:26
stevan true
mugwump I had a look at the hash tests, they don't seem to test hashes with object keys
stevan our objects are still strings 00:27
so we can get away with hash for now
mugwump a Set is just a Map with no values, whereas a Hash is a Map with string keys
stevan eventually when we have real object, then we can refactor into a proper set
mugwump oh true didn't consider that :)
stevan mugwump: ok, subclasses are now stored in hashes 00:31
%self<subclasses>{$subclass} = undef;
at this point all it does is avoid duplicates
mugwump ok 00:33
stevan should I be checking anything else when I add the subclass? I already check to make sure the superclass of the subclass is the class 00:34
putter I'm trying to understand a regexp test case m:words/.../. Has anyone heard of a ! ("shriek") modifying the interpretation of adjacent whitespace? 00:35
stevan let me commit so you can see it
mugwump perhaps check that it doesn't make a loop
stevan so that my subclass is no a superclass of my class 00:36
s/no/not/
putter As in ok( "abc! def" !~ m/abc! def/, 'Literal post-shriek nonmatch' );
and ok( "abc!def" =~ m:words/abc! def/, 'Word nonspace post-shriek nonmatch' ); 00:37
mugwump putter: according to S05, ! negates the meaning of a <> section, and makes it zero width. maybe the person who wrote the test mis-read the synopsis, or was basing it on a mailing list post that hasn't been incorporated into the synopses. 00:38
putter Both odd, and the last not even matching its own documentation. Perhaps its an historical artifact, but I havent turned up discussion. 00:39
mugwump then remove it. It can lie forever in peace in the version control system :) 00:40
putter mugwump: thanks. Ok, I think I'll just comment it out, rather than pursuing detective work into the Perl6::Rules implementation.
stevan mugwump: no loops (although that only applies to one level deep) 00:45
mugwump: commited and tested loop free (one level deep) 01:02
mugwump: should I write something which can check more than one level deep? 01:03
more like .isa() would behave?
all this meta-logic is making me hungry, bbiab 01:05
mugwump burps loudly 02:02
stevan: if .isa() works, enforcing no circular inheritance should be easy 02:03
or .clsIsa(), perhaps 02:04
meppl gute nacht 02:10
stevan mugwump: ok I will do clsIsa() since we don't have a working .isa() 02:12
mugwump: BTW - changing superclasses now removes the inv from the old super classes subclasses
try and say that 10 times fast :P
mugwump mmm. That one little axiom sure translates into a lot of code... 02:13
stevan thats how these things always go 02:14
I thought Class::Trait would be simple 02:15
but I ended up writing and re-writing it 3 times, and in the end it was almost 1K of code
s/code/LOC/
mugwump I've implemented this companion association thing before, in Class::Tangram 02:16
stevan I will have to look at that when i get to that part 02:17
(or you can do it :)
mugwump That ended up being 1,000 LOC; mostly due to the lack of an inheritance relationship between the four different collection types C::T supports (ref, set, array, hash) 02:19
(er, C::T in this case meaning Class::Tangram :)) 02:21
Class::Traits is like Luke Palmer's Class::Role, no?
stevan yes and no 02:22
it is based strictly on the Traits paper
Class::roles was more like the perl6 roles
Ovid actually start it
then ran out of time
and so I took it on
then I found a paper on the formal defintion of traits 02:23
mugwump What do you see as being the fundamental difference between these Traits and Perl 6 Roles, then?
stevan Roles can have state
Traits cannot
although apparently the newer version of Traits will allow state (according to some of the guys who are working on it) 02:24
other than that, most of the differences are minor (at least as far as I remember)
mugwump (from S12:)
A role may not inherit from a class, but may be composed of other roles. A role doesn't know its own type until it is composed. Any mention of its main type (such as ::?CLASS) is generic, as is any reference to the type of the invocant. You can use a role name as a type, but only for constraints, not for declaring actual objects.
stevan that is very much like Traits 02:26
SUPER:: in Traits refers to the super class of the class which implements the trait
mugwump where "superclass" means "the roles that this thing belongs to" presumably
stevan no, the $class->SUPER:: 02:27
basically resolution of SUPER is deferred until the last possible moment
mugwump "Stevan Little implementated Traits in Perl 5. In his series about the unfolding of the Perl 6 design, Larry Wall, the creator of Perl, gives a description of Roles, the variation of Traits that is planned as a fundamental language element of Perl 6." 02:31
-- www.iam.unibe.ch/~scg/Research/Traits/, linked from your Class::Traits module :) 02:32
stevan yup :)
i had some interesting talks with some of those guys
they were very interested to know about Perl6 Roles
mugwump cool. the Object equivalent of lambda calculus is roles-based rather than inheritance-based, too 02:34
stevan roles make so much more sense for code re-use too 02:37
although it is easy to get carried away with over-abstraction
which is basically why I never /actually/ did anything with Class::Trait 02:38
mugwump heh
stevan mugwump: I have questions about .clsIsa() if you have a moment 02:44
mugwump ok
stevan so if I check the .clsIsa() if a class, i should check it's super, and then the super-super and the super-super-super 02:48
until I find a match
always checking against the .clsName()
mugwump may as well be recursive
stevan yes
ok, so I thinking about this correctly then? 02:50
and should it be 02:51
$inv.clsIsa('Class Name') or $inv.clsIsa($class)
mugwump I think so, yes 02:52
stevan or both?
mugwump either? :)
stevan LOL
cognominal $ pugs /tmp/roundtrip.p6 05:01
No compatible subroutine found: &eval_parrot
what I should do to get that to execute?
mugwump did you install libparrot? 05:16
f0rth nmake win32 pugs failed for hours, src/Pugs/Types/Hash.hs:83:45: Couldn't match `Maybe' against `[]' 05:25
mugwump have you run `make clean' recently?
might be waving a dead chicken, but worth a try...
f0rth I've svn up a clean one 05:26
mugwump I only say that because someone was mentioning build weirdness with that file earlier 05:30
cognominal pugs -e 'my &say = sub { *say("toto") }; say "tutu"' 05:39
this loops...
I thought that would call the lexical &say that would itself call the global &say 05:40
mugwump night all! 05:42
gaal_ mugwump, oh, i just wanted to ask you a quick question :) 05:43
cognominal more like the morning here :)
gaal_ in Hack::Instances, shouldn't there be a global counter for instance ids as well as some randomness?
pasteling "gaal" at 192.115.25.249 pasted "good morning, the build is broken" (7 lines, 283B) at sial.org/pbot/10032 05:44
gaal_ oh, it's been noticed, sorry. 05:45
just backlogging
System.Posix.Env.getEnv and System.Environment.getEnv have different signatures. 05:54
f0rth, what platform are you on, windows, right? 05:55
f0rth gaal_: windows 2000 05:57
gaal_ yeah. i'm attempting a fix.
theorbtwo Define a function in Pugs.Compat that always has the right signature. 06:05
gaal_ yes, that's what i'm doing.
Trying to figure out how to transform IO String to IO (Maybe String)
clues welcome :) 06:06
theorbtwo do {foo <- thingReturningIOString; return Just foo}
gaal_ fixed in r2917. 06:26
06:38 castaway_ is now known as castaway
hachi I see that the syntax docs for p6 say : cannot be used as a quote delimiter "since it will always be taken to mean something else regardless of what's in front of it.", wouldn't q:::string: be a clear (to the interpreter anyways) way of delimiting 'string' ? 07:11
maybe there's some usage of : that I'm missing in that context, but as I'm reading all of this the third : has to be the delimiter, where am I going wrong? 07:12
Juerd hachi: It's technically possible, but requires huge workarounds and blurs a lot. Is being able to do this worth such a sacrifice? 07:14
hachi no, but the language design docs say there is some reason to which it cannot be done 07:15
the most likely thing then is that I'm missing some usage of : in string quote context, and I would like to know what it is
Juerd It can't be done with a single : 07:16
hachi of course, but that's not what it says
Juerd A single character is what works for all other characters. 07:17
(that are valid)
And it's rather stupid to *assume* exceptions, IMO
hachi hey, all I'm looking for is either the docs to be wrong, or me to be missing something... personally I think the latter is far more likely. You are saying that it's the former though? 07:18
Juerd You're missing that the documentation is the specification 07:19
If it says something cannot be done, and there is no technical reason for that, then it is *declaring* that it cannot be done.
hachi but if you read my quote, it says there is a reason... but it doesn't say what 07:20
Juerd I think you can safely assume that the reason is keeping people from having to learn weird exceptions. 07:21
hachi "it will always be taken to mean something else regardless of what's in front of it." -- Okay, assuming true, what does a third : mean in quoting syntax?
syntax error?
Juerd I hope so!
In fact, q:: is a namespace, so at that point it's already no longer q// 07:22
hachi ahh!
see, that's what I was looking for
not your opinion
Juerd I'm truly amazed.
hachi sorry, I need to not be an ass... bad start 07:36
07:44 chady_ is now known as chady
gaal_ what unicode terminal fonts are you folks using? I want one that has, for example, ?. (recently seen in MetaClass.pm) 08:00
Corion: ping 08:22
11:53 chady is now known as chady_
kolibrie autrijus: I'm always getting true for rules when using embedded parrot, but they work correctly when not embedded 12:22
Limbic_Region salutations all - anyone mind giving me an edumacation on binding := ? 12:42
I got a blast from the past email from someone who recognized my name on the p6.Language list 12:43
they sent me an email as they didn't want to appear "stupid" on the list
when taking a reference to a subroutine as in my $s_ref = \&foo;
they wanted to know why wouldn't (or perhaps couldn't) that be my $sub := foo(); # or something like that with the binding operator 12:44
oh good, perhaps gaal_ will have the answer 12:45
stevan Limbic_Region: $sub := foo() would bind to foo()'s return value
Limbic_Region: you would want $sub := foo;
Limbic_Region nope 12:46
he wanted $sub := foo()
gaal_ the gaal that has an underscore tucked to his name is not the true gaal.
Limbic_Region at least I think he did
stevan gaal_: what have you dont with the real gaal
Limbic_Region let me expand the question a bit further
stevan Limbic_Region: the () implies execution though
Limbic_Region right
12:46 gaal_ is now known as thetruegaal 12:47 thetruegaal is now known as gaal_
Limbic_Region foo returns a sub ref itself 12:47
which is a closure over a static variable
stevan ah
then I would expect $sub := foo() to work then
Limbic_Region he want's $variable to always be the value of the static variable inside foo
only it won't
since the return value isn't actually the static variable 12:48
stevan hmm,.. it won't in pugs? or perl6?
Limbic_Region it is an anon sub that when executed returns the static variable
stevan ah
Limbic_Region so now that I understand - I can explain it to him (and encourage him again to ask on list since it is obviously a good question)
stevan sounds like insanity to me :)
gaal_: ping 12:49
integral hmm, why are src/Pugs/Types/* not modules?
stevan integral: perl6 modules ? 12:50
integral sorry, I mean Haskell modules. They're just #include'd into Pugs.AST
stevan integral: no idea, autrijus is best to answer that 12:51
integral hmm
autark-jp module dependency maybe? 12:52
integral maybe, I do know at one point there was a AST.hs-boot
it's just that it makes Pugs.AST a right pain to compile optimized for me
stevan integral: yes there was an AST.hs-boot
integral: what would be involved in doing that? 12:54
integral doing what?
stevan changing src/Pugs/Types/* to modules
integral hmm, depends how many circular dependencies there are :-)
stevan you can always try it, if you have some time 12:55
integral *nod*
gaal_ stevan, pong, but i really have to go now for a few hours :(
&
stevan gaal_: just wanted to say Hack::Instances/Perl::MetaClass is mugwump's model and my code,... so any questions you can refer to me
gaal_: and the reason I dont use a global counter as well as the rand() is that rand() seemed to work fairly well 12:56
I created a 10000 key hash and found no dups 12:57
jabbot: seen mugwump 13:46
jabbot stevan: mugwump was seen 8 hours 3 minutes 34 seconds ago
pmichaud good morning (afternoon, evening) #perl6 14:33
kolibrie pmichaud: good morning 14:34
pmichaud hey, can someone briefly tell me what format pugs generates that is executable by parrot -- i.e., does it output PAST or PIR or ...? 14:35
iblech PIR
pmichaud is it easy to locate the PIR generator?
iblech src/Pugs/Compile/Parrot.hs 14:36
pmichaud excellent
many thanks
PerlJam greets pm 14:37
pmichaud greets, pj
PerlJam Are you working on the first stab p5 RE parser or the glob parser that you mentioned in your email to p6[cl]? 14:38
pmichaud pj: at the moment, no -- I'm working on a shift/reduce expression parser
w/operator precedence :) 14:39
PerlJam pmichaud: geez ... next thing you know, you'll be working on a perl6 grammar ;)
pmichaud It's likely I'll be evaluating P6 expressions by the end of the week
I already have the operator tables and precedence handling working 14:40
autrijus yo. 14:41
pmichaud so, it's just a matter of generating the resulting executable code (thus my question about PIR vs. PAST)
PerlJam wow. Didn't you start just yesterday?
autrijus pmichaud: I'm not sure PAST is in anywhere near usable
pmichaud autrijus: I agree with you -- I just wanted to verify it
autrijus pmichaud: I think though, it may make sense to simply represent PIR idioms as objects
pmichaud autrijus: I remember you were targeting PAST (however briefly) at one point and so I was going to follow suit
autrijus and build a bottom-up ast out of it 14:42
at least that's what I'm trying to do with PugsAST->PIR
by re-using the interpreter logic in a compiler setting
pmichaud well, I need to get just a bit farther on the expression parser and then I'll be looking at codegen in more detail :) 14:43
autrijus sure :) 14:44
btw, parsec's behaviour of addming a <commit> by default has been useful
i.e. it greatly speeds up parsing
pmichaud yup
autrijus however, there are some corner cases eg with
(int => 3)
where you need to basically lookahead to => as part of identifier/unary function token) 14:45
I'm handling them with explicit lookAhead calls... not sure if there's a cleaner way out of this
pmichaud yeah, I'm not sure there's a clean way out either
autoquoting is indeed a pain
autrijus yeah 14:46
pmichaud however, in my case I have the text immediately available, so when I see a => I can probably easily manipulate the operand/operator stacks and fix them as if we knew it was a quoted term beforehand
that shouldn't be too difficult 14:47
autrijus true, but you need to check that the left hand is an identifier-like too
~ => 3 # syntax error
if => 3 # works
pmichaud that actually makes it easier :-)
autrijus mm, nod 14:48
oh btw. any chance you know what should this print?
say (split /(..)*/, 1234567890);
sproingie i havent kept up with apocalypses. is => still a stringifying comma or does it make pairs now?
autrijus sproingie: pair
er, I mean
PerlJam sproingie: both
autrijus say (split /(..)*/, 1234567890).perl
pmichaud => is a pair constructor
PerlJam sproingie: => makes a pair with its LHS stringified 14:49
autrijus i.e. quantified captures in split//
pmichaud autrijus: no, I don't know what it should do
autrijus ok. pugs currently flattens each nested match
so '12', '34' etc
I'll ask p6l
pmichaud that seems reasonable
but yes, we need to re-examine split in light of the capture semantics
sproingie PerlJam: do those still automagically interpolate as lists? i still sometimes chain them, foo => bar => 123
autrijus that is (foo => (bar => 123)) 14:50
i.e. think => as consing
stevan sproingie: nested pairs :)
PerlJam doubts that nested pairs will flatten 14:51
pmichaud there's nothing to flatten there :-)
PerlJam pm: if you say @foo = a => b => 1, does @foo get ('a','b',1) or ('a', Pair('b',1)) 14:52
sproingie hm. not quite what I was used to. then again i've largely stopped doing the chaining thing
PerlJam ?
pmichaud pj: neither
@foo gets Pair('a', Pair('b', 1)) 14:53
PerlJam er, that's what I meant ...
stevan sproingie: we have all abused the "fat comma" to some degree, I am very guilty of it myself
pmichaud never abused the fat comma :-)
stevan pmichaud: you can admit it, we are all friends here :) 14:54
sproingie liked the stringifying aspect. harkened back to barewords, but in a safe sort of way
PerlJam I think Abigail is one of the people who is going to jump ship rather than use perl6 He abuses all sorts of features :)
autrijus finds stevan's metamodel work fascinating
sproingie: use <> to stringify now :)
stevan finds autrijus *work fascinating
sproingie oh i'm going to hate perl6. and love it. and oscillate madly between the two states.
autrijus PerlJam: Abigail will love [.]
[.] is da bomb
sproingie i already have that relationship with perl5, just not quite as passionately 14:55
stevan sproingie: passion is good :)
pmichaud perl 6 has lots more to love and hate, that's for sure :)
stevan autrijus: [.] is folding right? like foldl and foldr?
sproingie my swing with python is even flatter. it does what it does and nothing more or less
occasionally it surprises me, but rarely
autrijus stevan: no, reduce() is folding 14:56
stevan: [.] is folded method call
stevan ooooh
autrijus [.] $a, $b, $c
means
($a.$b).$c
PerlJam sproingie: so ... what do you think about guido's desire to make The One True Way list comprehensions?
sproingie wonders how far he can stretch the romantic metaphors. lisp: unrequited love, it just won't love him back
stevan autrijus: damn thats nice :)
sproingie: Lisp just gives you lots of hugs (), but no real love 14:57
sproingie PerlJam: i think he has no understanding of higher order functions. of course, python's lambda made that obvious. doesn't he want to get rid of that too?
PerlJam I believe he does
sproingie list comprehensions are great. show me how to compose a list comprehension as a function
they're a literal syntax. literals also need a functional equivalent 14:58
osfameron_ what's the difference between a comprehension and grep/map ?
PerlJam osfameron_: syntax :-)
pmichaud Is there an option to pugs to see the PIR it generates?
sproingie i would get that kind of prickly from lisp when i discovered my favorite feature was a macro
stevan aren't compressions usually lazy? 14:59
sproingie so the only way to extend it was with another macro
osfameron_ PerlJam: cool, I was worried I'd missed something...
sproingie now python's generator comprehensions are something interesting
autrijus pmichaud: yes... 15:00
$ ./pugs -CParrot -e 'say 123'; cat dump.ast
pmichaud perfect, thanks
(I know, I should probably rtfm, but...)
autrijus np :)
PerlJam I think guido is trying to take a page from perl by introducing any() and all() too (I believe he did that in the same message where he announced the demise of reduce/filter/etc)
autrijus yeah, there's this pugs -h :)
sproingie PerlJam: i'll give him credit for removing the rather broken reducee
autrijus pmichaud: the compiler backend is a mess, though. I eagerly awaits your codegen design :)
sproingie -e 15:01
pmichaud autrijus: me too :-)
PerlJam pm: Around here, I think it's usually that you need to wtfm rather than rtfm ;)
autrijus isn't it good that tfm is rw :)
PerlJam autrijus: indeed 15:02
stevan :D
autrijus pmichaud: want committer bit? :)
(so that tfm can be rw)
pmichaud autrijus: thanks, but not yet
PerlJam autrijus: When this is all done, you should write a little paper about the experience of handing out committer status so liberally.
pmichaud right now I'm working more on wtfc than wtfm :-)
integral wonders why AST.hs #includes Type/*.hs
autrijus integral: because .hs-boot is a greater evil than #include 15:08
integral it is? Having to maintain it?
autrijus that, and it triggers all sort of weird ghc bugs.
integral ah.
autrijus lots of "impossible"s
integral I'm looking at splitting some of the bigger files up a bit so GHC doesn't eat quite so much of my swap 15:09
autrijus so it ends up requiring razing src and rebuild
integral: I'd very much appreciate AST.hs refactoring
I've started by putting simple things to Types.hs
integral ok, cool :-)
autrijus but the codep is too strong :)
you're welcome to put even more things into Types.hs or AST/*
integral spots some easy targets like naturalOrRat
autrijus easy? :) 15:10
integral no cyclic dependencies ;-)
autrijus true :)
please be bold :)
and refactor mercilessly
integral switched to svk too
autrijus is happy that s/// and // Just Works now
integral svk++
autrijus it really feels like writing perl :)
pmichaud++
pmichaud I'm glad it actually works :-) 15:12
osfameron_ it's always nice when that happens ;-)
kolibrie it works for me not embedded, but all matches evaluate to true with PUGS_EMBED=parrot 15:15
autrijus kolibrie: really. 15:16
1 ~~ /2/
is true?
kolibrie yes
autrijus latest trunk?
kolibrie checked out latest parrot and pugs about six hours ago
ingy hola
autrijus kolibrie: oh. that's ancient!
try pugs trunk again :)
(parrot side doesn't matter)
yo integral
er 15:17
yo ingy
kolibrie autrijus: ok, I'll try again
ingy hi autrijus, how are classes coming?
autrijus ingy: stevan is doing most of the work
ingy cool
btw, I looked at LectroTest but couldn't get excited about it yet... 15:18
autrijus alright :)
ingy how is rules coming? 15:19
autrijus ingy: completely supported. 15:20
kolibrie autrijus: PARROT_PATH goes to parrot directory, or parrot/src?
autrijus well, as complete as PGE
ingy !
autrijus which is quite complete.
kolibrie: src
ingy are you doing native support?
autrijus yes.
er, wrong window
or external parrot
that works too 15:21
jhorwitz autrijus: i'm working out a new parrot/embed.h for leo. for now i'll add prototypes used by mod_parrot and pugs, and add public APIs where necessary. did you have any other ideas before i start? 15:22
autrijus no
that's pretty good
jhorwitz okay then. 15:23
autrijus kolibrie: I see what you mean. fixing
kolibrie autrijus: hovering mid air...
Juerd kolibrie: I'm guessing you're dutch 15:24
kolibrie Juerd: ik kan wel nederlands, maar ik ben amerikaner 15:25
Juerd kolibrie: Then why the dutch nickname? :)
ingy Juerd: I'm gueesing you're dutch
Juerd And "amerikaan" :)
("amerikaner" I think is afrikaans for american)
kolibrie Juerd: it's been a while since I was there
Juerd ingy: I'm not from this planet
ingy Juerd: I lay corrected 15:26
kolibrie Juerd: I wanted a nick I could probably get anywhere - and it seems to work -mostly
Juerd kolibrie: heh
kolibrie Juerd: if I go back six generations, I'm from Borssele, Zeeland 15:27
Juerd Ah, nuclear powerplant city 15:29
kolibrie lots of windmills, too
and that's about it 15:30
autrijus kolibrie: fixed. enjoy! 15:41
lumi Morning 15:43
autrijus yo.
kolibrie autrijus: ok, pulling now 15:44
lumi Should non-exhaustive patterns ever happen? 15:45
Juerd lumi: Not with normal non-code syntax
lumi: []* should be finite, even though there's a big bunch of nothingness matchable. 15:46
(That's (?:)* in Perl 5)
lumi Hmmm
Juerd svnbot6++
lumi Allow me to recontextualize, and berate myself for assuming waaay too many things
Juerd lumi: We couldn't forbid you even if we wanted. 15:47
lumi Should pugs die with a "non-exhaustive patterns in lambda" ever?
Specifically, '"foo" ~~ :foo' 15:49
I'm not sure what it should be doing, but prolly not dying with that?
theorbtw1 lumi: No, it should never die like that. 15:52
15:55 theorbtw1 is now known as theorbtwo
Corion gaal: pong # heh 15:55
autrijus lumi: write a test :) 15:59
PerlJam Would the :foo in lumi's example be a modifier on ~~ or just an error or what? 16:01
PerlJam votes error
autrijus obvious an error 16:02
PerlJam nothing is obvious to me anymore in perl6.
;-) 16:03
man ... my parrot compile is just zipping along while the pugs compile seems stalled.
Corion notes that the build worksforhim 16:04
lumi Ehm, anyone have p6 syntax for vim? 16:05
autrijus lumi: util/perl6.vim 16:06
Corion Does anyone know who changed the signature of getEnv from getEnv :: String -> IO String to getEnv :: String -> IO (Maybe String) ? 16:07
iblech lumi: There's util/perl6.vim, but I prefer the Perl 5 syntax highlighting over perl6.vim
PerlJam Corion: svn blame will tell you
integral hmm, I don't think it was me
Juerd For Perl 6, I don't use syntax highlighting
Corion PerlJam: I don't know which file to look at. Gaal fixed it in src/Pugs/Compat.hs, but I don't know what prompted that change.
wolverian Juerd: why not? 16:08
lumi Ack, I'll write it later, gots go
Juerd They first need to figure out a way to make %foo{"bar%baz{"quux"}"} and /foo { /bar/ and fail }/ look good.
kolibrie autrijus++ embedded parrot evaluates matches nicely
and s/// works awesome, too 16:09
thank you
autrijus kolibrie: np 16:11
kolibrie: and it's actually not slow :)
wolverian Juerd: I think you'll need perl6 integration in vim for that. 16:12
kolibrie autrijus: yep, seems pretty zippy 16:13
autrijus hmm :p6do in vim 16:14
Corion How do I convert IO String to String ? GHC complains that liftIO is not in scope - where would I find it or what should I use instead?
autrijus not hard really
Corion: you do it by using <-
liftIO can be imported with Control.Monad.Trans (liftIO) 16:15
Corion autrijus: D'oh ... I've done it so often, but still don't remember ;)
Now it at least compiles :)
kolibrie could p6 grammars be used to define syntax highlighting - then highlighting could be perfect 16:16
autrijus kolibrie: not sure vim likes that
but sure, write a p6rules or parsec -> vimsyntax compiler! 16:17
glory is yours!
kolibrie autrijus: it is a little different from the current .vim format
ah, the challenge
maybe 16:18
meppl guten abend
Corion autrijus: Should Compat.hs be split up into Compat.hs and Compat.Win32.hs ? I'm hacking in more and more Win32 specifics in the non-PUGS_HAVE_POSIX part ... On the other hand, nobody has complained yet ... 16:19
autrijus Corion: I'm pretty sure the only nonposix is win32 ;)
but if you want to refactor, that is fine by me too 16:20
Corion autrijus: Ah, wait until we need to port to Symbian, CE and V;S ;)
err - VMS
But I'll postpone refactoring until then
;)
putter Status update: I'm currently banging on Perl6::Rules t/*.t. Early "post"mortem - blech. Substantial rewriting was needed, not just to convert p5 to p6 (expected), 16:21
but because the regexp dialect was very outdated (unexpected:). Perhaps t/rules/Disabled should have been done first. Or brand new tests written.
autrijus putter: btw, t/perl6.t is almost there now (from Disabled)
been working at it
Corion Oooo - svnbot6++ # showing the comitter
wolverian svnbot6-- # said 2931 twice 16:22
putter autrijus: good to know. sorry - I should have followed your suggestion and done that first.
Corion Oh. And BTW, GHC is _very_ picky when linking DLL symbols - the case must match exactly.
wolverian: But two different lines from the comitt message, so that's OK
BTW, any other Win32 users here with a specific feature they are missing? I'm mostly working on implementing the Win32 equivalents of what POSIX provides, but maybe somebody can tell me a better thing to focus on ;) 16:23
wolverian Corion: no, he said the same two lines at 19:05 16:24
iblech wolverian: That's because autrijus' commit log was a multi-line message (it'd be simple to display the author and revnum only in the first line of each commit, but then we'd lose horizontal alignment)
wolverian same to iblech.
I can paste them if you want. :p
Corion wolverian: We trust you there ;)
wolverian that was.. 19 minutes ago.
osfameron_ Corion: would be great if threads/forking abstraction works well on Win32
integral believes wolverian ;-) 16:25
wolverian OH FOR THE LOVE OF GOD
:(
Corion osfameron_: Ugh. That part is with GHC ;)
iblech wolverian: You're right... Shouldn't happen
Corion osfameron_: Threads work, and forking never will.
osfameron_ Corion: yeah, and it's a biggy...
wolverian iblech: yay!
iblech And svnbot6 showed the two commits in reverse, strange 16:26
Corion And personally, I'll never touch fork emulation. It doesn't work with Perl5 and it won't with Perl6.
iblech is investigating
putter Before I go back and build pugs embedded, does anyone recognize: pugs> "a" ~~ /a/
pugs: waitForProcess: interrupted (Interrupted system call)
?
integral *blink* it did it again!
Corion putter: What platform? Sounds like your Pugs got sent a signal...
Looks like weird polling logic to me - it seems to use a sliding window instead of checking off stuff as "seen" and not posting it again. 16:27
putter Corion: x86_64, parrot HEAD, nonembedded,
autrijus jhorwitz: I want warnings and error control
jhorwitz: i.e. warnings.h should be exported
Corion putter: I don't recognize x86_64 as OS. 16:28
putter Corion: ;), FC3 current
Corion putter: But it still sounds like Pugs (resp. Haskell) got sent a signal and didn't handle it
putter Corion: tnx. I'll add it to my stack of potential bugs to check later, and try embedded. 16:29
Corion putter: Maybe because the interactive loop of readline/pugs waits for signals... 16:30
I don't have/use readline
gaal: I've unified your patch with the main (Win32) version of getEnv
iblech [svnbot6] I understand. It seems svn.openfoundry.org sent the log messages in reverse (once). So, $cur_svnrev was set to the smaller one (r2931). Then, 300s later, svnbot6 looked for new commits, and treated r2932 as new, because $cur_svnref was, at that time, r2931. 16:31
wolverian would it be smart to just refuse to decrease $cur_svnrev? 16:33
iblech wolverian: I think so. Unfortunately, we can't test easily if this fix works, as svn.openfoundry.org "normally" sends the messages in the correct order 16:34
wolverian spoof some packets? :) 16:35
iblech Not so easy with TCP ;) -- Fix commited as r2934 16:36
Limbic_Region seen Jonathan 16:45
jabbot Limbic_Region: Jonathan was seen 1 days 14 hours 51 minutes 47 seconds ago
Limbic_Region seen Jonathan_
jabbot Limbic_Region: Jonathan_ was seen 6 days 23 hours 37 minutes 16 seconds ago
Limbic_Region hmmm - I wonder if he knows his nightly binary builds of Pugs for Win32 hasn't worked since the 7th 16:46
autrijus _metaperl: 16:52
www.metaperl.com/talks/p6/hangman-e...lide2.html
the any<> is unescaped
you want &lt; &gt;
_metaperl oh thanks 16:53
autrijus np. great slides btw.
putter autrijus: Damian's email "Perl 6 rules capturing semantics" which you linked from the Journal is quite useful, and (recently) wasn't in pugs. Should it be?
autrijus putter: what wasn't in pugs?
putter www.nntp.perl.org/group/perl.perl6....uage/20985 16:54
pmichaud Damian is working on adding the document to S05
autrijus pmichaud: I added some error handling to pge
so perl6.t is running...
/a\N+?c/
segfaults badly
pmichaud ...that segfaults?
(is there a "good" segfault? ) 16:55
:-)
autrijus no :)
try it with your demo.pir?
it segfaults both here and there
ajs should "sub foo($a,$b){...} foo(1,2)" work? I keep getting an error invoking anything I define that takes more than one arg...
pmichaud indeed it does
okay, working on it
ajs kill('SEGV',$$) would be a "good segfault" ;-)
putter ok. it's rather critical for understanding rules.
16:56 Aankh|Clone is now known as Aankhen``
pmichaud if you think it's critical for understanding how to use rules, think of what it was like trying to write PGE w/o it :-) 16:56
Hmm, I wonder if it's the '?' that's doing it 16:57
Corion r2933 - datenzoo.de/pugs/win2k.html
pmichaud runs a trace 16:58
putter sigh. When doing PUGS_EMBED=parrot, does anyone recognize this one...
pugs> "a" ~~ /a/
pugs: internal error: adjustor creation not supported on this platform 16:59
Please report this as a bug to [email@hidden.address]
?
autrijus oh wow.
x86 64?
putter t
autrijus 6.4.0?
putter t
autrijus wonders if ghc-cvs helps
but, well, try using external parrot :(
obra autrijus: broken links in your last journal post
putter autrijus: re ghc-cvs... could be. would save me always doing +RTS -A500M... 17:00
autrijus: re external... that was the previous "waitForProcess: interrupted (Interrupted system call)" failure case ;) 17:01
autrijus obra: fixed
autrijus mumbles something about x8664.
autrijus mumbles something about ghc as well
try ghc cvs
and if it fails too, tell them to fix it :) 17:02
putter autrijus: will do. thanks.
putter remembers thinking "_64 porting problems aren't a _problem_, they're an _opportunity_ for writing more patches"... sigh. 17:04
er, just before deciding to buy one.
autrijus /a\N*?c/ segfaults too, but that's the same cause I think
pmichaud autrijus: /a\N+c/ now fixed, r8046
autrijus woot.
pmichaud oops,
autrijus: /a\N+?c/ now fixed, r8046
autrijus pmichaud++ fast 17:05
pmichaud forgot a parameter in the minimal match for quantified character classes
(try saying that quickly!)
autrijus I think I see an infinite loop... 17:06
pmichaud in PGE?
autrijus "a--" ~~ /^[a?b?]*$/
pmichaud yes 17:07
PGE doesn't handle repeats of zero-length subpatterns yet
autrijus k
Limbic_Region did anyone change any unlink() code recently?
pmichaud it's coming -- basically each group has to know to repeat only once if the position hasn't updated
autrijus but other than that, it's faring quite well :)
(huge numbers of failures still, but hey) 17:08
Corion Limbic_Region: Not that I know of
pmichaud what sorts of failures?
Limbic_Region Corion - looking into why a couple io tests are failing (according to your last smoke)
at least 1 I am responsible for 17:09
and it worked fine on my box
autrijus hm
Limbic_Region does the smoke run in order?
autrijus this is abysmally slow
"bbbbXcXaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa" ~~ /\NX(\N+)+X/
PCRE finishes it in 3 seconds
took PGE more than 1min and counting
I breaked it
Corion Limbic_Region: Yes it runs in order, but unlink on Win32 doesn't want to remove files still held open by the process.
Limbic_Region Corion - I know
pmichaud PGE's running unoptimized at the moment. Purpose was to get something that works that we can evaluate :)
Limbic_Region there is an explicit close before the unlink 17:10
Corion Hmmm...
Limbic_Region and it worked fine on my Win32 build
Corion Let's wait until the next smoke finishes :)
Limbic_Region ok
autrijus I keep seeing 17:11
SO=..
what is this?
Limbic_Region pasm
;-)
pmichaud it's in quantified ranges code somewhere
must be some debugging trace I put in
autrijus ok.
theorbtw1 It's letting you know that your significant other is one dot sort of a full ellipsis.
autrijus also try feeding this to pge:
rule /^\N**{3..4}(\N+?)$1$/
it's segfault as well (I think) 17:12
pmichaud works for me
at least as far as compiling the rule
autrijus try matching it with "foobarbar"
pmichaud yup, that segfaults 17:13
mauke what's \N again?
pmichaud not a newline
knewt reads the latest in the circular dereference thread and shakes his head
mauke ah, thanks
_metaperl let's really get English like: my @basket has <apples oranges bananas> ;
autrijus knewt: /ignore ;)
_metaperl my $age is 35; 17:14
knewt autrijus: tried joining #perl6 on rhizo for a moment, and found it to exist but with a channel key 17:15
autrijus not having anything to do with me
theorbtw1 was wondering earlier...
jhorwitz autrijus: back from lunch. you need Parrot_warn, etc.?
theorbtw1 wonders what it was that he was wondering earlier.
knewt yeah. when i saw the error i remembered the channel was actually over here
pmichaud autrijus: looks like it's a problem with the $ anchor
theorbtw1 Oh -- would it be possible to emulate the do { ... ; ... ; ... ; } form for non-IO monads in p6 / pugs? 17:16
; used to be an operator (and thus overloadable), but I think the recent [...] thing got rid of that.
pmichaud ^\N**{3..4}(\N+)$1 works
autrijus pmichaud: can you take the SO=.. away as well?
pmichaud done -- just haven't committed yet 17:17
autrijus okie
autrijus waits for pmichaud's commit for a rerun
.pbc++ # cross-platform, requires no recompilation
vm++ # damn good idea
pmichaud Committed w/o S0=, r8047
^\N**{3..4}(.+)$1$ works 17:18
hmmmm
and rule ^\N**{3..4}(\N+)$1$ works for me now for some reason 17:19
Limbic_Region autrijus - .pbc files work universally but depending on 32/64 bit may need to be re-aligned. The plan was to provide a utility so that you could convert 1 time instead of incurring the miniscule penalty each time it ran
Corion r2935 - datenzoo.de/pugs/win2k.html 17:20
Limbic_Region regardless - transparent to end user
pmichaud oh, wait, minimal match
Limbic_Region Corion - still failing
Corion Limbic_Region: Which one is the test you're looking at? 17:21
Limbic_Region t/pugsbugs/io_finalize_part2.t for instance
and t/builtins/io/io.t # which it was modeled after
pmichaud hmm, another bug in lazy cclasses... 17:22
autrijus pmichaud: can we backref to submatch?
Limbic_Region the weird thing is the test closing the file right before the unlink succedes
pmichaud autrijus: not yet in PGE
autrijus k
pmichaud autrijus: once I have it reasonably embedded with perl one will be able to do $1[0] for a backref to submatch
Corion Limbic_Region: Hmmm. Interesting ...
pmichaud (or any other variable name)
Limbic_Region It worked when I wrote/tested it 17:23
Corion Limbic_Region: I guess I'll look if I can put a trace() call into Pugs at the "right" position so I can see if/when the Pugs FileHandle object gets destroyed and if it flushes its buffers...
(although say() should never buffer)
Limbic_Region Corion - that's not the bug I am worried about right now 17:24
Corion Limbic_Region: Which bug is it then?
Limbic_Region that bug is something I think needs to be fleshed out with Larry and Chip
ok - let me start at the beginning
pmichaud I bet I'm going beyond the string boundary there
Limbic_Region I noticed that a filehandle wasn't flushed when the program ended (assuming no explicit close)
autrijus pmichaud: 17:25
/[:i a]b/
why is it bad?
pmichaud :i isn't implemented yet
Limbic_Region since I am not sure if p6 and/or parrot is guaranteeing timely destruction - going out of scope isn't necessarily enough to force the flush
pmichaud none of the case-insensitive things are implemented
Limbic_Region regardless - program ending should be
so I wrote a test to show it as a bug
pmichaud because I didn't have anything in Parrot to be able to do case-insensitive matches until this morning :-)
Limbic_Region but now in addition to that bug, the file is no longer being removed
pmichaud (oh, I could do a bunch of tolowers, I suspect) 17:26
Limbic_Region that is something that *was* working correctly
autrijus pmichaud: ok. parsefails:
autrijus.org/pge_run.txt
Limbic_Region Corion - got it?
pmichaud I don' t have character classes yet either (same reason)
or lookaround
Corion Limbic_Region: That still doesn't really tell me which test is now failing for you that previously did not. I assume that you are talking about test #4 in t/pugsbugs/io_finalize_part2.t, correct? 17:27
pmichaud I can probably get it to parse (but not honor) :i
Limbic_Region Corion - only 1 of the tests in the file should have failed
now 2 are
the last one should pass but fails 17:28
as does the last one in t/builtins/io/io.t
which is also unlinking a file (that has been closed)
wolverian Limbic_Region: there's a parrot op to force a dod sweep, and you can (if not now, eventually) mark which objects you want to be timely destructed. there is no default timely destruction in parrot, and perl6 doesn't require it 17:29
Corion Limbic_Region: Ah. So. I was correct in the presumption that you are talking about test #4 in t/pugsbugs/io_finalize_part2.t . This sounds to me as if Win32::API close() is not called/executed at the "right" time.
wolverian: We are not on parrot.
wolverian that's a good point.
:)
Corion wolverian: Of course does Perl6 require it, if you're trying to unlink a file on Win32.
Limbic_Region wolverian - besides, just because it is possible doesn't mean someone has spec'd what p6 will do
Corion - yes, test4 - which should not be failing now and was not failing in the past 17:30
wolverian I distinctly remember seeing a p6i post with larry saying perl6 doesn't need timely destruction (by default)
Limbic_Region wolverian - so do I but that was some time ago - he changes his mind you know
wolverian or on some other mailing list, I guess.
Limbic_Region: true enough. :)
pmichaud autrijus: what should PGE do when a parse fails? 17:31
right now it just kinda slops around :)
Corion wolverian: I don't care what Larry says the default is, but if close() and unlink(); is ever to work on Win32 it will either need special-casing magic or timely destruction/buffer-flushing.
pmichaud should it return an undef, or as larry puts it, an "interesting" undef?
Limbic_Region Corion - my mistake, your mind dump of what the problem might be sounded like you were attempting to solve the original problem I wrote the bug test for
we are on the same sheet of music
wolverian Corion: yes, of course. the objects that need timely destruction should be marked as such.
_metaperl what does "PGE" mean?
Limbic_Region I am just wondering why it mysteriously stopped working
pmichaud PGE = Parrot Grammar Engine
Limbic_Region Grammar Engine
pmichaud it's the piece that implements perl 6 rules and string matching
autrijus pmichaud: I made it throw Exception 17:32
_metaperl oh, so if ruby/python compile to parrot, they will be pcre... interesting
Corion Limbic_Region: Ah. Well, maybe "svn blame" on Prim.hs or src/Pugs/Types/Handle.hs can tell you something.
Limbic_Region Corion - I am out of the svn/svk game for sometime
no net connection from the new house until at least this saturday and no-go from work 17:33
Corion Limbic_Region: Ow! Well, maybe I find something that either allows us to delete open files or maybe I even find the bug there.
elmex did someone test parrot/perl6 on HURD ? ;) 17:34
Corion Hmmm. Pugs should get Win32::OLE and/or Win32::COM as soon as possible. Maybe GHC provides us with that already?
autrijus Corion: look at Win32.* 17:35
I think GHC gets us that already
Corion autrijus: So Just some glue magic is needed. Plus magic reflection, like $AUTOLOAD.
pmichaud autrijus: ^\N**{3..4}(\N+)$1$ now works
er 17:36
autrijus: ^\N**{3..4}(\N+?)$1$ now works
(I keep forgetting those ?'s)
autrijus great
pmichaud now to :i
Corion Hmmm. Should I give Pugs CreateHardLink() ? That would make Pugs only work on Win2k upwards ... 17:37
Limbic_Region Corion - well I am still trying to learn p6 so I can't be much help from the Haskell side
Limbic_Region is just poking people when things don't look right
and attempting to find bugs along the way of his self-edumacation
autrijus Corion: just go ahead :) let it fail on win9x 17:38
Corion: oh and it works with ntfs only
Corion autrijus: It would also fail on WinNT, which is what I have at work ;)
autrijus with ntfs you can do symlinks on dirs too
I think it can be made to work on winnt
see my cpan modules:
Win32::Hardlink Win32::Symlink
Corion autrijus: True - I forgot about the non-NTFS part ...
Oh. And Win32 has LZOpen(), LZClose() and LZRead() :)) 17:39
autrijus wonders how long it will take to write another 100ish perl6 modules
Corion ... it's been a long time since I looked through the Win32 API ;) 17:40
Limbic_Region autrijus - I was smart, I only wrote 1 public p5 module
won't take me long at all to re-write mine
Limbic_Region is waiting for proper OO though 17:41
out of curiosity - are there any examples of doing tied variables? 17:43
autrijus no..
stevan_ Limbic_Region: did you see Hack::Instances :)
Limbic_Region stevan_ (did you see "proper") :P 17:44
stevan_ proper?
Limbic_Region IOW - I am being lazy and making excuses
stevan_ LOL 17:45
Hack::Instances is kinda fragile, but your external code should be able to stay (mostly) the same
Limbic_Region ok so in p6 tied variables will be done with traits right as in my $debbie does Dallas?
stevan_ the internals will only need re-writing
pasteling "iblech" at 80.81.4.211 pasted "Limbic_Region: tied scalars are simple" (8 lines, 168B) at sial.org/pbot/10051 17:46
stevan_ anyway, lunch time &'
Limbic_Region interesting 17:47
pmichaud autrijus: [:i a]b now parses
r 8049
iblech the Proxy object is specced in S06, IIRC 17:48
pmichaud it gets a one-line mention 17:49
oh, as "Proxy" it's one-line
it's actually covered in more detail at Lvalue subs
Limbic_Region iblech - I am going to spend some time reading my Perl6::Bible
Limbic_Region needs to spend some more time reading The Bible too 17:50
iblech (BTW, FYI, at freepan.org/repos/iblech/Perl6-Bible there's a fully working p6bible and perl6doc (both written in Perl 6, though perl6doc calls perldoc after having set some %*ENV vars) 17:51
Limbic_Region iblech
grrr
iblech plus plus
ninereasons iblech++ # always making such useful things
Corion Heh. Looking at the length.t failure, it seems that PCRE doesn't cope well with unicode?
pmichaud pcre doesn't cope well with unicode in some respects, no 17:52
Corion pmichaud: Ah, so then there is no sense in looking into that ;)
iblech Stevan++ # /me's waiting for a POD::Simple written in Perl 6 :)
pmichaud well, I've run into problems with pcre and unicode in the past
integral hmm, what's the word you put after `import` to not import something? 17:53
Corion integral: You put "--" before "import" :)
integral heh. I mean just one function from the module
pmichaud import "something" but not really;
Corion integral: import qualified Module.Name
pmichaud :-)
integral hiding :-)
Corion And then you use Module.Name.Function
(or, as I do, localFunction = Module.Name.function) 17:54
autrijus Corion: the fault is mine. sec
Corion: basically the matchFrom and matchTo returned was bytes 17:56
Corion: but I'm too tired to fix. maybe you can take a try 17:57
it's in Prim.hs line 766
Corion autrijus: I'll look at that ;)
autrijus the key is that the offsets as returned from rv
are in bytes
even though PCRE correctly matches against unicode
i.e. the match is unicode
but the offsets are in bytes
p6 demands that we recalculate offsets in chars
one way to do this is to do the genericDrop etc to cs' 17:58
where cs' = encodeUTF8 cs
then recalculate from and len with genericLength of the portion dropped
pmichaud: btw, you were using latin1 << and >> in ascii encoding 18:00
pmichaud: so I texanised them
pmichaud yeah, I noticed that -- no problem 18:01
autrijus k :)
pmichaud I haven't got a good utf-8 editor here
earlier this week I was wondering what encoding we would be using for perl programs
theorbtw1 At present, pugs /requires/ that the input be utf8. 18:02
I think that's a very good default.
pmichaud yes, but there may be a goodly number of editors that cannot edit/display utf8
we may discover that someone will read a file into an editor and have it become garbage upon save
well, not garbage, but something that perl no longer understands 18:03
Juerd must be careful to stay friendly
integral 7-bit ascii is still a subset of utf-8...
ingy autrijus: 11:03 < purl> couldn't get the headlines:
wagner.elixus.org/~autrijus/cpan.rdf wasn't successful
Juerd All ASCII is 7-bit
ingy is that you?
autrijus ingy: wagner is offline 18:04
maybe permanently
Juerd integral: I think it's better described as: ascii is represented with unmodified bytes in utf-8
autrijus be patient while we deal
ingy ouch
integral Juerd: true
ingy autrijus: I'm a Perl programmer, but OK
theorbtw1 A pure ASCII file is valid utf-8, with identical semantics. 18:06
(By definition.)
autrijus pmichaud: yeah. problem is, a parser needs to be two-pass
to accurately accomodate for both utf8 and latin1
pmichaud sure, but let's say I grab a file from cpan that has utf-8 characters in it into my non-utf8 editor
autrijus pmichaud: either that, or allocate a fixsize buffer to guess
pmichaud or, my editor understands utf-8, and happily converts the utf-8 sequences into iso-8859-1 for me 18:07
autrijus pmichaud: the latter is rare
pmichaud: the first is quite possible.
(note that I was against the utf8 form of << >> and Y.)
(mostly because they are not in big5 ;))
pmichaud oh, I don't mind the utf8 forms -- I'm just curious how we'll end up dealing with them when it's all said and done :)
autrijus I still am, but it's part of spec, so I implement the utf8 form. :)
theorbtw1 pmichaud, if the editor is sane, you'll see two-byte sequences where there should be unicode chars, and if you don't touch them, you'll be fine. 18:08
autrijus theorbtw1: the "don't touch them" part is hard
integral maybe just easier to bail, and print "You have a weird idea of utf-8, run perl6-fix-non-utf-8-encodings"
autrijus because you may be writing new code.
theorbtw1 Big5 doesn't have Ā„? That seems like an odd omission...
pmichaud theorbtw1: yes, I know one can probably get by with simply not touching the "funny looking characters", but I'm sure there will be a lot of programmers who struggle with it for a time 18:09
autrijus theorbtw1: it has ļæ„
pmichaud for me personally it won't be a problem -- I know about charsets and char encodings :-)
ajs autrijus: keep in mind that the unicode << and >> are ISO-8859-1 as well as Unicode, which is why Larry suggested them. 18:10
I don't know of any editor that does not support 8859-1
pmichaud yes, but ISO-8859-1 is not utf-8
ajs yes....
pmichaud so, the parser has to have some idea of what encoding it's seeing when it's reading the file, or we have to decree that all perl programs are utf-8 or something like that 18:11
autrijus ajs: notepad on win32, chinese edition
does not support latin1.
I think it's the same for CJK. 18:12
also the Yen sign if typed in a JIS keyboard
will become backslash sign \ in ascii.
so I'd be very wary of using Yen sign in latin1.
Corion autrijus: Does wordpad support anything good? Because notepad.exe also doesn't support unix newlines. 18:13
autrijus Corion: the same, I think.
you can use utf8 or codepage
Corion autrijus: Ugh
PerlJam is *so* glad that Larry is the language designer ;)
autrijus and the default codepage is not latin1 for CJK people.
(actually you can also use ucs2.)
ajs If your editor spits out ASCII \ for yen, then I'm not sure there's much we can do for you. I honestly think you're better off writing in your native set and using a special module like "encoding" in Perl 5 (though hopefully better) 18:16
autrijus ajs: that's why JPerl :) 18:17
theorbtw1 wonders how hard it would be to support something like 'use encoding "foo"'.
autrijus ponders ļ¼„ļ½˜ļ¼œļ½™ļ¼žć€€ļæ„怀ļ¼„ļ½šļ¼»ļ¼„ļ½—ļ¼½
Corion PERL6OPT=-Mdecode=... :)
theorbtw1 autrijus, those are with the fullwidth chars?
autrijus theorbtw1: right.
theorbtw1: there's no halfwidth "high-bit" special symbols in big5 18:18
ajs remember that perl5 "encoding" is not sufficient, but hopefully Perl 6 won't have the same problems with unparsable text after the pragma.
autrijus theorbtw1: all of them are fullwidth
theorbtw1 Eh, ajs?
kolibrie autrijus: what font do I need to see those characters correctly?
autrijus kolibrie: I don't know... arphic maybe
ajs lemme get the quote from the docs....
iblech Hm... it seems P5 regexes, when inside a given { when rx:P5/.../ {...} }, don't capture correctly, but I was yet unable to deduca a test case
kolibrie autrijus: thanks, I'll try 18:19
theorbtw1 /who
Grr.
18:19 theorbtw1 is now known as theorbtwo
ajs Oh nevermind. Perl 5.8.1 fixed the encoding / Big-5 scan-ahead problme 18:20
er problem
autrijus right.
iblech P6 regexes seem to not work, either (when inside a bigger program)... /me continues to investigate 18:21
kolibrie autrijus: ttf-arphic-ukai or ttf-arphic-uming or both?
autrijus pmichaud: hey, is this legal?
rx /foo/
rx :P5/foo/
kolibrie: ming may be better
either would do
pmichaud well, it's going to be rx :perl5 /foo/
kolibrie autrijus: ok
autrijus er, larry hinted that it's changing to :P5 18:22
is that retracted?
pmichaud actually, it was :p5 for some time and then I noticed it changed in the synopsis
autrijus mumbles something about sunspot activity
pmichaud: but anyway. is
rx /foo/
legal?
pmichaud yes, why? 18:23
I'm pretty sure it is
you mean as a statement by itself?
or...
?
autrijus the space 18:24
between rx and /
I had it as legal 18:25
not terribly sure
pmichaud S05.pod: $rule = rx :g :w :i / my name is (.*) /;
iblech autrijus: Both P5 and P6 regexes capture correctly (and promote $1, etc. correctly), but $/ and $1, $2, etc. are not lexical, it seems -- you know that bug? 18:28
autrijus iblech: it's a XXX
iblech autrijus: Ok
autrijus not really sure how to do that 18:29
we can let each scope introduce its own $/
or, only add $/ to the pad when match happens
or, make it effectively a temp()
the last one is easiest
pmichaud I'd add $/ to the pad when match happens
stevan iblech: re: POD::Simple, what do you need?
autrijus pmichaud: right, but runtime pad?
sub foo { m//; return -> { $/ } } 18:30
maybe a static analysis at parsing time
pmichaud yeah, probably
iblech stevan: Currently, my perl6doc is just a wrapper around perldoc. It'd be cool if we had a pod2man/pod2text in pure Perl 6 :)
autrijus ok, I'll go with static analysis then
good that eval "m//" will terminate at the "" scope 18:31
kolibrie autrijus: ttf-arphic-uming installed - do I have to restart my terminal, or will characters show up properly if you were to send again? 18:35
autrijus kolibrie: ļ¼­ļ¼ļ¼„ļ¼ļ¼£Ā² 18:37
putter re static analysis: eval("m//") ?
autrijus putter: eval triggers reparse
putter k
autrijus putter: and the scope is in ""
so all's good
I worry about (m// + m//) though.
does the second m// see $/ from the first?
i.e. is the pad creation pushed to statement level? 18:38
pmichaud once a new match is started, $/ is bound to it
autrijus ok. but what about
pmichaud i.e., any $/ used in a match refers to the current match object
autrijus return (->{$/}, m//);
pmichaud (the one being built)
autrijus does the $/ bind to the pad created by m// ?
kolibrie autrijus: hmm. looks different, but still a few missing characters 18:39
pmichaud autrijus: I dunno about that one -- I'd have to review my notes a bit 18:40
putter wonders if "defined $1" sheds any light
Khisanth $/ is still kicking around in p6? I thought all those punctionation type variables were going the way of the dodo... 18:42
Corion Khisanth: $/ is now per =$fh , so I doubt that it's still called $/
Khisanth heh $fh./
how is that for ugly? :) 18:43
pmichaud $/ is the "replacement" for $0
theorbtwo $/ in p6 is something completely different from $/ in p5.
autrijus $fh.irs I think
irs is a weird name.
pmichaud $/ is the result of the last match
theorbtwo Reminds me of taxes, autrijus.
Corion IRS is perfectly fine if you know awk ;(
err - :)
theorbtwo (The US Internal Revenue Service.)
Corion (and shell, of course)
clkao inland revenue!
Corion Except there it's called "$IFS" :) 18:44
autrijus clkao: hey. congrats for svk1.0
Khisanth irs conjures too much hatred
clkao autrijus: thank you! and party!
autrijus: obra is flying over for the party. you can, too!
autrijus ... 18:45
I guess I can.
clkao no, but, have a party in taipei maybe
ha, for sure? it's this thursday 18:46
autrijus like, 48 hours from now?
mmm. challenging
clkao correct
autrijus but not infeasible
what's usd$2k among friends? ;)
Aankhen`` What the...? 18:47
Notices?
Aankhen`` cries.
autrijus yeah, a bot written in perl6
Aankhen`` msg #perl6 wasn't good enough?
iblech Aankhen``: The IRC RFC mandates that bots always use /NOTICEs to send stuff
autrijus oh there's an IRC RFC? 18:48
Aankhen`` Sadly, notices are rather annoying.
autrijus also it's not logged
maybe for the better.
obra autrijus: frequent flyer miles
Aankhen`` Anyway, it's cool to hear of the new bot. :-)
iblech autrijus: Yes, there'is RFC 1459 (from 1991 IIRC), and there're four new RFCs from 2000 (RFCs 2810 - 2814) 18:49
autrijus ah. didn't know about the 28xx one
kolibrie autrijus: did I read that right - 'rule' is now implemented? 18:50
autrijus kolibrie: rule{} yes.
kolibrie very nice
PerlJam yeah, autrijus++ and pmichaud++
iblech Aankhen``: Plus, as there's now a working Net::IRC, writing new bots is very simple :) You may want to look at examples/network/{svnbot,logbot,seenbot}.p6
Aankhen`` Will do. 18:51
autrijus pmichaud: rules live in their own namespace right?
pmichaud: i.e. you can have a grammar Foo, a rule Foo, a class Foo and a role Foo
pmichaud I haven't worked out the namespace issues yet
Aankhen`` I was recently called on to write a bot for someone, so maybe it'll end up being in Perl 6... :-D
Although I dunno. It's not likely they'll be able to run Pugs, since they'll be at the mercy of the server admin. 18:52
pmichaud can we have a class Foo and a package Foo?
PerlJam modules, classes and packages all share the same namespace I thought 18:53
pmichaud if that's the case, then grammars are there too :) 18:54
PerlJam er, no pun intended
pm: that makes sense to me.
pmichaud supposedly, grammar : rule -> class : method
PerlJam right, so grammar Foo and rule Foo can live peacefully together 18:55
pmichaud either that or grammar : rule -> module : method
putter Re eval... thinking about Ruby's "functions can participate in class declaration" (eg, "has $.foo is rw" is "attr_accessor :foo", 18:56
where atter_accessor is an ordinary function doing something like class_eval("def foo ..."))
how does one do this in p6? macros? 18:57
Juerd pmichaud: Yes, there can be a class Foo and a package Foo, but they share the same namespace :) 18:58
iblech putter: I think has() could be implemented as a macro, yes. Because of BEGIN{}, Perl programs can freely "chat" with the compiler, so a thing like attr_accessor is relatively simple implemented (I think)
wolverian don't we have a Class.methods.push( { ... } ) or so? :) 19:00
pmichaud juerd: in that case I'd say that grammar Foo shares the same namespace as well :)
stevan wolverian: maybe with the metaclases
pmichaud a rule is just a funny sort of sub
wolverian stevan: that'd make sense.
pmichaud that's certainly the way rules are implemented in parrot / PGE :) 19:01
putter iblech: so the next question is whether/how one can modify a class decl at later runtime. In ruby its class_eval or somesuch, in python $class.something_more_specialized_like_define_method...
wolverian ask p6l? 19:02
Juerd pmichaud: I've been wondering the past few days if rules need their own sigils
pmichaud: And perhaps not be called with <...> but be "interpolated" with their sigils, accepting arguments the same way a &sub does 19:03
pmichaud I think it's best to leave them as & and subs
PerlJam Juerd: no way!
Juerd PerlJam: Why?
It's a very major, important and omnipresent data type
pmichaud and the <...> for subrules works extremely well 19:04
Juerd pmichaud: Then perhaps, and this may sound horrible, <> should be the sigil :)
pmichaud syntactically it's almost exactly what we want. In that sense, you could think of <...> as being the rule's sigil :)
Juerd Outside of rules
putter Ruby's eval, method_eval, class_eval has been much more flexible than the python equivalents. though apparently they are slated for downgrading (eg, new vars decled in eval have only eval in scope).
pmichaud no, I much prefer having <...> as the {'...'} replacement
Juerd If they share namespace, they mustn't clash with ANY other identifier 19:05
pmichaud: I badly want '' to be {''}'s replacement
pmichaud: And $foo to be <$foo>'s replacement
pmichaud: And "$foo" to be $foo's replacement
It makes everything MUCH cleaner
pmichaud I'm not sure I agree that it makes everything MUCH cleaner
Juerd from simple matching, like '(' \w+ ')' to anything involving $foo
autrijus ok. < is now a sigil in Pugs's pad for rules :) 19:06
autrijus has been listening.
Juerd If <> calls rules, then <$foo> should call the rule NAMED $foo
pmichaud <...> is overloaded with lots of things 19:07
Juerd i.e. $key := (\w+) = $value := (<$key>)
pmichaud: Yes, and interpolation of strings or subrules doesn't belong there, imo 19:08
Certainly literals don't belong in <>
pmichaud in this instance, you're wanting <$key> to be the rule that was assigned to $key earlier in the match?
Juerd Oh, damn, I read your {''} as <''>
in rules
For clarity, I want:
<''> => ''
<$foo> => $foo
$foo => "$foo" 19:09
pmichaud you want quotes to be metachars in rules?
Juerd Yes! Very much!
PerlJam no way!
Juerd pmichaud: And <$key> should call <addr> if \w+ matched "addr"
PerlJam: For anything's sake, stop the lame "no way" replies, and bring in some reasoning. 19:10
pmichaud I like <$key>'s current interpretation better
theorbtwo Hm, I'm not sure I try to match on ' or " in REs any more or less then < and >.
PerlJam Juerd: sorry I'm talking IRL
Juerd PerlJam: Then stfu
pmichaud <$key> treats the value of $key as a subrule to be matched
theorbtwo In any case, I suspect I'll use p5 REs for simple ones, and p6 REs for complicated ones.
Juerd pmichaud: To be honest, I don't like <$foo> calling a rule named $foo
pmichaud: But I *hate* <''>, $foo and <$foo> as they are right now 19:11
pmichaud what exactly do you hate about them?
Juerd <''> because every <> matches something dynamic, but this one matches something static
<$foo> because it's ugly and can be $foo if $foo is "$foo", which makes sense 19:12
pmichaud <[...]> isn't dynamic
theorbtwo <> matches something strange. Trying to match nothing at all counts as something strange to me.
Juerd That way, you are left with only one literal interpolator: '' and ""
Instead of <''>, <""> and <$foo>
pmichaud would we do the same for matching arrays and hashes? 19:13
i.e., "@array", "%hash" ?
Juerd pmichaud: It's dynamic in the sense that <[a..z]> matches no dot, but one of a, b, c, ... z
pmichaud: "" would follow normal interpolation for q:2
pmichaud well, that wouldn't be the same as <@array> and <%hash> then, or @array and %hash as they are now
Juerd pmichaud: Making "$foo" the obvious way to match $foo literally in the same way that s/.../{ foo() }/ is s///e's natural replacement 19:14
This all greatly simplifies
pmichaud except for @array and %hash
gaal_ good morning.
Juerd No, those become more clear even
@array now intepolates a disjunction of *rules* 19:15
while $string interpolates a *string*, that is matched literally
That's inconsistent
pmichaud and not true
Juerd Oh?
pmichaud @array interpolates a disjunction of strings
theorbtwo wonders if you can do a C< rule :P5 {...} >
Juerd I see
pmichaud <@array> interpolates a disjunction of rules
Juerd Then at least it is consistent
I still don't like it, though.
pmichaud well, the nice thing about PGE is that everyone can write their own RE syntax :-)
Juerd And I want '' and "" to be metacharacters because <''> sucks 19:16
\( \w+ \) is hardly readable
<'('> \w+ <')'> is worse and hard to write as an extra bonus
'(' \w+ ')' is clean, easy to read and write
PerlJam Juerd: so is <lp> \w+ <rp> :-) 19:17
putter is there a reason Internals.hs imports Data.Ratio twice? i thought that was a no-no.
theorbtwo It is, PerlJam?
Juerd In a world of readable rules (as opposed to linenoise regexes), I would love to be able to avoid using the backslash to escape to literal interpretation
PerlJam theorbtwo: It looks clean to me and I found it easy to write :)
Juerd The same way you don't do $foo = '<' and not $foo = \<
Rules are a language, no longer you should consider them string-like, I'm convinced. 19:18
pmichaud indeed, rules are a language unto themselves. That's why they have their own compiler
PerlJam Juerd: so ... where's your proposal to p6l?
Juerd It's also why \( is needlessly ugly
PerlJam: I must read S05 again before I bug p6l.
ajs Juerd: \( \w+ \) isn't readable? Hmmm... I guess I must be too much of an old-school Unix guy. Looks clean to me. I always loved Perl for cleaning up the \ usage in regexen 19:19
PerlJam Juerd: Then I eagerly await your post. At least then I'll have the time to shoot at it appropriately :)
Juerd PerlJam: And for that I haven't yet had the time. Read logs of this channel to find that I'm pondering this for more than a month now.
ajs: Don't you think '(' \w+ ')' is more readable? 19:20
ajs: And what about [email@hidden.address] compared to either of foo\@example\.com and <'foo@example.com'>
Honestly, <''> breaks my fingers.
theorbtwo Juerd: You're using a sane keyboard layout, right? 19:21
Juerd theorbtwo: Yes, dvorak.
theorbtwo: It's *worse* with QWERTY.
ajs Hmmm... no, because '' is not visually balancing, so '('('('\w+')')')' is REALLY hard to read, even with spaces added appropriately '(' ( '(' \w+ ')' ) ')' where I find \((\(\w+\))\) to be much easier to count up \s
theorbtwo <'...'> isn't /that/ hard to type.
Juerd theorbtwo: Oh, yes, it is /that/ hard.
ajs: I disagree.
theorbtwo Is there a reason not to allow both <'('> and \( ? 19:22
integral agrees with ajs
Juerd ajs: But -- my thinking is not that \ should go away. It should stay.
ajs Juerd: your perogative
Juerd ajs: And '' and "" should be added to the list of metacharacters
PerlJam Juerd: I find <''> fine for longish strings, but I agree that it's slightly annoying for one or two char strings.
Juerd: I take it that something like <(> \w+ <)> wouldn't do it for you either?
ajs I have no gripe with your WTDI. There are cases where it would be clearer 19:23
Juerd PerlJam: No, it's better than the <''> thing, but not by far good enough. Besides that, I'd like to pull weird exceptions out of <>, rather than adding even more.
<> is way, WAAY too overloaded in rules.
PerlJam Juerd: Just like I'd like to have fewer metacharacters rather than more :-)
pmichaud sometimes it's better to put most of the overloading in one place 19:24
PerlJam Juerd: I think the overloading of <> is fine.
Juerd PerlJam: IMO, anything that is \W can be a metacharacter without trouble, *especially* if '' is added
PerlJam Juerd: It allows you to escape into yet another sub-genre of the rule language.
Juerd pmichaud: Yes, certainly. But in my opinion, it is always better to find a way without any overloading at all, if it's within reach and reason. 19:25
pmichaud the nice thing about <$file> is that you get a capture in the process
whereas that wouldn't happen with $file
Juerd pmichaud: Why? Following my proposal, $file would be identical to the current <$file>.
pmichaud you want string interpolations to automatically capture to the hash? 19:26
er, to the match's hash?
Juerd pmichaud: And - I have no real use for <$var>. The symbolic reference thing was thinking out loud.
pmichaud: So they can both be available even
pmichaud: No, I want $foo to be rule interpolation, and string interpolation to be strictly associated with quotes
autrijus so, while you're chatting... 19:27
pmichaud I think that's the wrong huffmanization
autrijus ...I've finished named rules and subrules support.
Juerd pmichaud: I think it's a better one than the current
I won't suggest this on p6l. There's too much opposition.
pmichaud I'm actually not opposed
I'm just pointing out possible advantages of the current system 19:28
PerlJam Juerd: suggest it anyway. Good ideas often spring from the discussion.
Juerd PerlJam: I'm a bit tired of floods of dozens of messages bashing my suggestion, then waiting a few months, and then seeing something that is suspiciously close to what I suggested. 19:29
PerlJam: Unless I think there are people supporting my ideas, I just wait until I see other people come up with whatever I thought.
qmole nice, autrijus
PerlJam of course, you may get another message from Larry to the effect of "I wish people would at least try to see that we've put alot of thought into the current system and that there are reasons why we made the choices we did" ;-) 19:30
Juerd pmichaud: I'm not really concerned with huffmanization for single character things, in rules. I'm much more concerned with aesthetics, typeability (from my perspective as a chronic RSI patient), and logic in communication ("->" looks like an arrow, not minus greater than, and + obviously has something to do with numbers -- "" to programmers clearly indicates something stringy) 19:31
Corion How do I index a (Haskell UTF8) String by character (instead of byte) ? If I have a byte index, I can get at the char index, but how to do the reverse? 19:32
theorbtwo I'm not terribly sure about that last bit, Juerd. 19:33
autrijus Corion: you can get at the char index?
theorbtwo If you want the value of $foo, a string, you write $foo, not "$foo", generally.
autrijus Corion: the only method I had was just encodeUTF8 and decodeUTF8
use them like Encode.pm equiv 19:34
Corion autrijus: I get at the char index by doing encodeUTF8 . takeGeneric . dropGeneric
autrijus: Which is what you did via encode/decode. So I'll try decodeUTF8
autrijus pugs> rule bar { 3 }; rule foo { <bar> }; say "Yes!" if 3 ~~ /<foo>/
Yes!
I imagine interesting things happening with this support :) 19:35
theorbtwo Woo, cool, autrijus!
autrijus like, a perl 6 grammar. ;)
or at least URI.pm can take advantage of this
by using named captures and subrules 19:36
Juerd theorbtwo: If you want a string, in Perl 6, you write ~$foo or "$foo" if you must have string interpretation. $foo is a value that can be anything. If you know it'll be a string, ~$foo and "$foo" are bad style indeed. However, in rules, I think it's more natural to assume $foo will contain a rule, as in "", you assume $foo is as string.
autrijus isn't technology great? :)
theorbtwo I'm not sure about that. 19:37
Perhaps $foo should do the right thing depending on if it does Rule or not?
Juerd theorbtwo: That would be nice too.
theorbtwo: The most important wish I have is that '' and "" be made to work without the surrounding <> 19:38
theorbtwo I'm not sure if that's mishuffmanization or not.
Juerd theorbtwo: "$foo" as string $foo flows from that, and allows <$foo> to become $foo, which isn't entirely important to me.
theorbtwo ' and " are awful common chars to make meta. 19:39
Juerd theorbtwo: Why?
theorbtwo: How often do you match them?
Corion Heh. The Prim.hs refactoring decreased the (unoptimized) rebuild time vastly!
theorbtwo Very often.
Juerd theorbtwo: I personally have not been able to find any example of that in a grammar where escapes for characters aren't used much more often 19:40
theorbtwo: And you can still match " with either \" (same as in a string!!) or '"' (same as you would declare it as a string)
integral great :-)
PerlJam Juerd: if the current <$foo> were changed to $foo, how would you pass arguments to it?
Juerd PerlJam: $foo(), with in (), expression syntax 19:42
PerlJam So would $foo.() work too?
Juerd PerlJam: This is a good argument for not doing this, by the way.
PerlJam: Still, I want the quotes.
PerlJam Juerd: Well, adding " and ' to the list of metacharacters doesn't seem too far fetched at the moment. 19:44
Juerd So currently, you've succesfully changed my mind about both <$foo> becoming $foo and <$foo> being symbolically referring.
PerlJam (though I'd prefer fewer metacharacters_
autrijus Corion: yes. refactor early refactor often!
PerlJam s/_/)/
autrijus (AST.hs particularly needs love, but I've said that)
Juerd I like the suggestion of bare $foo being <$foo> if $foo.does(Rule)
integral is loving AST
Limbic_Region question - does my $rv := some_routine(); bind $rv to the underlying variable being returned in the sub if it is like sub some_routine () { my $foo = 42; return $foo; }
or just a copy?
PerlJam Limbic_Region: yes.
Juerd PerlJam: It's not far fetched at all. It's really close. See also the many forms of BNF.
Limbic_Region PerlJam - so if that value changed $rv should change too? 19:45
cause if it isn't supposed to be a copy - I think Pugs is broke 19:46
PerlJam Limbic_Region: pugs mayn't be broken. I may be wrong.
Limbic_Region oh - lots of help you are
:P
PerlJam Limbic_Region: I don't think anyone has asked that question before (but if they have, it's on p6l somewhere) 19:47
Limbic_Region is writing p6.l now 19:48
Corion autrijus: I don't understand the doMatch and why it has UTF8 problems. What charset is Pugs internally?
Or bytes?
... or rather, what format are the Strings stored in?
knewt hmm. methinks it might take a while to sync down the pugs tree 19:49
theorbtwo We probably shouldn't rely on them being anything specific... but they are probably utf8. 19:50
Juerd tries to refrain from raising the usual perl 5 red flag: you mustn't care about the internal representation, assume it's vogon for all you care.
integral knewt: use --skipto
knewt integral: it's nice to have the history
integral *all* of it 8-)?
PerlJam integral: --skipto --mylou --mydarling? 19:51
knewt well ok, i could probably have done without the majority :)
but now it's started...
Corion theorbtwo: Hmmm. We need to rely on them being something, or rather, there is some cognitive dissonance between what PCRE returns indices to, and what is returned to Pugs / Run.hs, and how they don't fit together.
knewt ghc6 is what i need once i've grabbed it, right?
integral killing it is fine. You can tell I did this earlier ;-)
knewt: 6.4
knewt damn. only got 6.2.2 in hoary 19:52
it won't work with 6.2.2 ?
Juerd No
integral nope
autrijus Corion: Strings are stored in Char
Corion: Char are unicode.
Juerd knewt: Get ghc-cvs from universal/sid
autrijus Corion: but to pass things into PCRE, we need to encodeUTF8 them.
Corion: so think of it as converting a u-string into bytestring
Corion: now when PCRE passes back offset
they are in byte offset 19:53
even though PCRE does unicode matching just fine
the only thing we need to do
knewt Juerd: what, the source, and then build my own local package?
autrijus is to convert the array of offsets from byte to char unit
everything else will Just Work
end of explanation :)
Corion autrijus: Hmmm ... We also return a substring. I don't know what it is used for, as we also return the offsets.
But that must be converted as well I think 19:54
autrijus Corion: it's used to avoid keeping a copy of orig string around
Corion Aaah. OK.
autrijus Corion: which may or may not make sense :)
but yeah, if the offsets are correct
then just use offsets on the orig string (in u-char)
Corion autrijus: It will prevent later ugly hunting of weird interaction.
autrijus will yield the correct substring
yes, that too.
Juerd knewt: No, it's the cvs version, compiled 19:55
knewt: apt-get install ghc-cvs, after you've enabled universal
knewt Juerd: i'm running ubuntu, not debian 19:56
Corion Yay! Seems after talking through it again, I changed something that makes length.t Just Work! Thanks autrijus !
knewt prefer not to mix/match
Corion hurries to commit, just after the smoke test :) 19:57
Limbic_Region Corion - but do you know WHAT that something is and WHY it just works?
Alias_ what exactly is $bot<add_handler>("runloop", &svn_check)
is that a method call?
Juerd knewt: universal is an ubuntu repository
Corion I even realize what I had wrong. I thought that Haskell used UTF8, but it doesn't. It needs a gateway via encodeUTF8 on the way to PCRE and decodeUTF8 on the way from PCRE 19:58
Juerd knewt: in debian, it's called sid.
Limbic_Region Alias_ sort of
Juerd knewt: It's the same thing.
Limbic_Region since we don't have real methods yet
Alias_ ah
Limbic_Region using dispatch table
Alias_ So I'll almost never see it in real P6 code?
Juerd knewt: Do you want to run pugs, or not? You can (must) trust apt to handle things correctly and reversible. Besides, multiple ghc versions can be installed alongside.
Limbic_Region Alias_ - you will most assuredly see it in real P6 code, but just not for that purpose 19:59
Alias_ good enough for me
knewt Juerd: universe is an ubuntu repo, and i've got "hoary universe" enabled. but the ghc-cvs in that is dated 2004-07-25, and 6.4 wasn't released until march 2005
Juerd Limbic_Region: Sorry I missed your question when you asked it here. I had an answer :)
Limbic_Region: Mailed now.
theorbtwo $bot{'add_handler'} has a coderef.
Limbic_Region thanks Juerd
Alias_ so it's basically ->() ? 20:00
Limbic_Region likes instant gratification
Alias_ oh wait..
theorbtwo Yup.
Limbic_Region in p5 speak yes
Alias_ ->{foo}->()
Corion Ah great. And now when I want to test my stuff, the build is broken :/
theorbtwo The second -> is optional in p5 too.
Limbic_Region well, p5 only forces you to dereference at 1 level
Juerd knewt: That sucks. Then pick the package from sid. Still won't hurt your system, AFAIK
Limbic_Region but yeah
integral urk, (broken builds)--
Alias_ theorbtwo: But only by people that prefer terseness over clarity :)
Juerd knewt: If you really don't want to "infect" your system with perfectly compatible debian packages, install debian in a chroot with debootstrap
Khisanth Alias_: all the other -> would make for a lot of clutter 20:01
autrijus for how things will eventually look like, see svn.openfoundry.org/pugs/ext/Test-B...Builder.pm :)
Alias_ Khisanth: Easy to read clutter
autrijus build's broken? 20:02
Limbic_Region Alias_ - in any case, dispatch tables are going to be as useful in p6 as they were in p5 for the right situations - for OO though it is best to use OO
Juerd knewt: Which is a matter of mkdir ~/debian; debootstrap ~/debian sarge ftp.debian.nl/debian; cd ~debian/etc/apt; rm sources.list; wget juerd.nl/sources.list; chroot ~debian; apt-get install ghc-cvs svk
Alias_ Limbic_Region: Of course
Khisanth ->{}->[]->{}->() doesn't really add anything helpful
Alias_ Limbic_Region: But there's normal code, and then there's fun stuff
Khisanth and .{}.[].{}.() even less!
Limbic_Region Alias_ - we don't have OO yet so we get to say "We did OO before it existed" 20:03
Alias_ I heard :)
Limbic_Region ported Config::Tiny and thought it was a lot of fun
discovered 3 pugsbugs in the process
Alias_ Limbic_Region: It should have been. Config::Tiny is pretty... dense :) 20:04
autrijus Corion: unbroken
Limbic_Region I really need to study the Perl6::Bible as most of my p6 code is still just a straightforward translation of p5 20:06
Corion Yay. Build is progressing nicely and should soonish start smoking ;) 20:07
Alias_ Limbic_Region: Especially to find a way to implement a blessed HASH properly :)
Because I noticed you had to call a method to get the hash, and then get the value :)
Which is somewhat against the spirit of the original 20:08
Limbic_Region Alias_ - that's a bug I found
knewt Juerd: ok, see, unless i force the dependencies it won't work to install a sid package into hoary. [[ Depends: libc6 (>= 2.3.2.ds1-21) but 2.3.2.ds1-20ubuntu13 is to be installed ]]
Limbic_Region you don't really need the blessed hash to do the spirit of Config::Tiny 20:09
Juerd knewt: I personally have never had problems replacing ubuntu pakages with newer debian ones
Limbic_Region you just need to be able to store a hash reference as the value to a key in another hash (which will act as an accessor/mutator method)
knewt Juerd: except that now sid has a new version number on its libc6 than hoary does
Limbic_Region unfortunately - sticking a hash reference to another hash doesn't work correctly 20:10
knewt Juerd: and when a package depends on that version number it won't install without forcing (no thanks)
Limbic_Region subsequently I found the bug is only a bug if you are initializing/creating the hash
if you create it and then add the reference it works as expected
as soon as I get a machine I can commit with I will make it prettier
Alias_ - like I said, a lot of fun 20:11
Alias_ heh
Juerd knewt: I have replaced ubuntu's libc with debian's.
Alias_ I was curious how it would be done though
Corion Limbic_Region: You should work on my JavaScript text editor. Maybe you could make a telnet emulator out of it as well :)
Alias_ It was really just a curious confluence of Perl 5 oddities that allowed it to work in the first place
Juerd knewt: But I use bare debian now, because ubuntu was as limiting as many other distros
Alias_ I couldn't resist having a "plain" hash that you could call methods on
Juerd knewt: There's still the debootstrap+chroot method if you don't want to take risks 20:12
Alias_ especially for something like .ini files which are SO obviously just a HoH
autrijus journal up; g'nite *wave* 20:13
Juerd Sleep well
integral be prepared for a new AST.hs in the morning :-)
Limbic_Region Alias_ - it will work in the dispatch table with just 1 extra level
ajs Ok, I give up. What am I doing wrong: ./pugs -BParrot -e 'sub factorial($n){if $n>1 {$n*factorial($n-1)} else {$n}} say factorial(3)'
Corion BTW, there is one unguarded test (failure) in undef.t : pugs: user error (Cannot find the parrot executable in PATH) ... 20:14
Limbic_Region In fact Alias_ - I may be able to nopaste something for you - give me a minute or two
Alias_ sure
autrijus integral++ # I'll be prepared :) 20:15
ajs: I think, obviously, the bug is in the -BParrot 20:16
take it away and it will work ;)
(but seriously, Parrot.hs codegen needs to be rewritten using MonadEval to keep track of temp PMCs.)
ajs Ok, I thought I was doing something dumb. Let me poke around and see what Parrot is thinking
Limbic_Region Alias_ - after working around the bug I changed the code enough to make changing it back non-trivial (more than 10 minutes worth of work) 20:17
I should have net access again after Saturday
autrijus Corion: "Cannot find parrot executable" should perhaps to be tryIO'ed in Prim's evalPGE.
Limbic_Region will ping you then
Alias_ ok
autrijus Corion: i.e. catch it and rethrow using Eval monad's fail""
but I need to sleep. :) &
Corion autrijus: Ah. Maybe tomorrow. I need to sleep in one hour ;) 20:18
good night autrijus
autrijus nite
mj41 g'night 20:19
Juerd autrijus: I don't know how to write a test for it, but can we have __END__? 20:21
Juerd uses pod for now
Limbic_Region Juerd - is __END__ a parse error? 20:22
now I mean
Juerd AFAICT 20:23
push @foo, [ 1, 2 ]
Limbic_Region ok - so I think I figured out how to test it
Juerd Add 1 or 2 elements to @foo?
I'm really hoping for 2
eh
1
It's adding 1, 2 now
Instead of an arrayref
Is current behaviour a bug?
gaal Juerd, is __END__ specced? 20:24
isn't it spelled =for END or something
?
ninereasons =begin END
Juerd gaal: No, it's not specced.
gaal this is the end, my only friend, etc.
Juerd Okay - why do arrayrefs flatten in list context?
That way I can't nest them :(
Limbic_Region Juerd - WRT to testing for __END__ - create a module that has an __END__ and modify an exported variable after the __END__
in the test script that uses the module, check the value of the exported variable 20:25
Juerd Limbic_Region: Ah, good idea
Limbic_Region right now it would be a parsefail
gaal Juerd, do you just need something that ignores everythign after the __END__? i think i can add that
Limbic_Region after implementation it should be the last value before __END__
Juerd But I do wonder if __END__ is valid. I hadn't considered the possibility it'd go away.
gaal: I have =end now, which works too :)
Limbic_Region probably should ask @larry 20:26
Juerd gaal: I converted some script to p6 long time ago, I'm testing it for the first time now
I need to do it line by line
Limbic_Region cause people tend to like inline files
Juerd Because pugs isn't quite ready enough for this :)
Limbic_Region __DATA__
ninereasons @=DATA
gaal i think data has also gone away
yeah
Juerd pugs> [scalar [1,2, scalar [3, 4]]] 20:27
(1, 2, (3, 4))
Limbic_Region but has __END__ ?
gaal now accessible consistently and programmatically like ninereasons says.
Juerd This is HARD to work with.
gaal i think so, yes.
ninereasons SO2 Limbic_Region
Juerd AAAAAAAAARGH. I'm going nuts here :) 20:29
Limbic_Region ninereasons - [16:06] <Limbic_Region> I really need to study the Perl6::Bible as most of my p6 code is still just a straightforward translation of p5
Juerd I want very simple nested literal arrays :(
(flattening of references is madness. I wonder why it works like this)
ninereasons i'm pretty sure it's a persistent bug, Juerd
squashed and resurrected, repeated.y
_metaperl isnt Perl6 a community rewrite of Perl? can the community vote on your issue Juerd? so far, it looks like it is no longer a community language 20:30
Juerd _metaperl: It's not the community's language.
_metaperl: The community writes the code.
hachi "@numbers >>++;" is throwing an error in pugs, are the hyper-operators working?
_metaperl perl6 is a community rewrite of Perl
integral the community's input was some RFCs ages ago
Juerd Larry makes the final decisions.
Corion I don't believe in "voting". That assumes that all voters are equally competent.
ninereasons I suppose that someone might say that, "final" decisions are made in usage 20:31
Corion Of course, we are free to implement "useful" things in Pugs and see if they survive until Perl6 20:32
gaal forum2.org/gaal/pugs/smoke-cygwin.html incidentally. 20:33
Juerd Corion: My wish is a wish for rules. 20:34
Corion: Which has some impact on compatibility. 20:35
Corion gaal: I unified your patch with my reimplementation of getEnv and it all Just Works for the moment ;)
gaal yes, thanks - works great here!
Juerd Is there any haskeller around who could make arrayrefs in list context non-flattening?
Currently, [] and () are treated equally. 20:36
gaal my wish is for fluency in haskell like my fluency in perl
hachi ow!
Corion Juerd: I think I could find the place, but I only got one more hour today ...
Juerd Corion: Do what suits you best
Corion ... and I want to hack up some nice graph display thing in HTML, like GraphViz but more restricted. For displaying a MUD map. 20:37
So, if anybody has a nice MUD map renderer that I can access / feed from Perl, ... :)
gaal (not to say my perl is perfect of course, only that i don't get stuck with it so often as i do with relatively basic things)
Juerd Hm
Corion (actually it's a dual-lived project - MUD map renderer to edit a MUD, and also to display the dataflow/system structure for $work)
Juerd Corion: If you write it in perl 6, you'd need arrayrefs probably, so do that ;) 20:38
Corion ... but the data structure is quite similar.
Juerd: Hah. We'll have to wait until my boss adopts Perl6 as new development language :)
Juerd You write muds for work? Nice job!
castaway define nice, Corion ;) 20:39
Juerd 17466 juerd 25 0 180m 177m 10m R 96.3 35.1 3:38.42 ghc-6.5 20:40
holy puppy
It's very memory hungry.
hachi err, I thought it was eating up over 500 MB last night when I compiled
Corion castaway: Ah, at least like GraphViz, if not better. I'd prefer something like SVG or DHTML with drag-and-drop or a very good automatic layout of course.
castaway aehm..
well theres a pic of mine somewheres..
Corion Juerd: No, I don't write MUDs for work, but I realize that the (simplified) data structure of a MUD is the same as the data structure of our IT architecture and its dataflow :) 20:41
(and $boss ran a MUD in his university days)
crysflame are you neu too?
castaway doesnt like much of GraphViz' output 20:42
Juerd Corion: wheh
castaway Corion, desert-island.dynodns.net/perl/telnetclient.html
theorbtwo Your IT arch is like a dungeon? That's scary.
Corion r2948 - 4860 test cases: 4783 ok, 77 failed 20:43
(at datenzoo.de/pugs/win2k.html )
castaway hmm.. I wonder.. you could probably make pretty ones with zMapper actually :)
Corion theorbtwo: Well, the systems are like MUD rooms, the import/export interfaces are like doors and the data transport is a connection between doors.
castaway and that one uses a "real" DB
(and is customizable..) 20:44
that'd be interesting, to use it for a network layout..
Corion castaway: That's my plan, at least for $work ;) 20:45
anyway - afk for the moment
theorbtwo I'm convinced that it's possible to get useful graphviz output... I just haven't figured out how yet.
castaway which is the plan? 20:46
yeah, I'm still hoping you will :)
Corion theorbtwo: I don't know how to tell GraphViz that a certain passage has to connect to the "north" of a cell
Juerd theorbtwo: Ignorance isn't bliss in this case. It's a huge waste of time.
Corion still - afk &
Juerd PackFile_unpack: Bytecode not valid for this interpreter: version mismatch 20:47
Parrot VM: Can't unpack packfile /home/juerd/project/pugs/src/pge/PGE-Hs.pbc.
What does it mean and how can I avoid it?
theorbtwo bets the file is zero-length? 20:48
Juerd I have no idea which file and why. 20:52
stevan gaal: ping
Juerd Okay, I give up.
I'm going to hard code
Arrayrefs are utterly useless if they flatten
gaal plong! 20:53
stevan gaal: did you see my comments re: Hack::Instances
gaal nope! shall i grep the logs?
stevan nah I can tell you
gaal yes plz
revdiablo Juerd: fwiw, I wrote up some tests to show the weird flattening behavior in t/pugsbugs/flattening.t 20:54
stevan well, originally I had a counter
Juerd revdiablo: Thanks
stevan but I changed it to the rand number later
to get a more identifiable string
gaal a random number is more identifiable? :) 20:55
stevan in a weird way yes
but it would really function the same either way
gaal hahaha!!! lookie this:
./pugs.exe -V | egrep path_sep\|osname
osname: cygwin
path_sep: :
ajs grats!
ninereasons wow. many of the tests work now, in t/rules/Disabled/rules.t
gaal i think i know why pugscc &c didn't work on cygwin ;-) 20:56
ninereasons pmichaud++
ajs gaal: oh heh, I thought you were pointing out having gotten it working,
yeah, : isn't quite right is it :)
gaal stevan, a rand number is just a wee bit less safe, isnt' it?
stevan how so?
Juerd Anyone any idea how to use a literal hash ref?
gaal better have both imho
Juerd both { } and hash { } are broken
stevan I ran about 10,000 random numbers and got no dups 20:57
gaal: we can do both I suppoer
gaal ajs, this could have been the bug that was preventing it from work
stevan suppose
Juerd hard codes again
stevan gaal: it is also a very temporary hack (at least I hope it is)
Juerd: welcome to the wonderful world of Pugs hacking :)
gaal stevan: well, a reasonable person would say that you and me, and all the pugs hackers together might never get a collision until real objects land 20:58
but i am unreasonable! :)
stevan LOL
putter Fyi, rules are apparently not usable on FC3 x86_64. Embedded due to a ghc bug which may or may not have been fixed yesterday in cvs-ghc, and parrot standalone due to an undiagnosed interrupted system call failure.
stevan gaal: ok, adding the counter in now ...
gaal just make sure you don't *add* the counter :))
stevan LOL
gaal: concat-ing the counter :) 20:59
gaal this reminds me of a riddle (which of course will now be easy to solve)
join it witha delim, stevan :)
stevan yes
OBJECT;<class>;<rand>;<counter>
gaal n programmers are siting at a table and they want to know their average salaray
putter Fyi, as of ghc-6.4.20050506, on FC3 x86_64 the pugs-kill gc bug has _not_ yet been fixed (hopes to the contrary not withstanding;). A ghc bug report has been submitted. 21:00
gaal however, they don't want evryone to know each other's particular salary.
putter err, s/pugs-kill/pugs-killing/
gaal what scheme can they use to know the average w/o losing privacy?
n > 2
theorbtwo wonders if it's a card table. 21:01
stevan gaal: r2949
gaal stevan++
stevan gaal++
gaal stevan+=rand+$counter++ 21:02
stevan gaal: the answer is 0, they were all outsourced
theorbtwo Everybody gets some chits, closes their eyes, and throws the chips representing their salary into one hat, and the rest into the other.
stevan :D
theorbtwo This may involve more aim then programmers are capable of, though. 21:03
stevan more athleticism at least
wow, bsmith++
gaal svnbot6 is teh bomb 21:04
putter Re Perl6::Rules t/*.t port, I hope to clean it up tomorrow to reflect Damian's S05 draft post, and check it in largely untested. :(
stevan I love svnbot :)
integral Pugs.AST.Internals is still >1500 lines :-)
Corion bsmith is on a refactoring rampage :)
stevan jabbot: do you love svnbot?
jabbot stevan: We were discussing you -- not me.
Corion but that's cool, because that reduces (re)build times while developing
stevan jabbot: are you jealous of svnbot
jabbot stevan: What are your feelings now?
stevan jabbot: I am feeling that your days are numbered 21:05
jabbot stevan: Really, my days are numbered?
castaway ;)
theorbtwo wonders if anybody has a better plan.
stevan jabbot: yes, it is only a matter of time before ingo write you out of existence
jabbot stevan: åŽŸä¾†å¦‚ę­¤
putter wishes svnbot "notices" were logged...
Juerd Fsck. No gather/take yet? 21:06
stevan Juerd: nope
integral putter: svn log ...
Juerd grumble
ajs a tells b "s(a)+rand(1)", b tells c "s(b)+s(a)+rand(1)+rand(2)" c tells a "s(c)+s(b)+s(a)+rand(1)+rand(2)+rand(3)" a subtracts rand(1) and passes result to b, b subtracts rand(2) and passes to c, c announces "(s(a)+s(b)+s(c))/3"
Juerd Is "is copy" implemented? 21:07
Corion I doubt it... 21:08
stevan yes
although maybe not 100% for really big/nested/odd structures
Juerd I'm getting complaints about trying to change a readonly value
But I can't find it
gaal ajs++
except that rand(1) is not really needed
it can be any random number
oh
putter integral: colabti.de/irclogger/irclogger_log/...-05-10,Tue : "bsmith is on a refactoring rampage :)"... but you wouldn't know it, or have much idea what's being discussed, by looking at the irc log. :(
Juerd tnx.nl/3763VEZF
stevan Juerd: you can always make a new local copy
integral putter: oh, right 21:09
Juerd *** Error: Can't modify constant item: VRef <Hash>: NonTerm (MkPos "examples/matrix.p6" 38 18 38 31) at examples/matrix.p6 line 51, column 2-18
gaal you were using rand(1) as "a first randome number". yes, that's right.
Juerd That's the build_re line
The bottom line in the paste
gaal bbiab
integral and it's less a rampage, and more a let's wait for ghc to finish, _again_
ajs gaal: yep... sorry, subscripts not available ;-)
gaal integral: make unoptimized
is much much faster to compile.
integral yeah, still slow
stevan Juerd: which one is not able to copy? %had? 21:10
Juerd stevan: It still complains if I copy manually
stevan: I have no idea which one!
stevan my %new_had = %had;
as the first line of the sub
Juerd stevan: Still complains
sub build_re ($y, $x, $todoX, ?%hadX) {
my $todo = $todoX;
my %had = %hadX;
stevan Juerd: where is @matrix coming from 21:11
Juerd stevan: Defined earlier. It is in scope
As is @adj
stevan Juerd: then your problem is likely in @matrix
Juerd How so?
stevan I have had issues with this before
Juerd @matrix isn't changed
stevan oh 21:12
ok
lemme look closer
Juerd The function's pure with dependencies
stevan Juerd: what is line 38 21:15
Juerd stevan: my @next; 21:16
Oh, no, line numbers changed
I'll repaste current first
stevan ok
gaal beh, that bug i though i found, it wasn't a bug :( 21:17
whcih sucks because if it was, we'd probably have had a working pugscc on cygwin
Juerd tnx.nl/3764JXCT
I hate Terminal.app
It scrolls on selecting text
stevan Juerd: so which is line 40 21:18
gaal p5 $Config{'path_sep'} isn't the, uh, subdirectory separator, it's the $PATH separator :)
Juerd stevan: %had{...}++ and next
stevan Juerd: have you tried breaking that up a bit 21:19
%has{ $_<y> ~ "/" ~ $_<x> }++
Juerd I'll try that
Tried; doesn't work
Same error
It's in the push line 21:21
Now to find out why...
Hm, no, it doesn't have to be
stevan $y and $x need to be is copy maybe?
Juerd They're not changed
stevan ok
Juerd The problem somehow has to do with the inner build_re call 21:22
Hm, no, even if I put its arguments in a bare list it croaks
stevan how about try adding a topic var to the loop
no using $_ 21:23
s/no/not/
Corion slumps into bed.
Juerd Any use of %$_ is wrong I think 21:24
So let's try that indeed
gaal should i be getting this? pugs.exe: user error (Cannot find the parrot executable in PATH)
Corion r2950 - datenzoo.de/pugs/win2k.html
gaal: See the backlog. Autrijus hinted at what should be done about it.
(tryIO)
g'night
&
gaal night Corion
Juerd Argh, $_ is undef 21:25
mugwump morning all 21:26
Juerd Oh, fscking hell 21:27
Hashes are flattened too
So essentially, list context is fucked up entirely
I give up for now
Thanks for the help, stevan
stevan Juerd: your welcome 21:28
mugwump: hey there
mugwump: I have a few questions when you have a moment 21:30
gaal bug: s:p5/aaa/bbb/ goes to parrot, s:P5 does not. 21:41
s:perl5 works too. maybe i got it wrong and :p5 is invalid? 21:42
mugwump hey stevan. I notice that there are two dists in ext/ now ... 21:48
stevan mugwump: I moved it all to MetaModel 21:49
you probably need to remove metaClass manuaully
knewt woot, stage1 has finally built, and it's moved on to stage2 21:51
revdiablo gaal: I remember reading somewhere that you could use :P5 :perl5 and :Perl5 but not :p5
gaal hey, anyone looking for a Prim to add?
i'd like rename and chmod :) 21:52
octal is spelled like 0o755 now, right? 21:53
....and how do you say qx? is there a final word? 21:54
mugwump ah, right. I wonder why svn didn't do that for me 21:55
stevan mugwump: maybe you had local mods?
svn can be picky sometimes
mugwump that must have been it 21:56
it saw all my emacs ~ backups and just couldn't bring itself to delete them
stevan :)
mugwump: so, I am wondering about MetaAssoc 21:57
we are not using it to assoc properties to classes are we?
properties == MetaProperty 21:58
mugwump Not directly. The association between MetaClass and MetaProperty is simply a hash
stevan ok
I noticed (somewhere, not recalling right now) that methods can only ever be associated with one class 21:59
and same with properties
so should those two have a 'class' slot added?
so we can enforce that?
mugwump Sure
stevan ok
we surely can make Visibility and AssociateClassWith Roles when we have them 22:00
but for now I will cut and paste :)
jaap I'm trying to download pugs from the darcs repository, but the link on the pugs website doesn't work 22:01
stevan also I think the SuperClass -> Class -> SubClass relationship is pretty solid now
jaap i'm trying darcs get wagner.elixus.org/~autrijus/darcs/pugs
stevan I die if we have circular inheritence
jaap: I think wagner.elixus is down
jaap stevan: is that temporary, or should i use a different repo?
stevan jaap: I would recommend the SVN repo
I think i heard autrijus say it will be down indefinetly 22:02
jaap okay. such a pity, i much prefer darcs :-(, ah well, i'll get it from svn then, thanks
stevan mugwump: so, when a property is added to a class, I set the properties 'class' slot with $inv 22:07
s/class/metaclass/
and can properties be removed from metaclasses? 22:08
I assume they can
mugwump naturally, this sort of thing should be able to be abstracted 22:09
but I'd worry more about making that work in classes that will be generated by MetaModel objects 22:12
stevan mugwump: I am not sure what you mean
mugwump ok, well these assertions that we're making about the relationship between "superclass" and "subclasses" - at least, the generic one that membership in one implies a corresponding membership in the other - apply to all associations 22:14
stevan ok 22:15
but in meta-meta-models properties will more than likely be fixed?
mugwump I see this as an application of Set Theory style relationships to in-memory objects. But that's my connection, I don't have a thesis to back that up :)
It's unlikely that the set of objects at Perl::MetaModel::$Model (or whatever) will change at runtime 22:16
stevan I know nothing (much) about set theory, so I'll buy that
mugwump: so I can assume for now, that its set-once and then read-only
and maybe in later refactoring, ... we can add the ability 22:17
mugwump If you know how to use a foreign key or a link table in a database, you at least know the concept driving this relationship companionship...
changing at runtime basically means that a person is loading a module to parse another language or mutate the language
stevan mugwump: and they are on crack
so I wont worry about that then
mugwump :)
stevan I will only enforce the rule for inserting 22:18
I wont bother with taking it out
knewt feck. i just ran out of disk space. this means that the entire build process from bootstrap -> stage1 -> stage2 will run again. aarrrggghhh.
mugwump Yep. It's an interesting condition to try to assert. Basically you have to catch updates to the containers, do set operations to find out the "differences", then send messages to the affected objects
stevan mugwump: for now I was just going to make a removeClsProperty method and make everyone go through that 22:19
mugwump right. Now you're arriving at another conclusion I came to in Class::Tangram, this is very good :) 22:20
stevan :)
meta-meta-hacking is fun 22:21
I also found a book on MetaCLasses on my bookshelf
I dont have it with me at the moment
mugwump cool... any good?
stevan havent fliped through it yet
got it a while ago in a tech-book grabbag on ebay 22:22
never had a use for it until now
ok next question 22:23
what are methodParams?
the method args?
or something else 22:24
more mysterious maybe :P
mugwump it's a signature, really. there are lots of parts on the MetaModel that are pre-echos of what will only really be needed on the next level
stevan ok 22:25
so should I break that up? returnType, argTypeMap, etc?
ooh, the Metaclass book has symbol glossary, 22:26
which means no more guessing at what all those funny bits mean
mugwump we want all that magic to happen on the Class Model, it happening on MetaModel is ok but perhaps superfluous for "bootstrapping" 22:28
stevan ok
so I leave params as just a list? or hash? 22:29
hash seems to make the most sense
mugwump if it makes sense to you that a hash represents a parameter specification :). 22:33
stevan mugwump: well it all depends on the details of the parameter spec, a hash seems pretty flexible 22:36
mugwump It needs to be something like the Params definition in src/Pugs/AST.hs 22:38
sorry, src/Pugs/AST/Internals.hs
but again - doing that in full might end up being unnecessary in the Meta-Model.. 22:39
stevan line 561 -> data Param = MkParam 22:41
that one?
mugwump yes, that's it
the metamodel objects we're building would ideally be generated from these parts of the haskell source 22:43
perhaps, anyway.
stevan ok
well I think a list is better
as we will likely have a list of param objects
mugwump yes. besides, the MetaModel is supposed to be as simple as possible 22:44
stevan ok

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!