Pugs t-shirts www.cafepress.com/pugscode | Pugs 6.2.9 released | pugscode.org | pugs.kwiki.org | paste: sial.org/pbot/perl6 | www.geeksunite.net
Set by stevan on 15 August 2005.
stevan putter: pong 00:20
buu stevan: Explain that -- 00:21
stevan buu: explain what?
buu stevan: \-1.0 ?
stevan buu: I have no idea
buu =[ 00:22
stevan ?eval my $a = 0; $a--;
evalbot_7085 0
stevan ?eval my $a = 2; $a--;
evalbot_7085 2
stevan ?eval my $a = 2; --$a;
evalbot_7085 1
stevan buu: the difference in your results can probably be explained by evaluation order 00:23
as for the why you get \-1.0 I have no idea
buu stevan: Right, the first one was a mistake.
stevan buu: I am not sure what the state of type coercien is in Pugs 00:24
?eval my $a = 2; --$a; $a
evalbot_7085 \1
stevan ?eval my $a = "a"; --$a; $a
evalbot_7085 \-1.0
stevan ?eval my $a = "a"; --$a;
evalbot_7085 -1.0
stevan ?eval my $a = "z"; --$a;
evalbot_7085 -1.0
stevan hard to say, I am not familiar with the internals much these days 00:25
putter stevan: pong? 00:33
stevan hey putter
putter hi stevan
stevan I am looking over Perl6::(Value&Contanier) as well as PIL-Run 00:34
putter how much of mm2 works now, compared with mm1?
stevan well, most of it I think
there are some changes
putter questions welcome. the code has gotten a bit crufty.
stevan looks like most of those will effect Perl6::(Value & Container) more though
putter: to be honest, I dont know enough about PIL and all that to really grok what is going on in PIL-Run 00:35
putter I've been trying to keep the P6::V/C interface concentrated in ApiX, but it's leaked out a bit.
well, let's see... 00:36
lib6/Test.pm is a simple version of Test, tuned to what is/isnt working in pilrun at any moment. the regular Test.pm is just to complex. 00:37
lib6/P5Runtime/PrimP6.pm is just some primitives, implemented in regular p6. the equivalent of pugs's src/perl6/Prelude.pm
svnbot6 r7086 | putter++ | PIL/Run/ApiX.pm: implemented parameter default value. Param->new() is now called with invocant/lvalue/lazy/named/writable flags, though Code.pm does not yet implement them.
putter crude_repl.pl is the "main" program. but is mostly just a read-eval-print-loop, and some command line argument handling. 00:38
lib/PIL/Run/MainX.pm require/use's everything 00:39
stevan EvalX?
it looks like it is doing the PIL -> PIL-Run conversions
putter lib/PIL/Run/Type/Macro.pm and Object.pm are leftovers from when P6::V/C weren't used. They still dont have a Macro, but one could/should(?) create a Macro is Sub and finish throwing away lib/PIL/Run/Type/. 00:41
lib/PIL/Run/PrimP5.pm is some primitives writen in mutant p6/p5. the sub arguments are p6, the bodies are p5. 00:42
it's the equivalent of Prim.hs
lib/PIL/Run/ApiX.pm provides a bunch of p6_mumble functions, intended to (mostly?) wrap the runtime and mm. to keep design decisions localized, simplify swapping them, etc. 00:43
lib/PIL/Run/Root/ is nothing. well, its currently the root of the P6 namespace in p5 space. but no files go there. 00:44
stevan yup I saw that one 00:45
ok, well I think mm2.0 might be ready for experimentation
I am not sure it has all the features from 1.0 yet 00:46
and some things have changed
I was going to try and convert chromatic's Test::Builde to MM2.0 and see what happens
putter lib/PIL/Run/EvalX.pm has the PIL node classes, which just provide ->expand() which yields p5 code. and a couple of functions at the bottom to take p6, call pugs, get pil, eval it, expand() it, and eval the p5. There is some initialization cruft at the very bottom, which should probably go elsewhere. 00:47
and just to wrap up, lib/PIL/Run/Match.pm is the Match objects returned by regexp matching. It probably belongs in P6::Value. 00:48
ok... 00:49
stevan so what is class support like in PIL
IIRC is wasnt that great
putter no... 00:50
the key thing is you find out about classes incrementally. an accessor method shows up. another method. at some point you see a...
stevan ok, well the MM can support that 00:51
just dont use the class 'Foo' => { } stuff 00:52
putter &Class::_create('C') or Role::_create or whatever. A hack to tell the runtime its a Role/Class/Package/etc.
stevan and build the classes manually
One major difference is that I dont use p5 symbol tables for classes
putter though one could grovel over the pil looking for Mumble::_create('Foo') flags if one needed them ahead of time.
stevan I am not sure how much of an issue that might be
putter hmm... currently name lookup works like 00:54
stevan it doesnt
classes are just scalars
name lookup can be added
putter (defined($scalar_xxx)?$scalar_xxx:p6_lookup('scalar_xxx',1)||do{warn "autovivifying scalar_xxx\n" if $main::global_debug;$scalar_xxx = Scalar->new()}) 00:55
Ie, try a lexical $scalar_xxx or $array_xxx or etc, and if there isnt one, call a package-local-sub p6_lookup(), which then crawls up the package tree looking for it. 00:56
and in this case, autovivifies it if it isnt found
stevan ok 00:57
putter so if each class has some package associated with it (which the p6_mangle()r can rename vars to), p6_lookup could hook into the mm. either with the mm providing lookup, or providing a list of package names to check. 00:58
stevan mm2.0 implements it's own ::Package 00:59
putter ie, in A::B::C, $x is searched first in C, then B, then A, then *.
so if the mm says ['A::B::C','A::B','A','*'], or perhaps their mangled forms, then p6_lookup could take it from there. 01:00
stevan I will need to think about this 01:01
the MM2.0 really tries not to use any p5 packages at all
all package use is in chaos.pl
putter though given $X::y , $y has to be searched for in A::B::C::X, A::B::X , "A::X", "X" the last time I checked (a while ago) 01:02
stevan it is mostly to either 1) tag types so I can check them with blessed() and 2) provide a place for the dispatcher to exist
putter that's fine,
stevan putter: I am not sure I understand the way you are nesting packages
are those scopes? or package namespaces? 01:03
stevan thinks maybe he should read more of PIL-Run
putter pilrun could provide the mm-to-pkg link. like it does now.
re nesting,
my fuzzy recollection of namespaces, perhaps now obsolete as there have been changes, is given package A{package B{package C{say $X::y;}}}, 01:04
one looks for $y in... A::B::C::X, A::B::X , A::X, X. 01:05
stevan oh yes,.. I think you are right on that 01:07
although the Apoc are supposed to be very very out of data
date
and they are very sparse on package/module information
it is not a "fun" topic I think
putter it would be nice if the mm provided a "list of names of classes in currently lexical stack", or perhaps even better, a "given a variable name, a list of classes to check for it in". that way, when p6l changes things, only the mm needs to be updated :).
right now it's just hardwired into p6_lookup(). check self, parent package, parent's parent, etc, up to root. 01:08
s/self/current pkg/ 01:09
soo... what does Package etal do in the way of namespace support? vars? subs/methods? multimethods? 01:10
stevan multi-methods are provided by Class::MultiMethods::Pure
putter our $var;? 01:11
stevan all the multis are inside a single sub,.. which makes them easy to deal with
nope
well,.. yes and no
I was talking with autrijus about this 01:12
we decided that my's would be block/file scoped
and that the package need not know about them
putter k
stevan and our's would be my's which also had an alias in the package namespace
so
this means I only support ours
putter hmm... are non-code vars importable? 01:13
stevan my $FooPkg = $::Package->new('$:name' => 'Foo');
then you do:
$FooPkg->STORE('&foo' => sub { ... });
then $FooPkg->FETCH('&foo');
I do manage the magical $?PACKAGE variable for it with the MM
putter: any type of var is supported 01:14
I dont check anything though
so you can stick a hash in a slot with a @sigil
also,.. ::Module, ::Class and ::Role all implement the FETCH and STORE 01:15
(actually ::Role doesnt yet, but the stub is there)
putter our $x; ==> $FooPkg->STORE('$x', ... Scalar->new() ? Int->new() ?
stevan ::Class->FETCH('$.var') returns the meta-attibute
putter yes, that would be it
actually more like this 01:16
package Foo; our $x;
putter hmm... getting the package name, full and short, back from a package object? 01:17
stevan $FooPkg->name gives you the short
putter and is there any connection between package objects. eg, those for A::B and A?
stevan Packages dont have long names though 01:18
putter err, "A::B::C"?
stevan $FooPkg->STORE('::Bar' => $::Package->new('$:name' => 'Bar'));
rafl j #madwifi
putter ah, ok
rafl Whops.
putter :)
ok, so given that Bar Package, can one crawl up link to FooPkg? 01:19
or otherwise get "Foo::Bar" 01:20
stevan nope, not really
putter (the string/name)
stevan ::Package is not very fleshed out yet
putter k
stevan I suppose that could be done
I just need to think about it
putter ok... so my initial straman idea is to have a package PkgNNNN associated with each Package. PkgNNNN::p6_lookup($var) gets the lexical context, a list of Package's, or their names, from it's associated Package. 01:23
hmm... but since Package/etal will handle any import issues, I think its FooPkg->give_me_packages_to_look_for('$x') 01:25
or better, especially if import does renaming, give me a list of absolute varnames to check. 01:26
stevan Module does the export stuff 01:27
Package is really just a dumb namespace
putter so MumbleBar::p6_lookup('$x') calls BarPkg->look_where('$x') and gets back ['$Foo::Bar::x','$Foo::x','$*x'] 01:28
ok, but hopefully they all have the same api.
stevan yes, they all use FETCH/STORE 01:29
though I need some other things to list all the items in the namespace too
putter ok, let's see.... 01:30
oh, you mentioned some things had changed from mm1. could you summarize?
api wise 01:31
stevan the lack of packages is a big thing
there is no more MetaClass
Perl6::Object is now $::Object
MetaClass is now $::Class 01:32
calling class methods requires the use of a pseudo-package
$Foo->class::bar();
and $Foo->meta has gone away
$Foo is the meta already
putter how badly do we loose if for class C{has $.v}, a C::v method is created but no other note is taken of v's existance? 01:35
actually, looking at the pil, it will be possible to figure out what's an attr. sigh. I look forward to pil2. 01:37
re PIL, if you want to see the PIL for anything, you can do 01:39
./pugs -B JS
jspugs> :pil.yaml class C{has $.v}
ok, are there any questions I should have asked but haven't? 01:41
:)
stevan :)
no idea
putter ok...
stevan I will look over stuff more,.. and think about expanding upon Package's API
putter I was thinking of enabling method foo() is perl5 { q!print $SELF->name()," in p5 code\n";! } and then generating Int/Num/etc/etc from normal p6 files, with the backend creating the appropriate MM stuff. Rather than using hand edited p5 files. Thoughts? 01:44
Any downside immediately apparent? 01:45
Ie, P6::Value/Container would become a Prelude.p6, with some embedded p5 use. 01:47
stevan putter: I really am not familar enough to answer in any meaninful way 01:49
and I am being called by my wife,.. movie time
putter: I will read more of the code tomorrow and have some more questions then
stevan &
putter thanks! 01:50
thanks stevan.
02:07
Does p6 oo support preemptive mixins. Ie, class C does/is D, and if there is a conflict, _D_'s version takes priority. Rather than the usual inheritanc C-has-priority, or Rule mixin cry-conflict. 02:09
s/\./??/ 02:10
It simplifies C is D is CTemplate. Especially if C can be parameterized on D. 02:12
&] 02:19
buu ?eval 1 03:05
evalbot_7086 1
buu ?eval say 1
evalbot_7086 1 bool::true
buu hrm
f0rth ttttttttt
svnbot6 r7087 | fglock++ | * perl5/Hash - an Array or a Pair can be stored into a Hash 03:54
r7087 | fglock++ | %a=(1=>2) works now
KD6PAG Q: Anyone have know where there's sample code that would allow me to do a POST operation with a type="file"? 03:55
obra KD6PAG: see #perlhelp 04:11
buu There's a #perlhelp?! 04:23
xinming_Beijing buu: hmm, I think you should try #perl 04:29
buu Why/ 04:33
(It turns out there really is a #perlhelp, with all of 3 people) 04:34
gaal morning! ah, missed putter :( 05:42
autrijus going to airport in 30 minutes 05:49
it seems that erik meijer is doing some most amazing work in his language-of-choice that is VB
gaal autrijus, .es? 05:50
autrijus with lambda expressions, query combinators, local type inference, etc
gaal eirk meijer, isn't he the function guy with the yellow flower?
autrijus .ee yes
gaal ee, yeah
autrijus I don't remember the yellow flower, but likely yes 05:51
his work is under the LINQ banner
gaal there's a "Hugs98 for .NET" link on his home page :) 05:52
autrijus # msdn.microsoft.com/vbasic/future/de...erview.asp
also the equivalent of $obj.$method 05:53
xinming autrijus: Hi. :-) 05:54
autrijus hi xinming 05:55
GeJ heya all. 05:56
gaal any general tips on debugging monadic code?
autrijus gaal: isolate them into small functions and write unit or quickcheck tests for their behaviour? 05:57
xinming autrijus: I've brought A3E in BeiJing. :'( There is no M2C or M2N left for me...
autrijus failing that, liftIO (print)
gaal autrijus: that you can do Select in the language is quite amazing
autrijus xinming: aw. does it work okay?
gaal: it's just sugar for .Select(c => c.foo)
gaal well! that's something too (to get Dim to answer to .Select) 05:58
xinming autrijus: hmm, Well, Except the speed... I think It performs much poor than yours. :'(
autrijus: by the way, How much does your laptop cost?
autrijus .select:{.foo, .bar}.where:{.val < 10}.orde_rby:{.baz}
xinming: ntd$40k I think 05:59
xinming autrijus: I've just done setting up the Linux System... Might be back for soon...
autrijus gaal: the amazing thing to me is that they made the query combinator range over objects, XML and relational DB as domains
it's like Class::DBI done right ;)
xinming autrijus: ??? How much Us dollars? :-S
obra autrijus: sample code?
xinming: divide by 33
autrijus obra: msdn.microsoft.com/vcsharp/future/linqsamples/ 06:00
obra just got suckered into the new laptop game too :)
autrijus they are in C# but it's just sugar; the underlying dynamism and inferrence works for all CLR languages
erik mentioned on LtU that most of the stuff is ported en masse from HaXml and HaskellDB etc 06:01
which, if true, is very exciting.
xinming I bought my laptop too expensive maybe... - -b 06:02
obra gets excited for hot new language features 06:04
autrijus oh, apparently they range over WinFS too
gaal why not XPath? 06:05
they have sometihng like template::extract to query xml
autrijus gaal: this to xpath is like XML::Twig to xpath
gaal re: your remrk autrijus from the other day about parsers not being symmetrical with queries
(hoping i got that right) 06:06
autrijus nod. I remember XQuery
but you can't use XPath/XQuery as a templating language
you need XSLT for that which is nonreversible 06:07
gaal I'm not sure I understand the analogy viz. XML::Twig.. 06:08
they have the example Dim CountriesWithCapital As XElement = (reverse template)
i should study what they're doing more, probably, 06:09
but if it's a *query* then a reverse template is certainly quick to write, but an xpath expression is probably more powerful?
but never moose, I'll read this more carefully while you're on the plane :) 06:10
autrijus moose :) 06:11
gaal soon in an Acme::MetaSyntactic module near you :-)
masak :) 06:16
masak still hasn't really figured out what this "moose" is 06:17
gaal forum2.org/moose/
masak reads
weird 06:19
and i don't really see how it relates to XSLT :) 06:21
or is "never moose" just a silly expression? yeah, probably
ok, never moose :P 06:22
gaal meditate on it and you'll see it
you're certainly getting there :-)
hcchien I saw the control quickref, and interesting in "for @foo -> $x, $y, $x { ... }". how does it work?
autrijus hcchien: it counts arity of the lambda argument, see 3, then take 3 a time 06:23
hcchien autrijus: yes. I think so. but $x will be 1, 4, 7 if the array = (1..10)
won't 3, 6, 9 overwrite the $x? 06:24
autrijus I think it should be an error myself :)
masak nod 06:25
having the same argument twice leads to insanity
and why would one want to?
gaal i can imagine using undef to skip args 06:26
hcchien if I just want to skip the first one? :-)
autrijus or _
but undef is more perl6ish
I mean perlish
but I need to run 06:27
see ya again in a day
gaal for @foo -> undef, $x, $y -> { ... }
autrijus waves &
gaal bye a
_ will end up synonymous with undef :)
(actually, no, undef in some places, Any in others...) 06:28
masak ?eval my @a = 1..10; for @a -> $x, $y, $x { say $x } 06:29
evalbot_7087 1 4 7 Error: Insufficient number of required parameters: 1 actual, 2 expected
masak ha-haa!
?eval my @a = 1..10; for @a -> undef, $y, $x { say $x }
evalbot_7087 Error: unexpected "f" expecting ";", statements or end of input reserved word
masak aww :(
add a test? :) 06:30
hcchien hmm...., _ doesn't work for me.
masak hcchien: i don't think it should 06:31
gaal i think it's not specced, so p6-l it first? (better check the synopses first)
masak all of them? that's like, O(n)
gaal hccien, it shouldn't work... it's how that kind of thing is spelled in haskell
masak, heh, no, s04 should be enough :) 06:32
hcchien oh, right. so autrijus said that undef is more perl6ish.
masak gaal: oki :)
hcchien but evalbot doesn't like undef.... :-)
gaal hccien, right, i was joking about how p6-l posts from people like yuval and luke are getting haskellish 06:33
masak, `The "for" statement` doesn't mention it.. i'd look for the definition of pointy subs, maybe s06 06:34
masak hcchien: no, it's not a feature that's in perl6, yet. i'm going to suggest it on the mailing list
gaal: oki
gaal .....which doesn't mention it either :) p6-l query time i think.
masak yep, i'm on it :)
gaal masak++
hcchien masak++ 06:35
and pugs won't supprt for (;;) anymore? 06:38
gaal oh, see the "loop" statement 06:42
in s04 i believe.
yes, s04. loop $i = 0; $i < 10; $i++ { ... } 06:43
you can also spell infinite loops as loop { ... }
hcchien oh, ok. I saw that. 06:44
masak gaal: henceforth, i will mercilessly exploit you as my index for synopses :P
at least s04 and s06 which you seem to know well 06:45
gaal laughs
masak does @_ still contain all the arguments in a sub in p6? 07:05
if i write a sub which sometimes takes a hashref as a first argument, and sometimes not, is that a good time to use a multi sub? 07:09
gaal masak: @_ still contains them all
and probably yes :) 07:10
but currently at least, @_ is only defined if you don't provide an explicit signature
i don't know if this is specced actually. 07:11
but s06 will tell you :)
masak :) 07:12
reason i'm asking is, i've started to look at GetOptions in Getopt::Long, and it does a lot of @_ checking 07:13
i'm thinking of the best way to p6ify those idioms
gaal that's a sure sign you want multis with explicit args.
masak i think so too
a lot of code will just disappear into signatures :)
gaal i gave a talk in my local pm group about p6 for p5 programmers 07:14
the warmup was about function sigs
shooting fish in a barell
barrel even. 07:15
GeJ gaal: in your post yesterday you suggested that one might not expect the arguments/options to parse to not necessarily belong to @ARGV, I'm a little lost there. Do you have an example in mind? 07:16
masak shooting fish in barrel? is that a classic example?
GeJ s/belong to/come from/
gaal GeJ, one example is tests. you want to enter the getoptions function many times; setting a global var each time is, well, quaint but not pretty. 07:17
another example is when you provide a CLI of your own from within a tool
eg., something with a "console".
masak, what i meant is that the talk was intended to show people who already like p5 why p6 is better/needed 07:18
and we've developed blindness to many quirks and misfeatures
nothingmuch morning 07:19
GeJ morning nm
masak morning
gaal function sigs save you (well, me) from the stupidest bugs, and save lots of typing
they have other benefits as well of course :)
(reflection)
heya nuffin
GeJ gaal: hum, I see... interesting, didn't think about this one. Something I'll have to keep in mind as I'm considering doing a small CLI client for a project I'm working on. 07:20
gaal so take a bunch of "we're doing just fine" p5 programmers, who might be worried about their significant investment in p5
and get them to like p6 07:21
masak gaal: luke palmer took your _ joke seriously. that's sort of funny in itself
gaal not hard really, since p6 really is wonderful :)
'tis. :)
GeJ Well, I know that to some people around here, damian is not the more popular person in the world, but his IO::Prompt seem sexy enough to give it a try. I wonder how he handles this. 07:22
gaal Damian's great! His code's just hard for mortals to understand. 07:24
GeJ nods
gaal GeJ: you can completely frontend GetOptions with something that does the @ARGV localization for you. Trivial code, makes your client code much prettier 07:26
in a p6 version we'll just have multi sub GetOptions(%spec, ?@source = @*ARGS is rw) returns Getopt::Long 07:28
GeJ masak: you're on this? 07:29
gaal hmm, need to be clever with the other variant so that GetOptions(opt1 => $target.... source=>@*ARGS) doesn't preclude an option called "source" 07:30
masak GeJ: sort of, still muddling over it, but yes 07:32
GeJ m'kay 07:33
gaal [very off topic] say, anyone know the reason for the naming of java's reflection "forName" method? what does "for" signify there anyway? 07:35
nothingmuch Base64.t still hangs
GeJ I saw that Mister A. set the 6.2.10's poem. A release schedule is planned already? 07:37
gaal doesn't know
GeJ Just to be sure, a Haskell guru will implement stevan's MM2 into Pugs, and then all (or most of if @Larry decide to change things) the OOP syntax will be available. Am I correct? 07:41
That's the way I understood it. MM and then MM2 was sort of a blueprint for autrijus to code it in Haskell. But I'm prolly wrong. 07:43
gaal yes and no; having much of the implementation in Perl means other compiler implementations can reuse it 07:45
nothingmuch coughs 07:47
gaal what? 07:50
nothingmuch the Blondie project is all about that 07:51
have the language define the prelude by itself
(horribly inefficient, but reusable and useful)
and have the compilers provide builtins for prelude functions as they see fit
gaal is it clear that all of the MM can be expressed in the prelude?
nothingmuch i don't know, but most of it can 07:52
gaal putting aside circularity for a moment.
nothingmuch and the stuff that can't can be stubbed
forcing the compiler to provide a version
as for circularity - it's also useful
gaal sure
but it's a different problem.
nothingmuch if the prelude is defined circularly, the compiler can choose where to break the circle
gaal bbiab 07:55
gaal nothingmuch: i acknowledge this is cool and potentially very useful 08:00
i'm just saying, it's an experiment, and if it turns out to be too hard 08:01
or too slow
pugs can still do OO prims in haskell
nothingmuch it's actually both easy and fast
gaal good :) 08:02
dada pugs build is borked
nothingmuch (the output... I don't know about compilation yet. It works reasonably fast, but it's not optimizing, or anything)
gaal dada, error?
dada NMAKE : fatal error U1073: don't know how to make 'pugs.exe'
Stop.
gaal oh, that's annoying frontend stuuf again. 08:03
nothingmuch (the process itself is very simple though - take the hash of each node being applied, and see if it can be replaced)
gaal dada: make sure you have a file called "pugs" (no extentsion)
that's a perl5 sctipt
then, create a file "pugs.exe", never mind what's in it
dada yes, I have it
gaal then run nmake again
dada let's try 08:04
seems to be working 08:07
gaal this is a weird extension thing happening on win32 since `pugs` has been made a script 08:11
i'm trying to catch putter to see how bad we need this script... i can see that it saves work, but it also causes problems 08:12
following autrijus' mention of LINQ, this is an interesting and well-written paper: research.microsoft.com/~emeijer/Pap...l2003.html 08:40
autrijus rehi from CKS airport 08:48
boarding in two hours :) 08:49
GeJ rehi :)
clkao Mr Tang 08:50
autrijus Mr Cow
clkao Mad Tang
autrijus :) 08:51
rafl How about a release timeline for 6.2.10? Is there anything known? 08:52
autrijus I wonder if I can do that en route to estonia 08:53
but more realistically, it will be on time for euro oscon
since I have ~2wk of time
gaal cool, i have a talk about pugs on 6 oct 08:54
rafl Fine. I thought the release was already planed for the beginning of this month..
gaal: Where?
gaal maybe we can release then :)
my local pm group
tel-aviv 08:55
rafl Ah, OK.
gaal autrijus, was the flight delayed, or were you just very lucky with check-in and that stuff? 08:56
rafl Maybe someone of you is from germany and wants to give a pugs/perl6 talk as well? I'm co-organizing a linux-conference here.
gaal iblech maybe?
autrijus gaal: no, I misremembered the time 08:57
fortunately on the right direction
gaal heh
rafl gaal: Hm, havent seem him for a while..
gaal i know a chronic latecomer who sometimes manages to fool himself by setting his clock wrong
autrijus, can you maybe take a look at the lexical pragma thingie? 08:58
the idea is that in BEGIN time i gather pragmas to be installed
and the parser puts them in the AST
autrijus which test is that?
gaal but it's putting it in the wrong place 08:59
t/packages/lexical_pragmas.t
nothingmuch autrijus: remember what happenned with my flight?
;-)
gaal nothingmuch: i met a girl in mongolia who didn't understand how come her visa to china was so much longer than everyone else's 09:00
turned out she read the expiry date as dd-mm-yy
it was yy-mm-dd
nothingmuch oh crap
gaal yeah, she had to leave in a hurry 09:01
GeJ rafl: where are you from?
nothingmuch china is not the place to be doing that, is it?
gaal luckily she had two days left
nothingmuch phew
nothingmuch finished the setup bug... that was a pain
rafl GeJ: Germany, Saxony, Chemnitz.
gaal she wouldn't have gotten into any *real* trouble:
just denied reentry, and have to wait a bit longer in mongolia to get a new visa 09:02
costly, but mongolia is not a bad place to be in :)
in the meanwhile that is.
nothingmuch aye
nothingmuch would like to hike/bike in mongolia
GeJ rafl: ok. Germany, Bayern, Munich myself.
dada !!! 09:03
rafl GeJ: That's next to where iblech lives, afaik.
dada "pugs.exe is not a valid Win32 application." :-)
gaal daad: yeah, run pugs_bin.exe
s/ad/da/
we have to sort that out :) 09:04
dada sheesh
gaal alternatively, perl pugs
dada doesn't work either
Can't exec /bin/sh at pugs line 1.
gaal beh
perl -X pugs 09:05
dada same
gaal sorry, lowecase -x
dada oh yes, works now 09:07
sort of...
I see intermixed shell and pugs prompts!
and when I quit pugs with :q, the shell closes! 09:08
gaal autrijus: oh, t/packages/lexical_pragmas.t won't parse now, change the install local func in Paser:913 to install in the local env rather than the caller
that isn't right either, but at least the test will run :-/
dada, then run pugs_bin.exe directly :( 09:09
dada gaal: yes, that's what I'm doing
autrijus k 09:11
hm, maybe we should make a pugs.bat
dada oh well, nmake test doesn't work either
autrijus for win32
dada or just: copy pugs_bin.exe pugs.exe 09:12
autrijus but that will prevent pugs/perl5 from running
which is the raison d'etre
nothingmuch has anyone seen iblech lately?
autrijus another thought is move the wrapper into haskell space
Main.hs
gaal nothingmuch: not on channel, but he has made some commits
autrijus: that's what i thought - is it just command line parsing the wrapper is trying to avoid? 09:13
autrijus yes
nothingmuch wants to get people to help with Blondie
any volunteers for learning?
autrijus gaal: you can probably do that?
dada who's Blondie?
nothingmuch dada: an intermediate language that i'm developing to prototype some compilation approaches
gaal autrijus: i'll give it a whirl; wanted to talk with putter to make sure i understood the motivation first 09:14
nothingmuch type inferrence based pessimistic unboxing
and selective replacement of prelude
autrijus gaal: the motivation is -BPerl5
nothingmuch has problems with the wrapper
it should be built by the makefile to use $^X 09:15
not base on $PATH like it does now, or /usr/bin/perl like it did before
gaal nothingmuch: it'll still have problems on w32 then
and anyway there should not be a perl5 interpreter involved for most programs
autrijus right
gaal i bet this slows down smokes by a lot
dada $^X is ok on win32 09:16
svnbot6 r7088 | autrijus++ | * Import GHC.Exts for DrIFT.Perl5; that's the only change required by GHC 6.4.1-release
dada it always returns the full path to perl.exe
autrijus C:\work\pugs>perl pugs
Can't exec /bin/sh at pugs line 1.
dada autrijus: perl -x pugs
gaal autrijus: so the underlying motivation was to get p5 $^X in pugs? we can sure throw that into haskellspace during the build too. 09:17
nothingmuch yes, and we should
the perl that the user selects to run Makefile.PL should be consistently used
not guessed later
gaal++
autrijus yes 09:18
dada: not acceptable really
as old versions of pugs will have a pugs.exe installed
gaal what's the helper named perl5 anyway? i don't have an executable with that name on my system
dada autrijus: you mean the copy is not acceptable? 09:19
autrijus "perl -x pugs" is not
gaal: mm? it's perl5/PIL-Run that runs perl5 helper
dada oh well, I did a pugs.bat which contains "perl -x pugs" 09:20
seems to work
autrijus dada: will it win over .exe?
dada I deleted the fake .exe, of course
autrijus no it would not
gaal a .com might win... :-p 09:22
nothingmuch eep
gaal what? .com is great
load to ram, execute
no header muckiness
gaal hides
autrijus riight
gaal (p5 $^X in pugs coming up as soon as this sanity build finishes) 09:23
dada I can't come up with a proper name for a pugsbugs test 09:24
gaal btw: i can't seem to build a pugs with precomp prelude and with -O0 -- ghc dies horribly in the second phase, apparently because an unoptimized precomp prelude is huge 09:25
dada ohhhhhhhh wait wait
gaal wait actually
why should it be any different in size?
dada t/examples/examples.t is borked!
it looks for pugs.exe, hardcoded in the file
now that I have pugs.bat, it miserably fails 09:26
nothingmuch www.instructables.com/
autrijus mm I'm seeing 09:27
C:\work\pugs>pugs_bin t\01-sanity\08-test.t
pugs_bin: user error (***
unexpected ":"
expecting operator or "!!"
at C:\Perl6\lib/Test.pm line 100, column 40
t\01-sanity\08-test.t line 1, column 1)
probably bad old c:\perl6
aye 09:28
dada I don't have 08-test.t there 09:29
svn up again?
mmm no, is there
1..1 09:30
ok 1
ok, what are pugsbugs tests supposed to do? fail if they hit the bug? 09:34
svnbot6 r7089 | gaal++ | add perl5path to config (contains the $^X of the building perl5)
r7090 | gaal++ | add tests asserting a module is loaded once, but &[un]?import hooks get called many times
gaal you can :todo<bug> them 09:35
make sure they do pass once the bug is fixed
it's okay to put parsefailing bugs there, usually in a file of their own 09:36
though in the release they'll probably be skipped.
autrijus "nmake smoke" on win32 fails due to the missing pugs.exe 09:37
gaal i haven't tried a smoke since before my vacation ;-)
we really have to unwrap pugs 09:38
i'll work on it today.
autrijus wonderful 09:39
seeing that people maybe interested to svn co and try it out in 48 hours
that's none too soon :)
dada wow, committed :-)
autrijus You probably don't want to export an operator name; instead
define a new variant on the new operator (eg. multi sub *infix:<+>):"&infix:%"
at "blib6/lib/Set/Symbols.pm" line 145 column 1
gaal eep, why are installbin and other config directives pointed at perl(5) paths? 09:43
please nobody `make install` :-)
autrijus @x = (a => %x, b => %y) 09:44
svnbot6 r7091 | dada++ | added test for bug: bool::false not parsed correctly on LHS of an operator
autrijus this is kosher right? no need to
(a => hash(%x)) or (a => \%x) or whatnot 09:45
gaal i hope so? doesn't => impose Item context? 09:46
autrijus right
fixed
gaal hopes to pick up that "be blazingly fast" trick while autrijus is in .il :) 09:47
so, Pugs.Main:84 already had run ("-B":backend:file:_) = slurpFile file >>= doCompileRun backend file 09:48
what was wrong with that? just the $^X? or something else?
ahhh, the comment in /pugs:41 about the hack needing FindBin? 09:49
svnbot6 r7092 | autrijus++ | * NameGen: Hash in pair value position is just fine as => imposes Item context 09:50
nothingmuch goes to have lunch 10:00
autrijus gaal: yes, precisely that 10:11
also the runtimes needs to be installed from perl5/
but that is somewhat orthogonal
gaal i've probably no time to sort out the perl5 installation path issues - please someone take a look, this is critical. 10:12
azuroth how do I go about getting a job in melbourne? 10:13
gaal autrijus: i'll definitely do the unwrappage.
autrijus gaal: wunderbar.
the perl5 installation foo needs a rethink
I think we don't want relocate perl5
yet separate Makefile.PL sucks 10:14
so probably sometihng intelligent in the mainMakefile.PL
gaal why is there any perl5 installation info at all? very scary :)
autrijus mm?
gaal why does pugs contain any perl5 paths, except for the $^X it was built against? 10:15
i mean right now it just looks like a bug someone accidentally injected:
./pugs_bin.exe -V:installsitebin
installsitebin: c:\perl\bin 10:16
make install will do terrible things again
autrijus right... it's a bug
though it's somewhat fine, as it only installs .exe files
and iirc has been like that for a while
gaal if we avoid reusing these names, we can maybe avoid that scenario
oh?
autrijus I agree it's wrong
OT: is there some way, via CSS 10:17
to scale fonts so that it stretch to fit?
gaal i hope we get an alternative to EU::MM for p6 anyway (though i ain't working on it :-)
not that i know. you are going for takahashi eh? :) 10:19
autrijus yes. :)
gaal (my css-fu is very limited)
autrijus nod
gaal that ruby presentation used images, 'fcourse
autrijus mm font-size-adjust, font-stretch
gaal wow, zvon.org do it again 10:21
they have references for so many interesting things :)
nnunley phears Blondie.
gaal autrijus: those properties don't seem to work on my browser (firefox 1.06/win32) 10:23
autrijus apparently font-size-adjust is the way to go 10:24
ok, so a Font::AFM and font-size-adjust is totally the way to go 10:25
either precalculated or via javascript
I should be working on the content but yet I'm hacking away at the style 10:26
what's wrong with me :-/
gaal you must be human! 10:27
careful, that's mildly contagious and eventually lethal 10:28
autrijus purrs 10:31
GeJ hum, don't want to sound like an ass or something, but : 􏿽xAB[10:49] <autrijus> boarding in two hours :)􏿽xBB and it's 12:40. 10:40
Happy hacking in an airport is fun, but missing the plane is a pain.
just my $0.02 though
autrijus 16:55 < autrijus> gaal: no, I misremembered the time
GeJ Oh, oki. My mistake. Sorry for the noise :) 10:41
autrijus np :)
I'm downloading visual haskell
hopefully will finish on time
gaal visual haskell!? 10:44
visual haskell++#6 :-) 10:45
autrijus: when in a rush on shared media, decrease everybody elses MSS! and tcpkill their connections! be ruthless! we don't need ruths! :-p 10:46
autrijus ROFL 10:48
nothingmuch returns from lunch 10:49
nnunley: want to get involved? 10:50
azuroth what sort of things need porting to perl 6? 10:57
nnunley nothingmuch: Would love to
autrijus azuroth: your favourite program :)
nothingmuch azuroth: Getopt 10:58
azuroth masak's working on getopt, I think
nothingmuch nnunley: what do you know?
azuroth: in that case, XML::SAX on steroids
the perl 6 object model lets us do such things more cleanly
azuroth hokay. I'll look into it
nothingmuch another option is taking the lessons learned from DBIx::Class, and trying to consult stevan on a CDBI persistence layer based on the metamodel 10:59
such that the meta class gives the persistence, instead of the super class
nnunley nothingmuch: Be more specific. :) I have some basic compiler theory under my belt. A long time ago I looked at doing an AST based reduce/apply engine, but didn't get that far.
nothingmuch nnunley: i mean specific to blondie
the blondie todo:
nnunley nothingmuch: Ahh. I've just been tracking it since you added it. 11:00
nothingmuch refactor Param into Thunk
make the type inferrencer support higher order functions
and compile to PIR
oh, and support union types (it can be either this or that, we can only find out at runtime)
do any of these interest you? 11:01
nnunley nothingmuch: Union types, ala SSA?
nothingmuch if so i can help you get started... i'd really love for someone without my bias to look at the code
i dunno
let me find out
what's ssa?
autrijus mm 463 fails, a mild releng challenging
nothingmuch structures semantic analysis? 11:02
nnunley nothingmuch: Single Static Assignment, IIR the acronym correctly.
nothingmuch ah
autrijus "tagged" union types?
aka variant types
nothingmuch we don't have assignments yet
autrijus because there are also untagged union types aka (+)
nothingmuch i think you could say that blondie is already SSA 11:03
nnunley nothingmuch: en.wikipedia.org/wiki/Static_single...nment_form
nothingmuch because parameter binding is
autrijus plugs ANF in stead of the isomorphic SSA
nothingmuch so what does a union type have to do with SSA? 11:04
gaal (i must do some $work, will only be able to commit the unwrapping later)
nothingmuch en.wikipedia.org/wiki/ANF
autrijus: please fill it in
autrijus nothingmuch: A Functional Perspective on SSA Optimisation Algorithms 11:05
nnunley nothingmuch: it seems to be similar in intent to the theta operator for SSA.
autrijus it's the functional equivalent of SSA
it's much happier than SSA because all it has is binding and lambda
PIL was glorified ANF.
nothingmuch cannot find theta on the ssa page
autrijus: what is the TLA?
autrijus Administrative Normal Form
nnunley Err... Sorry, Phi, not theta. 11:06
nothingmuch nnunley: please wait, reading
autrijus: please put a link in this page: en.wikipedia.org/w/index.php?title=...ction=edit 11:07
autrijus nothingmuch: I'm boarding _right now_
nothingmuch oh, okay
autrijus so probably not this moment
gaal have a nice flight!
autrijus ask #haskell to fill it in
gaal "Administrative Normal Form"!? probably in triplicate.
nothingmuch autrijus: don't crash or anything 11:08
autrijus aka A-normal form
nothingmuch:
maybe I'll land to an island with fiber optic network etc
nothingmuch hah 11:09
nnunley: no, not really like phi 11:10
simpler
nnunley Ok.
nothingmuch assume you assign a value into the array element 1
and a value of another type into element #2
then you take out element i
the compiler cannot know (unless the compiler is really crazy) what type willc ome out of the array 11:11
so the array must return boxed types
as for phi - the moment we have assignment we will add ssa ;-) 11:12
nnunley Couldn't the type be widened until it encompases both the types of elements 1 and 2? (Of course, I'm making this a 'simple matter of coding')
nothingmuch: Got it.
nothingmuch yes
that's the point of the excercise
the type of the array becomes [PV|IV] if the array is assigned 1 and "foo" 11:13
and every value coming out of the array must be checked for at runtime
nothingmuch doesn't like acm 11:16
they should give me all their articles for free
nnunley They definitely should. It's a nice big black hole.
I guess that's why I almost always hit citeseer first for research.
nothingmuch was just clicking on links 11:17
the problem is that most articles are unaccessible to me
i lack the vocabulary
so it's really not worth $10 to figure out that I don't understand the language 11:18
nnunley nothingmuch: Right. Same issue with me. I have to muddle through the notation. Which is one of the reasons why citeseer is nice -- it's free.
Aankhen`` Hi. 11:19
nnunley But i guess it would be good to try to put together a primer.
nothingmuch citeseer.ist.psu.edu/cis?q=administ...s&cs=1
Aankhen`` What is CiteSeer?
The search engine?
I see.
nothingmuch nnunley: so, what would like to do for Blondie?
Aankhen`` Don't mind me, just talking to myself.
nothingmuch hola autrijus
err
Aankhen``:
nnunley Aankhen``: ok. ;)
Aankhen`` Hi. :-)
Aankhen`` strangles nnunley. 11:20
nnunley nothingmuch: Good question. What are you having problems with?
nothingmuch nnunley: nothing really, except lack of time 11:21
nnunley nothingmuch: Got it.
nothingmuch and the fact that I'm the sole author of Blondie
which means that i get no code level critique
and the truck number is high (this is dangerous with me, i tend to burn out)
nnunley nods.
nothingmuch err, low 11:22
Aankhen`` Argh... how do I stop the Parrot bugs from being e-mailed to me?
nothingmuch Aankhen``: procmail?
Aankhen`` No, try again. :-P
Oh. Unsubscribe from p6i. 11:23
nothingmuch aha
citeseer.ist.psu.edu/flanagan93essence.html
nnunley nothingmuch: Where is your todo document?
nothingmuch nowhere
i'll commit one
nnunley Good good. 11:24
Hrm. So you have several placeholder nodes in Nodes? 11:27
nothingmuch huh?
nnunley Sorry, reading through Blondie::Nodes 11:28
nothingmuch one second, i'll figure out what you're talking about ;-)
what placeholders? 11:29
Stub is an unimplemented thunk
and the placeholder type is used by the type inferrencer to pass types from syms to params
nnunley Sorry, I'm probably going to make you explain it to me like I'm a 2 year old. The Blondie nodes file's intent is to create a reasonibly direct mapping into PIL? 11:31
nothingmuch nope
it has nothing to do with PIL
the Nodes file defines the nodes of the blondie AST
they are documented in Blondie.pm
nnunley rereads. 11:32
Hrm. So why the pattern of Thunk(Stub())? 11:36
Instead of just Stub?
Hrm. The flanagan paper is definitely pretty good. 11:38
'Oooo. Shiny.' 11:39
nnunley prints it out. 11:50
svnbot6 r7093 | nothingmuch++ | r8223@syeeda: nothingmuch | 2005-09-22 14:45:34 +0300 11:51
r7093 | nothingmuch++ | TODO for blondie
nnunley nothingmuch++ Thanks. 11:53
nothingmuch: BTW, minor grammatical nit (I'm sure you'll catch me in many) -- "its" is possessive, "it's" is a contraction of "it is" or "it has" 11:55
masak what should the signature be of a function that is called like this: Getopt('a' => 'b', 'c' => 'd')? 12:00
There can be any number of pairs (I assume those are pairs) in the argument list.
nothingmuch returns 12:15
Thunk(Stub()) is because... errm 12:16
i forget
bottom line: because a stub thunk can take params for the purpose of type checking
(it doesn't right now, because this info is thrown away too early) 12:17
and a stub is mearly something that the compiler should refuse to pass on
think of Stub as the '...' operator
you son't say 'my &foo = ...;'
you say 'sub foo { ... }'
masak: a list 12:18
for the getopt sig
*@pairs
hola kolibrie 12:19
kolibrie hola nothingmuch
svnbot6 r7094 | nothingmuch++ | s/it's/its/ 12:20
masak nm: thx
nothingmuch masak: ? 12:21
oh, oops
masak for the signature
nothingmuch i confused you with nnunley
masak but the answer was for me, right?
nnunley grins. 12:22
masak nothingmuch: anyway, it didn't work :)
kolibrie so, I'm trying to port HOP make_lexer (section 8.1.3 page 374) to Perl 6
I'm having prototype problems, I think
nothingmuch yes, it was
and I'm sorry ;-)
kolibrie: is can you link to the online code sample? 12:23
kolibrie the original? /me looks for url
nothingmuch aye
kolibrie hop.perl.plover.com/Examples/ALL/Lexer.pm 12:24
nothingmuch kolibrie: please explain in a few words roughly what it does 12:25
masak aha!
masak discovers 'is export'
nothingmuch aha!
masak :)
now everything works 12:26
kolibrie the tokens subroutine expects an input iterator, and then a token definition containing label and regex
masak or rather, all my tests fail instead of not running
nothingmuch =)
what is a label? 12:27
kolibrie the make_lexer subroutine takes an input iterator, and a list of token definitions, and strings a bunch of calls to 'tokens' together
the label is a string identifier
anything you want
nothingmuch okay
oh, i see 12:28
make_lexer stacks tokenizations on top of each other
kolibrie my problem is figuring out how to prototype make_lexer, and then to appropriately pass to tokens
correct
I have something like: sub tokens ($input, $label, $pattern, ?$make_token = sub { ... }) { ... } 12:29
nothingmuch sub make_lexer ($lexer is copy, *@definitions) { for @definitions -> $args ( $lexer = tokens($lexer, @$args) }; return $lexer }
yes, that's about right 12:30
kolibrie that looks pretty
nothingmuch oh wait 12:31
svnbot6 r7095 | iblech++ | * Usual svn props.
r7095 | iblech++ | * New t/var/autoref.t: $arrayref = @array, (key => @array), etc.
r7095 | iblech++ | * t/pugsbugs/parsing_bool.t: s/eval/try/ so other backends no capable of eval()
r7095 | iblech++ | have a chance of passing this test.
nothingmuch you need to flatten *@args
i forgot how
to allow it to be positionated ;-)
(make an array into a list of positionals)
kolibrie S04? says *@$args
nothingmuch right
kolibrie or **@$args
nothingmuch i'd go with *@args
no need to steamroll 12:32
kolibrie :)
nothingmuch just explode
kolibrie does that compile for you? I get an error
ah, just a s/<'('>/{/ 12:36
svnbot6 r7096 | masak++ | number of tests in plan corrected in gol-linkage.t
nothingmuch oops... sorry =) 12:40
kolibrie at least your's compiles now 12:41
I can't remember what I had, but it kept not compiling
ah, now I'm getting a runtime error: *** No compatible subroutine found: "&tokens" 12:47
how is that? 12:48
nothingmuch nopaste the code? 12:52
kolibrie in feather.perl6.nl:/home/kolibrie, see Lexer.pm, and pugs extract_date.p6 test_date.txt
nothingmuch try giving tokens a prototype 12:55
oh wait
do away with the 'sub tokens { ... }'
i think it's confusing ppugs
kolibrie hmm. same error still 12:58
svnbot6 r7097 | nothingmuch++ | more todo 12:59
nothingmuch een.wikipedia.org/wiki/GENERIC # i want to compile to this after PIR 13:02
and also another interpreted backend
kolibrie oooh, that would be nice
nothingmuch the more backends Blondie has the more we can learn about what is the true role of an AST as high or low level as Blondie
and give this knowlege back to PIL 13:03
nothingmuch is going to to a blondie interpreter in C later
it shouldn't be too much of a mess
gaal nothingmuch: re your reply to masak: we can do better and specify it's a list of pairs, no? i think the syntax is `*@list of Pair` but have no idea if it works :) 13:04
rehi btw :)
nothingmuch right
hi ho =)
nothingmuch leaves right about now
gaal bye nothingmuch. 13:10
interesting idea for build speedup: 13:45
# www.isi.edu/~hdaume/HAllInOne/
creates one big file instead of many source modules 13:46
doesn't support ghc extensions yet
obviously, not good for when you hack on one file a lot 13:47
but something to follow and possibly enable in releases one day.
autrijus nod 13:55
hi from .hk btw
xinming_Beijing autrijus: are you in a aeroplane? 13:58
autrijus no, in .hk for transit 14:00
gaal hello! :) 14:02
Khisanth you are at that giant HK airport? :) 14:03
xinming_Beijing cool... the wireless lan is really cool. :-P 14:05
gaal how long is your connection? how long is your next flight? 14:06
nothingmuch ftp://gcc.gnu.org/pub/gcc/summit/2003/GENERIC%20and%20GIMPLE.pdf
autrijus Khisanth: aye
nothingmuch autrijus: please put a note in your journal, that blondie is looking for backend ideas 14:07
nnunley nothingmuch: I was under the impression that GENERIC wasn't exposed.
autrijus nothingmuch: you can write my journal for me in docs/journal/ :)
or you can wait for two days...
nothingmuch i want to try as many backends as possible so that we force Blondie into a reusable shape and then throw that away
nnunley nothingmuch: LLVM could be a good target. AS would JASMIN
nothingmuch and reimplement our knowlege in PIL
autrijus nothingmuch: okay. I nominate YARV 14:08
nothingmuch nnunley: link for jasmin?
gaal the unwrapping stuff, by the way, is needing a reasonable way to handle pathnames in haskell :-(
autrijus the standin for parrot in my mind -- after clr that is
gaal: Distribution.Compat.FilePath
nnunley jasmin.sourceforge.net/ 14:09
autrijus my flight is in 1hr
I'm finally starting to write slides
nnunley I'm sure there's a better java assembler out there, but that's the one I remember. 14:10
gaal autrijus: i'm getting errors about that being "locked", asking on #haskell
nothingmuch is YARV stack based?
nnunley nothingmuch: Yes. 14:11
autrijus yup.
nothingmuch good
nnunley www.atdot.net/yarv/insnstbl.html
autrijus but has first class continuation support
nothingmuch ooh
autrijus has to, since ruby has callcc 14:12
bbiab 14:13
back
kolibrie anyone want to diagnose my: *** No compatible subroutine found: "&tokens" 14:21
on feather.perl6.nl:/home/kolibrie 14:22
pugs extract_date.p6 test_date.txt
fire alarm &
autrijus try to get some electricity & 14:31
svnbot6 r7098 | autrijus++ | * first few bits of CUFP talk. 14:39
Juerd Anyone with me on the issue of interpolation and stringification resulting in different values being a bad idea? 14:44
nothingmuch yes, me 14:45
QtPlatypus Juerd: Can you give me an example?
Khisanth QtPlatypus: well $foobar1 = "foo"~$bar and $foobar2 = "foo$bar" and $foobar1 ne $foobar2 :) 14:48
nothingmuch Juerd: who's idea is it to separate the two? 14:55
svnbot6 r7099 | nothingmuch++ | Blondie propaganda 14:57
kolibrie back
nothingmuch autrijus: journal/2005-09-22.kwdi
Juerd nothingmuch: Damian's 15:07
nothingmuch suspects that Damian is a double agent
Juerd Message-ID: [email@hidden.address]
nothingmuch by adding insane enough semantics he will force us all to use C# eventually 15:08
Juerd So far, I have more confidence in Damian than in Larry 15:09
Khisanth is there suppose to be some advantage in making them different? 15:11
Juerd Just that they can return different values
I think that this "advantage" is a very big disadvantage. 15:12
I think ~$foo, "$foo" and $foo.as(Str) should all return the same thing
And, of course Str($foo)
Khisanth agrees 15:13
Juerd Anyway, please do participate in the p6l discussion if you have an opinion
nothingmuch "bah $foo moose" should be a special form to make stuff into "bah " ~ ~$foo ~ "moose"
Juerd Where the unary ~ is redundant
Because binary ~ already forces Str context
nothingmuch was only making a point 15:14
nothingmuch still can't think of any scenario where "$foo" and "" ~ $foo ~ "" should behave differrently 15:18
or ~$foo for that matter
kolibrie if you want a different context, couldn't you: "" ~ +$foo ~ "" 15:31
nothingmuch yes
kolibrie so ~ should definitely always be string context, like "$foot"
"$foo", that is
svnbot6 r7100 | stevan++ | Perl6::MetaModel 2.0 - 15:33
r7100 | stevan++ | * A (mostly) complete port of Perl 6 Test::Builder to the MM 2.0
r7100 | stevan++ | - including tests as well
Juerd Read [email@hidden.address] about pairs stringifying differently in string context than they interpolate 15:35
nothingmuch i disagree with that 15:37
svnbot6 r7101 | stevan++ | Perl6::MetaModel 2.0 - 15:39
r7101 | stevan++ | * some minor changes to support the Test::Builder port
kolibrie if you want different interpolation semantics: "hello { +$foo } " 15:40
nothingmuch errm, what about just
"hello { $foo.as(...) }" ? 15:41
kolibrie that works too 15:44
nothingmuch Juerd: i sent some protest with some heavy reasoning 15:48
i hope it helps
svnbot6 r7102 | stevan++ | - docs/journal/ -> adding information about the Test::Builder port 15:50
r7102 | stevan++ | - Test::Builder - fixing a bug I found while porting
expeditio hi. is this a perl help channel? if not, where is there one? thanks. 16:00
nothingmuch #perl
this a genera perl 6 channel
expeditio k, thnx 16:01
svnbot6 r7103 | stevan++ | Perl6::MetaModel 2.0 - fixing attribute accessor and $:singleton property of Test::Builder 16:38
nothingmuch is addicted to wikipedia
s/addicted/destructively addicted/ 16:39
putter hi all 16:52
briefly...
nothingmuch hi ho
putter gaal: ./pugs does not currently well support Win, as I lack that platform. Two possibilities: 16:53
It is not possible to portably write a CPAN module which installs perl scripts. That seems unlikely, but I supposed is just borderline possible. Then unwrapping is definitely needed. 16:54
If it is possible, then, well, someone needs to read man MakeMaker or whatever we are using and _write it_.
Sorry if I was unclear. 16:55
fglock hi putter 16:56
putter Though writing the haskell infrastructure to do argument handling et al could also be done. Seems a long way around to do it though
hi fglock
But haskellizing it does have advantages. 16:57
Your call.
fglock Hash is fixed now - what is the url to check the smoke test?
putter Bye all & back ... 16:58
fglock:
m19s28.vlinux.de/cgi-bin/pugs-smokeserv.pl has smokes, but you have to contribute them yourself
unit testing is easier, just ./pugs -B perl5 -Iperl5/PIL-Run/lib6 t/whatever.t 16:59
fglock ok, thanks
putter running smoke-perl5 still requires some hoop jumping, as some tests non-terminate (a couple even with timeout set). some may still exhaust memory unless one externally sets a memory limit. 17:01
fglock np, I'm testing specific things (not fully automated) 17:02
putter ok. notes added to README. 17:08
svnbot6 r7104 | putter++ | PIL-Run/README: added notes on the hoop jumping currently required to run make smoke-perl5.
putter fglock: all set?
fglock yes - most things work :)
putter I'll be back in 48 hrs or so. 17:09
! :)
fglock :(
putter I noticed the most recent test had some regressions. Haven't gotten a chance to look at it. Both the "latest" and the "high water mark" are on the smoke serv. 17:10
fglock are there any higher priority bugs?
putter nothing comes to mind 17:11
fglock ok, I'll create some :) 17:12
putter might try the mm2 switch. stevan said that's becoming plausible, though there may still be gaps/problems. he was starting to explore switchover. 17:13
fglock++ stevan++ gaal++ 17:14
bye all & 17:15
stevan fglock: re: putter's comments on MM2 17:16
fglock hi stevan
stevan I just commited a port of the perl 6 version of Test::Builder,.. that is probably a good place to start looking to see how things have changed
hey fglock :)
stevan is finding it really hard to concentrate on $work today 17:17
fglock :)
gaal oh, crap, just missed putter again 17:18
rehi btw
stevan hey gaal 17:19
gaal stevan :)
fglock tries to convince a debian package to install in a different dir and port
hi gaal
gaal hey fglock! 17:20
gaal fights haskell file path libs
arg, anyone remember how to allow the c preprocessor in .hs source files?
ah, -cpp. how simple :) 17:21
nothingmuch dum dee dum 17:23
gaal time for tea! 17:25
eric256 hola....are there any known issues with building pugs on windows? i mean new issues...while compiling Pugs.Run GHC crashes with an access violoation error 17:28
gaal eric256 - welcome, long time no see - you are running out of ram i think 17:29
fglock I can't build it in Windows - I get a makefile error
gaal are you using the optimized build? which run is it, first or second?
eric256 hmm could be...yea life got busy, moved to a new state then bought a house. ;) takes a bit of time and energy 17:30
gaal fglock: "can't make pugs.exe?"
eric256 unoptimized
gaal unoptimized no longer works for me on win32 when precompiling.
fglock gaal: checking the error message - will take a while to get there
eric256 hmm i can try optimized then
gaal fglock: it's the wrapping issue. create a file with that name, never mind the content 17:31
fglock: when pugs builds use pugs_bin.exe
"pugs" has been made a script but it causes problems. i'm working on it.
nothingmuch nnunley: ping
eric256 thast not a very friends out of memory error for GHC to throw. ;) 17:32
"NMAKE : fatal error U1073: don't know how to make 'pugs.exe'"
i'm cleanings and starting over.
nothingmuch any volunteers for blondie pair programming or coordinated hacking? 17:33
fglock gaal - this is a very slow computer, XP + antivirus + slow disk, leave only 10% CPU available, I think
nothingmuch heh
eric256 blondie? isn't that a band? ;)
fglock eric256 - same message here
nothingmuch eric256: no, it's anAST
eric256 nm: yea i know. ;) 17:34
fglock gaal - create a file with which name?
eric256 hmm. clean build same error.
nothingmuch anyone?
pleease?
eric256 for one has no idea what you are even asking nm 17:35
stevan nothingmuch: what do you want to do?
eric256 fglock i did "cat test > pugs.exe" and now it seems to be building
nothingmuch the todo is long
make a pick
fglock nothingmuch: I'd like to, but I'm on $work now - not enough spare cicles
nothingmuch svn.openfoundry.org/pugs/perl5/Blondie/TODO
eric256: i want help on Blondie 17:36
gaal fglock, sorry, what eric256 said
nothingmuch: i have my hands full
stevan nothingmuch: anything I can do which does not involve re-reading all the compiler theory books I have
fglock ah, ok - you create a new file, and then nmake ignores it 17:37
gaal fglock: yes, clever innit.
nothingmuch stevan: you've already got an advantage over me
if i could get this far you can help too =)
eric256 nm....just curious...but why do you want to build another AST? this whole thing with javascript backends and all that confused me. I'm assuming there is a benefit i am just too dense to crasp un aided
;)
nothingmuch eric256: PIL/Perl6 is too big to be rapidly prototyped on multiple backends
i'm trying to explore less of the what needs to be done to make a backend, and more the how it can be done 17:38
so Blondie is per definition minimalistic (dumb, hence the name blondie)
eric256 but why?
nothingmuch to learn
eric256 ahh... ok
nothingmuch learning with PIL is a bit too hard right now
but once we've learned we can backport
it's essentially a compilation prototype 17:39
eric256 so its just a way for you to sharpen your skills to then attack pugs/pil/parrot/perl6 ?
nothingmuch a bit more than sharpen our skills... we're (i'm) trying to implement two ideas which are not so well explored
one is a circular prelude 17:40
for the benefit of portability
and correctness
and the other is dynamic/static type inferrence
svnbot6 r7105 | nothingmuch++ | s/seemless/seamless/
nothingmuch evalbot should check if system(qw/cmp pugs.new pugs.old/) 17:41
eric256 well i'd help if it wasn't entirely over my head. ;)
nothingmuch eric256: it isn't, I promise
i'll gladly walk you through
i want others involved in addition to myself
eric256 learning through teaching. ;)
freack GHC died agian...how much memory do you need!!! and why did it balloon up to where i can't manage it? 17:42
nothingmuch it needs a lot of memory 17:43
hundreds of MBs
eric256 didn't use to kill my machine though.
nothingmuch Pugs.Run isn't that big though... *puzzled*
eric256 i've got 350mb free when i start nmake 17:45
gaal nothingmuch: it #includes the precomp prelude 17:46
nothingmuch oi
heh
eric256 i'm guessing that is large?
dduncan I've found that GHC runs but slowing the whole machine down under 256M ... runs very well on a 768M machine
fglock meeting & 17:48
eric256 down to 260mb and falling. lol....240mb......230mb
so if this doesn't work i can turn of precompiled prelude and be okay? 17:49
gaal yes, but you really need the precompiled prelude :(
eric256 then i hope i freed enough memory. 17:50
gaal :223
eric256 is there any work that can be done to reduce that?
gaal oops, wrong window
not anything i know how to do :(
eric256 compile perl6 requirments....2gig of memory. ;)
so are the OO changes occuring in pugs? or is it still perl5 based modules? 17:52
gaal well eventually -CPugs output will be binary with DriFT, but we need one of the hardcode labmdaheads for that.
integral is there no possibility to serialise the prelude as perhaps binary, and then recreate the actual tree at runtime, so that GHC isn't trying to optimise all those compiler?
gaal integral: be a hardcode labmdahead! :)
eric256 just dropped below 100mb of memory and ghc "encountered a problem" 17:53
i've got the error report window up...anything in there that can help?
integral hmm, afaik all the DRiFT output stuff is working, it's just that there's nothing in place to load the stuff, right? 17:54
gaal integral: i don't know, it was all done when i was away. DriFT.Binary exists, fwiw, but i don't know if it's lossy or not.
oh no, if we don't use perl5 to wrap pugs then we need to use pugs to wrap perl5 when -BJS and -BPerl5... but "exec" isn't portable 17:56
gaal sighs
eric256 Prelude.pm doesn't seem that big. 17:58
Juerd The scary part is it being monolithic 17:59
I hope that's just a bootstrapping thing
integral perl code is a very efficient way of encoding very big ASTs :-P
Juerd integral: Except it's hard to parse :)
eric256 gives nmake unoptimized one last chance just for kicks 18:00
would it fit in memory better if it were broken into parts? or does the world not work that way?
integral hmm, this looks like it may not be *too* difficult! 18:01
the actual machinery seems to be there in PIL1.hs, and Binary.hs 18:02
eric256 here we go...300mb free when it hit compile Pugs.Run for the second time...might actualy make it 18:03
isn't it odd for that to be an out of memory error when there is still 100mb show as avialable? 18:04
wow...and now 86mb and the same error....must mean its lunch time ;) 18:05
putter whew. ride running late... 18:06
gaal: ping?
any windows perl usersers / CPAN contributers about? is runperl.bat the current state of the art in perl scripts? 18:07
gaal putter! hi :)
putter hi gaal :)
gaal pugs-the-script is a pain - but i can see what pain it prevented :) 18:08
putter my hope was that someone more familiar with windows than I would do the "obvious right thing", whatever that was, to make the pugs script work there. doesnt seem to have happened.
a quick google suggests runperl.bat might be the "state of the art". but I have no first-hand experience. 18:09
integral bother. the precompiled prelude isn't PIL1, it's AST
gaal putter, even if we could use runperl - which i'm not sure is the case - a wrapper would be a huge hit on smokes 18:10
putter oh, btw, I saw a "how much memory does ghc use" go buy. I recently noticed an oddball, where some p6 code (unfortunately not easily isolatable) resulted in a practically unbounded (killed at > 1GB) memory usage. a small (1 line) code change avoided the problem. but this is on an old ghc development snapshot, so the experience is likely not relevant to any current problems. 18:11
gaal: re huge hit, really? that seems surprising. 18:13
Juerd There exists spamming software that uses mail.domain.com to send mail to domain.com, without checking any mx record. 18:14
gaal why? you have 7000 tests, that means 7000 perl5 runtime inits
Juerd Hatefuly.
s/y//
My mx is mx1 - mail I use for customers, who authenticate 18:15
But, of course, mail accepted mail for the domain unauthenticated as well, as it's a standard qmail setup
putter On an optimized build, a 01-sanity test file which _doesnt_ use Test takes about 1sec to run at 1GHz. running perl on the same machine is 2 or 3 orders of magnitude faster.
Juerd Now, I am manually hacking qmail to always deny relying
putter hmm, perhaps I have some aspect of perl swapped in... 18:16
gaal as you're likely to do if you run it 7k times ;-) 18:17
Juerd Fortunately, qmail is very well written and easy to hack.
Pfew.
gaal so you're saying this is noise
putter gaal: 7k times? err, we have <500 test files...?
gaal and that i'm worrying too much about something that doesn't matter?
oh, right, i was confusing tests with test files 18:18
duh
okay, how about...
we write the wrapper in c? :)
i don't know how to fix the who-gets-invoked-first problem 18:19
putter ok, with an alternate version of perl which I'm sure isnt cached, the first run took 1/10 sec. and I imagine the .bat file game will take some time. any idea how long the runperl.bat hook takes? (eg, running a .bat file) 18:21
gaal not sure i even have one 18:22
let me look
putter Its in the perl distrib as of ... something.
apparently
gaal right, lessee 18:23
ugh, how do you time on windows?
putter at 1/10 sec overhead, 1 min on a 1 hr smoke, its clearly not a problem. At 1sec, it would be. Does perl take 1sec to start up on windows? It seems unlikely, but...
actually, does runperl.bat take 1 sec to start perl on windows. there may be path searches in there... 18:24
gaal: re time on windows. I have so no idea. anyone? 18:25
dada I benchmarked something on win32
gaal runperl is a BAT hit, sometimes another CMD hit, and a perl.exe hit -- before the pugs Perl5 wrapper
but i agree that performance may not be the first priority here.
putter perl -MTime::HiRes -e '... system("..." :) 18:26
dada dada.perl.it/shootout/hello.html
putter: I've used some Win32::API trickery
gaal ew, my path contains c:\perl\bin which is where runperl.bat is, but runperl.bat refuses to run! 18:27
dada a simple "hello world" 200 times takes 6.13 secs with Perl, in contrast to 1.10 secs with a C-compiled executable
take it for what it is
(note this is the startup time for perl.exe itself, not for the .bat trick) 18:28
putter is the fs cache wiped between calls? 18:29
dada no, not really
putter gaal: apparently you rename runperl.bat to say pugs.bat which would call the pugs script
dada but you aren't going to wipe the fs cache while you're smoking, no? :-)
putter search.cpan.org/~nwclark/perl-5.8.7...runperl.pl 18:30
true. I was thinking worst case. Perhaps meaninglessly worst.
dada anyway, putter: exec() behaves like _evil_ under win32
putter as long as someone else has incanted the circle of protection, np ;) 18:31
err, and it doesnt leak
dada struggle to avoid it if you can
putter dada: so, is runperl.bat not the right thing? 18:32
dada I don't know exactly what it dows
does
oh, now I know 18:33
and it doesn't look right
I guess "pl2bat pugs" is better 18:35
putter the things I briefly surfed suggested runperl is the new-better thing. no? 18:37
dada not for potentially interactive scripts like pugs 18:38
the fact is that exec()ing messes the I/O of the script with the I/O of the console itself 18:39
let me show you 18:41
pasteling "dada" at 85.35.33.226 pasted "bat things happen!" (48 lines, 1.4K) at sial.org/pbot/13253 18:42
dada after that, I have to kill the command prompt window with the X
putter gaal, dada: can you guys take care of this? either haskellize ./pugs script, or add pl2bat to Makefile.PL? or someone? 18:43
err, data, does your example actually use pl2bat? 18:44
dada putter: no, it uses the pugs script, which does exec()
and my nick is *dada* :-)
putter oops :)
dada so, well, you're doomed anyway 18:45
gaal ...and execing from pugs is not easy too 18:46
is this facade really that important, though?
first of all we can unify jspugs.pl and runjs.pl 18:47
haven't looked at either of them, but i'm sure it should be possible...?
oh, is the importance for alternate smokes? 18:48
too bad :( Also, i'm starting to run out of wakies 18:49
putter gaal: pugs_bin doesnt do much command line argument processing. that would have to be written. and if exec() itself doesnt work, and I cant imagine hs's exec working better than perl's (no?), then getting the js/perl5 interactive repls wont work. 18:51
so if exec() is broken on some (how many?) windowses, then a unified pugs interface, at least for interactive work, isnt going to fly. 18:52
gaal putter: i already wrote much of the cmdline stuff it into pugs
only then i realized there was the exec problem
putter: where is the unified interface used now? do we really need it? was i right that it's to make the smoke test invocations simpler? 18:54
putter It would seem unfortunate for some (how many?) broken windoeses to force the api changes.
gaal well, when the api changed it wasn't tested on windows in the first place :) 18:55
putter I'd suggest waiting until we have someone who understands perl/doze well enough to say "it effects N% of perl windows installs" or some such. If its just that some installs are broken, then we just say "get a non-broken one".
gaal it seems to affect all windows *builds*, already. 18:56
putter exec() doesnt work on _any_ perl windows installs? that seems inconsistent with what I've read.
gaal and (see the "don't know how to make" make error in the backlog)
s/and //
putter I have to go in about 3 min. 18:57
dada putter: I'm not saying that it doesn't work. it just behaves stupidly, and AFAIK it does so on any windoze I've seen
gaal `perl -x pugs -e "say <hi>"` does not work for me, fwiw. 18:58
well, it flashes a terminal
nope, it doesn't work at all: 18:59
`perl -x pugs -e "say <hi>; sleep 1000"` doesn't show anything either.
dada gaal: try this: perl -x pugs -e "say<hi>;" 19:00
putter perlrun.bat manpage says
It relies on the exec() operator, so you will need to make sure that works in your perl.
gaal dada: nope. 19:01
(why would there be any difference?)
putter since the page is only for win, exec() must work on some win().
dada gaal: it works for me
putter I have go in < 3 min
dada C:\dada\pugs>perl -x pugs -e "say<hi>;"
C:\dada\pugs>hi
(sort.. of.. works)
putter: there is a difference between "work" and "work well like it does in Unix" 19:02
gaal putter: i don't know... i have the feeling that this should be reverted, because we don't know how to make it to work right now
and it seriously breaks things.
putter dada: you might try removing the cruft at the top of pugs and replacing it with a simple #!/usr/bin/perl
dada no, it won't help
it doesn't, in fact 19:03
putter but experimentally trying to recover portability info, from two boxes, which is in someones brain, seems... suboptimal
gaal at least in the next 48 hours, and again if we don't figure it out for the release
dada gaal: the difference is the space in the argument 19:04
putter feel free to break PIL-Run. never been really announced. no sign anyone uses. should check with iblech however to make sure he doesnt depend on it now.
gaal dada: neither workforme.
dada that's strange 19:05
gaal putter: sorry :(
dada which windows?
gaal xpsp2, native perl compiled here with vc++2003
dada same windows here, using activeperl
xp pro, no home edition, right? 19:06
gaal oh wait
i'm blind
sorry!
dada hehee
putter wonders "is windows really that broken?", or do we just lack the experience.
dada the "hi" is on the prompt line?
gaal dada: it only works *sometimes* though. there must be a race condition? 19:07
dada no, I get consistently "hi" on the next prompt
gaal ah, no, there it is: you're right, the spaces are what break it
putter eep. bye
&
gaal is too tired to hack, unfortunately 19:08
dada is going home too
eric256 returns from lunch. 19:14
now to remember how to stop precompliing prelude
dada bye 19:15
eric256 considers maybe he should just use a feather accoutn to play with pugs and forget compiling it on windows 19:19
dunno if anyone is still around.....finaly got it to compile by changing my configy.yaml to no precompile prelude...in the past when you then ran pugs it would say "loading prelude" but now it doesn't...should i be concerned? 19:20
masak i also have a late-night question: does $^W still exist in perl6? 19:22
eric256 could setting the ghc_heap_size too large in config.yml make a big difference? is 19:25
gaal eric256: it doesn't load the prelude because it has no prelude. maybe we need to revive the inlined source prelude as an interim measure. 19:27
masak: no, but you'll have 'use warnings' 19:28
eric256: try making do with pugs -MPrelude
eric256 i'm sorry gaal i think i missed something there. what is Perlude.pm then?
gaal eric256: you need to copy it to your @*INC first though
eric256: it's just a file that isn't automatically loaded 19:29
eric256 ohh i think i see...well i changed the GHC heap size, maybe it'll compile now..i had it set at 320mb
gaal it used to be inlined into the pugs image as source code, and evalled at runtime
eric256 didn't it use to load that for you if you didnt' compile it in?
ahh
gotcha.
thanks 19:30
out of curiosity...why did they stop doing that?
gaal i forget, do you do haskell? i think we left some of the mechanism for this in, but commented out: if you want to do it, go ahead
eric256: there used to be no need: precompiling worked for everyone.
eric256 i know no haskel....unfortunatly, and i'm not sure i have the strength to learn it right now. ;) 19:31
gaal and we wanted to avoid bloat.
eric256 gotcha. the heap size change seems to be working...its at least getting farther
gaal Run on the second pass is the critical step.
eric256 yea. i spoke too soon. 19:32
its realy annoying to have a program crash when you are out of memory...seems rude. ;)
gaal well, this is me for tonight. things are messy but i'm not focused enough to fix anything :-( 19:33
see ya tomorrow &
eric256 later
hmmm....are we positive that its a memory issue i'm having? could it be an bad version of GHC or something else? I'm never getting withni 50mb of being out of memory, unless its trying to allocate a huge chucnk just all the sudden that doesn't seem like the right condition for running out of memory 19:52
eric256 's computer looks like a NASA launch controlloer...scrolling text in one window and two sets of graphs constantly changing...jsut makes you feel important. LOL ;) 19:54
geoffb hello, all 19:59
eric256 hey
eric256 crosses his fingers and tries to complie with GHC 6.4.1 20:27
arg 20:30
"ghc.exe: internal error: update_fwd: unknown/strange object 0" 20:59
putter hi. 21:16
fglock hi
svnbot6 r7106 | putter++ | pugs: Attempt to placate windows. Created pugs.bat with perl 5.9's pl2bat. ./pugs now uses system() rather than exec() on windows. Hopefully that will permit interactive ./pugs. This may all be scrapped for a haskell-only implementation.
putter could someone on windows try running "perl -x ./pugs", with either the pugs from r7106, or an older one with exec() changed to system()? 21:17
fglock sorry, I just recompiled - and I've moved the files around in order to make it work 21:18
putter Err, ideally first checking that the version with exec() is _not_ working for you. 21:19
fglock: the key question is whether perl -e 'system("pugs_bin")' works, even when perl -e 'exec("pugs_bin")' doesnt 21:20
hmm.. and I think those two commands are a sufficient test.
fglock testing
eric256 i would but sadly i can't get a compile....and now i'm getting that last error i posted..;(
putter backlogs
fglock putter: both work for me 21:21
putter oh, drat. 21:22
fglock: you used the resulting interactive pugs and everything was fine?
is there anyone online for whom perl -e 'exec("pugs_bin")' doesn't work?
dada. gaal: when you get a chance, please try perl -e 'exec("pugs_bin")' and perl -e 'system("pugs_bin")'. hopefully the first will have the problems you encountered, and the second wont. 21:23
eric256: what cpu and version of ghc? 21:24
fglock putter: my .pl files are associated with the editor - so pugs -BPerl5 opens crude_repl.pl in the editor :) 21:25
putter oh.. well, that's a separate problem... but the normal pugs interactive works?
fglock yes, it does
version 7086 21:26
putter on both exec() and system()? dada/gaal reported seeing what looked like a mix of command shell and pugs output.
eric256 6.4.1 on a Windows XP Pro SP 2 with a Pentium 4 2.4GH and 504mb ram
putter apparently exec() is broken on some windows boxes but not on others. need someone with a broken one to tell me if replacing it with system() works, or whether they are both busted. 21:27
fglock oops - yes, I see it - it looks like it has forked a new process, and both are reading from the windows terminal
cmd.exe + pugs.exe
putter fglock: is that with system() too?
fglock testing again 21:28
putter eric256: has 6.4.1 been released? if you have an old snapshot, especially an unstable one, you might try a new one.
fglock system works, exec has this problem
putter whoot. 21:29
eric256 just downloaded it today
putter ok. so, hypothetically, r7106 should work under windows.
eric256 because 6.4 was giving me the same problem
putter eric256: oh, drat.
eric256 your telling me. lol 21:30
putter fglock: could you try pugs.bat from r7106? 21:31
fglock ok
putter (note that r7106 touched Makefile.PL... just in case that's a problem for you)
eric256: did you check ghc bugzilla? 21:32
svnbot6 r7107 | stevan++ | Perl6::MetaModel 2.0 -
r7107 | stevan++ | * Test::Builder port now takes case of loading paths for @INC
r7107 | stevan++ | - (it makes running the tests easier)
r7107 | stevan++ | * removed calls to resolve() in Shapes example, they are not needed anymore
r7107 | stevan++ | * 32_parameterized_classes.t now calls does() as well as isa() to check types
r7107 | stevan++ | to make chromatic happy ;)
r7107 | stevan++ | * added new 33_BUILD.t test to check attribute handling
r7107 | stevan++ | * added another parameterized classes test to test things like:
r7107 | stevan++ | - Array of Array of Int
eric256 no. i was busy with realy work. lol. keep hoping i'll get just the right combination of settings...but i am about out of hope.. 21:33
putter the ghc folks actively maintain and lookat and fix problems in bug manager. there's a link off the ghc homepage. 21:34
if it's not a known problem, lets help the ghc folks out by reporting it. I seen quick turnaround. and it would be nice to have 6.4.1 robust enough to run pugs without problem, thus escaping from the "to run pugs, ... big involved cruft re ghc ..." to a simple "ghc 6.4.1 required and sufficient". 21:36
eric256 i;ll check it out
fglock putter: the new pugs.bat works - it gives an error message ("No such file or directory at pugs.bat line 53") when pugs_bin terminates 21:37
putter works... as in you can interactively use pugs_bin? 21:38
fglock yes - note that the line number is for pl2bat'ed version
putter oops, doh. I left the post-exec() die outside the if(win){system}else{exec} fixing... 21:39
fglock: ok, hopefully fixed. 21:47
hmm... or... I don't think I really understand the error message. could you try the new version? 21:48
svnbot6 r7108 | putter++ | pugs.bat: corrected bogus error message. fglock++ 21:50
fglock testing
ok - the error is gone :) 21:51
putter so... it's all working... ????!???
fglock let me try with system($0,$bin,...) - this might fix the crude-to-editor problem 21:53
nothingmuch TSa is making me tired 21:54
svnbot6 r7109 | putter++ | pugs.bat: improved error checking/reporting 21:56
fglock putter: I will probably need to use findbin to get 'perl.exe' before crude_repl.pl in the system() call 22:02
nothingmuch MEANADJ-->FURLIKE on the BODYLOCATION
from search.cpan.org/src/MSCHWERN/Bone-E...y/Rules.pm
QUESTION-->do you have a sister who isn't so MEANADJ 22:03
COITUS-->give my FURLIKE dog AILMENT 22:04
AILMENT-->ADJ lovin'
ADJ-->retarded
give my hairy dog retarted loving
putter fglock: from brief and superficial scan of windows perl literature (eg, man perlwin32), it seems the usual thing to do is have .pl (hmm... bleeding on keyboard... where did that come from...?) configured for execution, rather that editing... 22:05
fglock PxPerl asks what you want to do on installation - and I usually edit programs more often than I execute :) 22:06
just adding 'perl' works for me
it shouldn't hurt, since most people have perl in the path 22:07
putter gaal: so using system() rather than exec() may work in the haskell version too.
fglock "pugs.bat -Cperl5" is working fine now 22:08
putter what about pugs -Cperl5 , no .bat 22:09
fglock after "rm pugs.exe" - it works fine
oops - it is not compiling the lines: I get "Unrecognized char \x90 at ...\pugs_bin.exe line 1." 22:11
putter gaal: the currently contemplated "work around foo.pl triggering editor" hoop jumping is the kind of thing I thought might go more easily in pl than hs...
but it has its own disadds. might be worth doing a tradeoff list 22:12
err, what? 22:14
fglock the error is cause by something like "perl pugs_bin.exe" - but I have no idea why this is happening - is it because I "fixed" pugs.bat?
putter oh, you tried just putting in perl to the system call...? no, it also system()'s pugs_bin... 22:15
you would have to do something like unshift(@args,"perl") if $bin =~ /\.pl$/; 22:17
fglock ok
putter when PxPerl asked edit-or-exec, did it give a default? 22:19
putter womders just how common this case is...
fglock I don't remember - but I think it recommended the editor - I find it safer this way (if I want to make it executable, I can make a PAR or bat) 22:20
cool - it works now
(that's because I almost always call perl programs from the command line) 22:21
(and edit using double-click) 22:22
putter hows this 22:26
my(@args)=($bin, split($v, $_));
# PxPerl is said to give a choice, at installation, of the .pl suffix
# triggering either execution, or an editor. This works around that.
unshift(@args,$^X) if $bin =~ /\.pl$/;
# exec() is broken on some windows. Interactive use of pugs_bin gets
# command interpreter output mixed in. So we use system(). 22:27
if( system(@args) == 0 ) {
?
fglock ok (windows only, maybe) - it is nice to comment why you are doing that 22:28
time to leave! bye & 22:29
putter & 22:30
thanks for your help
yet more that wouldnt have gotten done without it 22:31
svnbot6 r7110 | putter++ | pugs.bat: invoke $^X explicitly when calling .pl helpers, to avoid .pl suffix configuration issues. The pugs-as-a-script should now be working correctly on Windows. gaal++ dada++ fglock++ et.al. 22:44
nothingmuch wow
if i was a superhero i'd call myself aqualad 22:45
en.wikipedia.org/wiki/Aqualad
putter nothingmuch: TSa?
nothingmuch tomas sandlass
nothingmuch didn't understand a word from his previous email 22:46
putter wikipedia... Encyclopedia Galactica in the making...
Khisanth the number of people in that group seems to be growing :)
nothingmuch Khisanth: it's not that it went right over my head 22:47
i didn't understand the sentance structures
nothingmuch is probably too tired
Khisanth I already guessed that! :)
pasteling "putter" at 66.30.119.55 pasted "a non-executable quick experiment" (2176 lines, 62.7K) at sial.org/pbot/13260 22:51
putter fglock, stevan: as a quick experimental hack last night, I explored what a "entire pilrun prelude in one file, in source-filtered psuedo-p6" would look like. 22:53
the motivation was 22:54
G2 When's Perl6 out? ;-) 22:56
G2 ducks
putter the right thing would be to use real p6, and a :p5 trait, so one has method foo () :p5 { q! ... ! }. but without pil2, pugs just isnt up to it. so the idea was to source filter psuedo-p6, extracting from it the equivalent of Perl6::Value/Container, PrimP5.pm, and PrimP6.pm.
I was hoping the p6 would be sufficiently less verbose than explicit p5 metamodel that things would be come notably smaller and more manageable. 22:57
It basically worked for the mm parts, but I didnt realize how much pure p5 code was in P6::V/C. It looked like there was some posibility of compressing that aswell, but I didnt try. 22:58
The core idea was to move everthing into one place/(terse)form, where one could see and work with the whole thing. I'm not sure if it's a win or not. 23:00
Regardless, comments would be welcome on the type hierarchy at the top. We'll need to implement that regardless of form. 23:01
any thoughts? 23:02
If one were to pursue the idea for real, the pasted file probably shouldnt be used as a basis. I played a bit fast-and-loose with the conversion. Just wanted to see roughly what it would look like. 23:03
G2: unclear. speed of progress has varied a great deal. in part based on how much of autrijus's time we get. but we weren't able to dig up $3k/mo, so he's now at best part time. at one point I thought something plausible could be running by end of year. now it's less clear - timing is now more a social/economic issue than a technical one. 23:06
s/in part based/in large part based/ 23:07
joao Hi 23:09
putter hello
joao my @array = ("a","b"); $last_index = @array.last; print $last_index; -> this should work, or am I mistaken? 23:10
putter ?eval my @array = ("a","b"); $last_index = @array.last; $last_index 23:11
evalbot_7110 Error: Undeclared variable: "$last_index"
joao sorry
putter ?eval my @array = ("a","b"); my $last_index = @array.last; $last_index
joao I forgot the my :)
evalbot_7110 Error: cannot last() outside a loop
putter ?eval my @array = ("a","b"); my $last_index = @array; $last_index
evalbot_7110 ['a', 'b']
putter ?eval my @array = ("a","b"); my $last_index = +@array; $last_index
evalbot_7110 \2
joao well, I'm reading "Perl 6 and Parrot" and they give the @array.last example
saying that it gives the highest index in an array 23:12
putter, can you explain me, please? :) 23:13
putter Two posibilities - the language continues to evolve, so the book is already significantly out of date; and pugs is not a complete implementation of p6 yet. I'm pretty sure this is an example of the first case.
joao Oh
?eval my @array = ("a","b"); my $size = @array.elems; $size 23:14
evalbot_7110 \2
putter hmm... anyone want to write a "book compatibility" module? method last (Array @self:) { +@self } etc... 23:15
oh, yes, there is that. better.
joao nothingmuch, it's different
elems give the number of elements
putter yes 23:16
joao last should give elems-1
:)
nothingmuch huh?
joao the last index?
putter "Perl 6 and Parrot" apparently speaks of a .last method on arrays, equivalent to p5 $#array
joao yaps 23:17
page 26 O:)
$#array gives the number of elements or the last index?
putter I'm off. Be back in 40 hrs or so. Cheers. Thanks again to gaal, dada, fglock, etal for their help. 23:18
joao: in perl5, yes.
joao putter, thank you 23:19
putter err, yes, last index.
joao yes, I've checked :)
buu joao: Last index.
Note that $#! is illegal!
But @! isn't.
The mind boggles.
joao What other methods do arrays have? 23:20
putter So, I believe the recipe for pugs svn up and build, is... first delete "pugs" and "pugs.exe", and then everything should work. 23:21
bye all &
joao ?eval :1(2) 23:28
evalbot_7110 ('1' => 2)
joao ?eval :1('a')
evalbot_7110 ('1' => 'a')
joao ;)
QtPlatypus ?eval (1=>2) => (3 => 4) 23:32
evalbot_7110 ((1 => 2) => (3 => 4))
joao hmmm :: ((Int,Int),(Int,Int)) 23:33
Pair of pairs?
?eval :1(2) => 1(2)
evalbot_7110 Error: cannot cast from VInt 1 to Pugs.AST.Internals.VCode (VCode)
joao ?eval :1(2) => :1(2)
evalbot_7110 (('1' => 2) => ('1' => 2))
joao :key(value) <- key can't be int? 23:34
QtPlatypus ?eval ('root' => ('left of root' => ('Leftmost leaf') => ("Middle leaf)) => ( ('Right of node')=> ('Middle leaf') => ('Rightmost leaf') ) 23:45
evalbot_7110 Error: unexpected "M" expecting block, "\\", "$!", "$/", "$" or "\""
QtPlatypus ?eval ('root' => ('left of root' => ('Leftmost leaf') => ("Middle leaf")) => ( ('Right of node')=> ('Middle leaf') => ('Rightmost leaf') ) 23:46
evalbot_7110 Error: unexpected end of input expecting term postfix, operator, ">>=><<", "\187=>\171", "=>", "::=", "**=", "xx=", "||=", "&&=", "//=", "^^=", "+&=", "+|=", "+^=", "~&=", "~|=", "~^=", "?|=", "?^=", ":=", "~=", "+=", "-=", "*=", "/=", "%=", "x=", "Y=", "\165=", "|=", "^=", "&=", "=" or ")"
QtPlatypus ?eval ('root' => ('left of root' => ('Leftmost leaf') => ("Middle leaf")) => ( ('Right of node')=> ('Middle leaf') => ('Rightmost leaf') ))
evalbot_7110 ('root' => (('left of root' => ('Leftmost leaf' => 'Middle leaf')) => ('Right of node' => ('Middle leaf' => 'Rightmost leaf'))))
kolibrie I'm having prototyping problems 23:48
I have a subroutine that I can call normally like this:
my $lexer = tokens($input, 'date', rx /(\d\d\d\d)/)
with a prototype like this: 23:49
sub tokens ($input, $label, $pattern, ?$maketoken = sub ($label, $value) { [$label, $value] }) { ... }
and that works great
when I try to call that subroutine from somewhere else, like this: 23:50
$lexer = tokens($lexer, *@$args)
then the subroutine is not found
*** No compatible subroutine found: "&tokens" 23:51
if I add an asterix in the prototype before the first two args
the subroutine is found, but the code ref is an undefined scalar 23:52
any ideas? 23:53