perl6-projects.org/ | nopaste: sial.org/pbot/perl6 | evalbot: 'perl6: say 3;' | irclog: irc.pugscode.org/ | UTF-8 is your friend!
Set by moritz_ on 4 May 2009.
pugs_svn r26687 | lwall++ | [STD] more work on lexically scoped language braids 00:04
r26687 | lwall++ | allow :temp and scope declarators in regex language
r26687 | lwall++ | [gimme5] translate temp to maintain P6 semantics wrt old value
00:08 PhatEddy joined
PhatEddy just to note that the role/enum conflict I tested earlier might be a variant of RT #65022 00:10
rakudo: package A { class Baz{ } }; package P { class Baz{ } }
p6eval rakudo 53f705: OUTPUT«Re-declaration of type Baz at line 1, near "}"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)␤» 00:11
jnthn Oh, they probably all boil down to the same underlying issue. 00:13
PhatEddy since I am doing package tests I will probably associate the last example with the RT and add a comment on the variations 00:14
jnthn Sounds fine. :-) 00:15
00:38 payload left
pmichaud back. 00:38
but only for a bit. dinner. 00:39
00:42 eternaleye joined
jnthn thinks he has a working first cut of WALK. 00:42
00:44 bacek__ joined
PhatEddy rakudo: package A{our $fifty = 42; my $thirty = 30}; say $A::fifty ; say $A::thirty 01:01
p6eval rakudo 53f705: OUTPUT«42␤Use of uninitialized value␤␤» 01:02
s1n PhatEddy: looks like a bug
PhatEddy Am I right in thinking that $A::thirty should have complained harder? Am I over-testing packages?
jnthn How so?
It maybe should have complained harder, not sure. 01:03
It certainly should not have printed 30, which it didn't. :-)
s1n does 'my' always imply private?
jnthn It implies lexical.
$A::thirty is a package lookup.
s1n what scope is 'our'? 01:04
jnthn package 01:05
well, lexical alias to a package variable...
01:12 Kyosuke_Kiryu left 01:14 Kyosuke_Kiryu joined
TimToady note however that STD currently assumes that "my $A::b" is short for "my package A { our $b }", more or less 01:15
without the declarator, $A::b scans outward for an A package 01:18
decommuting &
PhatEddy rakudo: package A { constant $pi = 3.14 }; say $A::pi
p6eval rakudo 53f705: OUTPUT«Use of uninitialized value␤␤»
PhatEddy realizes not sure what that was supposed to do 01:21
dalek kudo: c0480a5 | jnthn++ | src/classes/Object.pir:
list is just a method, and should not get a namespace entry. We need to do this more widely, but this scratches an immediate itch.
 01:23
kudo: 902f999 | jnthn++ | src/parrot/ClassHOW.pir:
Need to lie about top of the Perl 6 object hierarchy.
01:27 wknight8111 left
dalek kudo: b7b3c60 | jnthn++ | src/classes/Routine.pir:
.name method for Routines.
 01:34
kudo: 71c69d0 | jnthn++ | src/setting/Object.pm:
First cut of WALK; doesn't know all of the various orderings, but knows how to handle canonical and super.
jnthn erm, 3:30am. Sleep time. 01:36
jnthn ide spat. Dobru noc. :-)
01:57 cognominal left 02:23 PhatEddy left 02:42 mikehh_ joined 03:20 charsbar_ joined 03:37 lucs left 03:40 orafu joined 04:04 SamB left 04:06 hercynium left 04:07 fridim_ joined 04:15 alester_ joined 04:23 fridim_ left 04:29 skids left, Kisu left 04:41 Kisu joined 05:03 renormalist joined 05:08 renormalist left 05:15 iblechbot joined 05:28 justatheory left 05:31 alester_ left 05:38 jhuni joined 05:42 jhuni left 05:54 spx2 joined 06:15 wollmers joined 06:19 iblechbot left
moritz_ good morning 06:27
06:27 nihiliad left 06:34 ejs joined
patmat good morning moritz_! 06:46
moritz_ there's a game called "world of padman", and I always mis-read "patman" as "padman" on first sight ;-) 07:04
07:10 DemoFreak joined, eMaX left
Matt-W Morning 07:10
moritz_ rakudo: sub foo(){}; say &foo.name 07:15
p6eval rakudo 71c69d: OUTPUT«foo␤»
moritz_ rakudo: module A { sub foo(){}}; say &A::foo.name
p6eval rakudo 71c69d: OUTPUT«foo␤»
pugs_svn r26688 | wollmers++ | [t/spec] add verbosity for diagnosis 07:16
r26689 | azawawi++ | [S:H:P6] version 0.52 released which includes latest STD 07:17
r26689 | azawawi++ | [S:H:P6] updated Moose dependency to 0.77
r26689 | azawawi++ | [S:H:P6] removed dependency on Readonly
07:17 Kyosuke_Kiryu left 07:47 meppl joined
pugs_svn r26690 | wollmers++ | [t/spec] typo 07:57
07:58 ejs left 07:59 AzureStone joined
moritz_ rakudo: say 'a' ~~ /[ a | <fail: "no a"> ]/ 08:08
p6eval rakudo 71c69d: OUTPUT«a␤»
moritz_ rakudo: say 'b' ~~ /[ a | <fail: "no a"> ]/
p6eval rakudo 71c69d: OUTPUT«too many arguments passed (6) - 1 params expected␤current instr.: 'parrot;PGE;Match;fail' pc 2275 (compilers/pge/PGE/Regex.pir:83)␤»
moritz_ so how do I fail from a regex?
rakudo: say 'b' ~~ /[ a | <panic: "no a"> ]/
p6eval rakudo 71c69d: OUTPUT«Unable to find regex 'panic'␤Null PMC access in invoke()␤current instr.: 'parrot;PGE;Grammar;' pc 308 (EVAL_16:138)␤»
08:10 riffraff joined 08:23 payload joined
pugs_svn r26691 | wollmers++ | [t/spec] simplified and unfudged 08:26
08:28 minazo left 08:30 donaldh joined
pugs_svn r26692 | wollmers++ | [t/spec] minor cleanup 08:33
moritz_ wollmers: are you sure that is() autothreads? 08:34
ok $?OS eq $osnames, ...; would be safer 08:35
08:39 masak joined
pugs_svn r26693 | wollmers++ | [t/spec] eq is safer, moritz++ 08:40
masak greetings, pickcamels.
moritz_ h h 08:41
are we known to be that picky? :-)
Matt-W pickcamels?
masak oh, don't overanalyze. just accept it. :P 08:42
08:45 nemstep` joined 08:58 nemstep left 08:59 bacek__ left
moritz_ masak: if we don't overanalyze, we have nothing to say :-) 09:04
masak moritz_: uh, correct. everybody please resume your overanalysis. 09:05
pugs_svn r26694 | masak++ | [S32/Containers] fixed minor typo 09:06
09:10 nemstep` left 09:11 nemstep` joined 09:12 smtms joined
pugs_svn r26695 | wollmers++ | [t/spec] fudge tests for rakudo 09:21
09:27 ZuLuuuuuu joined
pugs_svn r26696 | wollmers++ | [t/spec] fudge tests for rakudo 09:29
moritz_ wollmers: there's no use in fudging tests for rakudo if rakudo doesn't implement the feature being tested 09:32
lunch & 09:33
09:33 donaldh_ joined 09:35 donaldh_ left 09:47 cognominal joined 10:39 aindilis joined
jnthn H H 10:49
Matt-W OH HAI 10:51
10:52 payload left, masak left
jnthn needs to do some stuff for $OTHER_WORK...will get onto Perl 6 things a bit later today. :-) 11:08
11:09 mikehh_ is now known as mikehh 11:20 donaldh left, donaldh joined
pugs_svn r26697 | wollmers++ | [t/spec] added TODO: clarify 11:23
11:32 pancake joined
pancake are private classes declared as the apocalypsis says? 'class :Name' ? does this changed or its just a missfeature of rakudo? 11:35
11:35 Jedai joined
moritz_ don't read the apocalypsis, they are historical documents only 11:36
the synopsis are the specification
I'd guess it's 'my class Name { .. }' these days
std: my class A { }
pancake yep, i know, but im so clicks far from the synopsis and prefered to ask :)
p6eval std 26697: OUTPUT«ok 00:02 35m␤»
pancake ok :) 11:37
my looks better
but didnt seems to work 11:38
pastebin.com/m5571ef12 11:39
buubot pancake: The paste m5571ef12 has been copied to erxz.com/pb/17440
moritz_ then it's NYI
(NYI is "not yet implemented") 11:40
pancake yep, easily deductible acronim :) 11:41
11:43 payload joined 11:48 xinming_ joined 12:04 pancake left
pugs_svn r26698 | wollmers++ | [t/spec] added TODO: 12:06
12:07 wolv is now known as wolverian 12:13 azawawi joined
pugs_svn r26699 | wollmers++ | [t/spec] added TODO: 12:14
12:36 ruoso joined
jnthn std: class Foo:ver<0.0.1> { { 12:37
std: class Foo:ver<0.0.1> { }
ruoso Hello!
p6eval std 26699: OUTPUT«##### PARSE FAILED #####␤Unable to parse blockoid; couldn't find final '}' at /tmp/SCWYpneOmP line 0:␤------> ␤ expecting statement list␤FAILED 00:02 37m␤»
std 26699: OUTPUT«ok 00:02 35m␤»
jnthn hi ruoso
ruoso jnthn, do you have any plans on implementing contextual variables in rakudo? 12:38
jnthn It's on the "things that should probably happen soonish" list. 12:39
azawawi hi
jnthn I think pmichaud wants to get the support for them into PCT too, or will also need them in PGE, so I haven't rushed to get them into Rakudo just yet. 12:40
azawawi ruoso: hi, did u manage to find out why make mildew was croaking on kubuntu...? 12:41
ruoso azawawi, no... I haven't found... 12:42
azawawi ruoso: oh well, im nearly finished with the run mildew/rakudo option on Padre::Plugin::Perl6... 12:43
ruoso azawawi, great... don't be worried about the problem in mildew tho, because we're getting the refactored smop/mildew on shape pretty soon, and that shall fix the problems you're having 12:44
azawawi jnthn: what is the status of PGE syntax highlighting that i previously disabled (or hijacked :)) in Padre Perl6? 12:45
jnthn azawawi: No idea - I think Gabor added it at the Hackathon? 12:46
But not sure it's had any changes (or needed them) on the PGE side since then. 12:47
azawawi jnthn: today im going to finish the STD/PGE syntax highlighting switch... 12:48
jnthn OK, nice.
azawawi hmmm, rakudo parrot build failed on win32 xp... was working yesterday. 12:49
jnthn That's odd - works here for me today. :-S 12:50
details?
oh, and make sure you have the correct Parrot revision..it was bumped up yesterday.
moritz_ azawawi: did you remember to 'make parrot'? 12:51
azawawi make realclean; &
jnthn ruoso: S12 specs a method ^isa() (as opposed to ^isa($possible_parent)) to get the parents of the current class. 12:53
I'm thinking by default that would want to return just the immediate parents?
Perhaps with a flag that makes it get the whole MRO?
Your ClassHOW for s1p is missing this, btw.
ruoso jnthn, a flag doing the opposite
jnthn You think the mro should be the default? 12:54
ruoso for .^methods and .^attributes TimToady agreed to have a :local named argument to fetch declarion local to that type
azawawi jnthn: i ran perl Configure.pl --gen-parrot; mingw32-make failed; but im going to try it again after mingw32-make realclean
jnthn Oh, so the default is "everything"?
ruoso jnthn, yes... because, usually, you don't care how it is defined, you only want to know if it is
jnthn And you pass :local for "just this class"?
OK.
ruoso yes 12:55
jnthn kthnx, will implement.
moritz_ and spec, please
jnthn aye, S12 wants an update.
ruoso about ^isa
I think we have an API inconsistency there...
jnthn yes
ruoso because ^does receives an argument 12:56
jnthn $foo.^isa(Int) # is $foo an Int
$foo.^isa() # get me the list of parents of Foo
Oh, you mean
roles list of roles
ruoso I thought about it too... but it looks weird for it to return different types....
jnthn Probably wants to de 12:57
does
*be
?
ruoso jnthn, yeah... maybe ^roles and ^super
jnthn If they're different multi variants we're fine.
ruoso jnthn, but it looks weird
maybe the parameter is a filter
jnthn Well, not ^super because that implies "only the direct superclasses" 12:58
You could do it as an optional parameter.
ruoso ^isa(Int) return all the types in the hierarchy that matches Int
jnthn I think that we should probably have isa and does being the arguemnt and non-argument form.
*or* have a separate ^parents and ^roles
e.g. name them both matching or both different. 12:59
But not one matching and one not.
ruoso jnthn, true....
what about the filter idea?
jnthn Don't like it.
azawawi jnthn: mingw32-make failure log: sial.org/pbot/36440
jnthn I want ^isa(Foo) to tell me Fast.
ruoso right
I'd be pretty much ok with ^roles and ^parentes 13:00
s/tes/ts/
jnthn Yeah
ruoso do we need TimToady's blessinh?
jnthn And part of me thinks requiring multi-dispatch and optional parameter stuff for something so fundemental and up in Object is probably not great.
ruoso or we just tell him we implemented that way :P 13:01
jnthn For sure, Rakudo doesn't have a bootstrap issue there, but I can appreciate for smop that may not be the case.
ruoso jnthn, and using multis to change the actual meaning of a method is weird... you usually want to customize behavior, not be an entirely different method
jnthn Yeah, that too.
OK, let's go for ^roles and ^parents and stop overloading meanings onto one method name. 13:02
ruoso good
jnthn We could always apply the forgiveness vs permission rule. ;-)
moritz_ that's the general trend anyway
see the reverse -> .flip split
reverse -> reverse|flip|invert actually :-)
jnthn I'll patch S12 and the sepc tests and Rakudo.
13:03 abra joined
ruoso cool 13:03
jnthn and you can patch src-s1p. ;-)
moritz_ hattrick!
13:03 riffraff left
ruoso jnthn, should roles and parents be available in Object as well? or just as meta methods? I'd argue to be just metamethods... 13:04
jnthn I think just as meta-methods. 13:05
Matt-W seems like meta-method thing to me
jnthn .^methods is for sure only in the meta-class.
ruoso: Oh, I wanted to pick your brains on one other thing too.
in smop as far as I understand it you have the metaclass as a "singleton"
wolverian can I specify that a role Foo can be composed only into classes that also compose Bar
s,$,?,
ruoso jnthn, not really singleton, but shared amongst a lot of types 13:06
jnthn OK, but is there a single "instance" of ClassHOW essentially?
ruoso in practical terms, yes
jnthn I appreciate instance is maybe not quite the word in the prototye-OO world.
OK, so that's the direction I'm thinking we'll head in Rakudo too. 13:07
With that in mind though, I'm not too clear on:
ruoso jnthn, yes... ClassHOW is not really an instance
jnthn (from S12)
Alternately, you can associate a class method with the current metaclass instance,
which as a singleton object knows your package, and can function as a more traditional
"class" method:
our $count; method ^count { return $count }
ruoso but conceptually, you could have something that was really an instance
jnthn, I think that part of S12 is just broken ;) 13:08
jnthn OK, we should probably ask TimToady about it.
I can't see how we can have the metaclass both as a "pure prototype" (is that how you call it in smop?) or as we'll have it in Rakudo more of a singleton, and yet still have it know about individual classes... 13:09
Oh, unless the representation API knows the package...
ruoso jnthn, it does
.^!who
jnthn (as in, provides a way to get at it)
ah, OK
ruoso in SMOP, the "class" information is in the protoobject^Wtype object 13:10
the HOW knows how to use it
that's why you have the same ClassHOW being used by most classes
a more class-based OO would have the methods stored in the HOW itself
jnthn Rakudo is heading in a pretty similar direction.
ruoso jnthn, that's very much cool 13:11
jnthn Sure, but I don't see how you can really do that and do the represenation stuff too.
ruoso think of the representation stuff just as a way to get the actual data stored into the object
jnthn Right. 13:12
ruoso in SMOP it is a method call, but there's no reason for you to do that in Parrot
since parrot has a much more defined low-level structure
in SMOP it is a method call, simply because that's the only thing SMOP knows how to do
(call methods)
jnthn It turns out that a subset of the Parrot vtable methods actually match up pretty closely with your representation API. :-)
Not enough that we can really easily unify it. 13:13
ruoso exactly... so you can map the code using .^! directly to vtable calls
jnthn, I'm open to work through getting it unified
jnthn Well, it's not always going to be a direct mapping.
One awkward point for us is your repr API's assumption that instance storage is a HoH. 13:14
ruoso right... that's very Perl6-ish
which is not very parrot-ish
how would parrot need for it to look like? 13:15
jnthn Well, thing is, as I understand it the repr API is meant to allow us to easily interoperate with other object models.
And it seems like too big an assumption in general.
Well, Parrot knows how to get an attribute, set an attribute, and those two can optionally take the class the attribute belongs to, otherwise it will just grab the nearest attribute up the mro. 13:16
ruoso yes... my idea is that we can interchange HOW and REPR as pleased
so you could use a different representation with ClassHOW, as long as that REPR provided all the API ClassHOW requires
i.e.: using a Glib object with ClassHOW
jnthn Right, but does a GLib object really conveniently provide a HoH view or its attribute store? 13:17
*of its...
HoH is not really an efficient object layout.
ruoso jnthn, that's why I'm very open on getting the REPR unified... It wasn't still tested with other REPR implementations 13:18
jnthn OK.
ruoso btw... wasn't the mro supposed to be implemented by the HOW itself?
jnthn Interesting question. 13:19
I mean, ^dispatch and ^can are two places that are affected by this.
And those are fine...
Oh, hmm 13:20
What does $!foo (getting at instance storage) compile down to in smop?
ruoso I mean... I should be able to implement a different MRO by using a custom HOW
jnthn e.g. class Foo { has $!foo; method x { $!foo } } # the $!foo reference in the method
ruoso self.^!instance_storage.{$?PACKAGE}<foo>
jnthn Is attribute lookup meant to be via the metaclass too?
OK, so that goes straight via the repr API? 13:21
ruoso jnthn, S12 makes it very clear that private attributes are not virtual
so... I think it's ok to go through the REPR api directly
13:22 exodist joined
jnthn Sure, Parrot provides for them being, and we'll have to customize that in p6opaque or something. 13:22
Or maybe not even there.
Just in emitting something that specifies the package always.
(We currently are going through what I'd call our repr API today, e.g. the vtable methods, for attribute lookup.)
ruoso yes... in mildew the above line is pretty much the result for $!foo 13:23
jnthn (But I'm pretty sure we ain't always getting the right answers.)
Ok, so we're consistent-ish in smop and Rakudo there already. Great.
13:23 skids joined
jnthn So, I think the best way forward, is I'll dig into getting Rakudo moving in the right kinda direction, and then once I've got our HOW/repr distinction straightened out some, we can look at how we might be able to unify it some. 13:25
azawawi: Looking at your nopaste, I suspect it may be due to some recent Parrot changes in the way dynpmcs related to Parrot. 13:26
13:29 Jedai left 13:30 sri_kraih joined 13:31 masak joined 13:36 PhatEddy joined, smtms left
pugs_svn r26700 | jnthn++ | [spec] Rename ^isa() to get list of parents to ^parents so there's no confusion with .^isa(Foo) and consistency with .^does(Foo) and .^roles(). Also fill out the introspection specification a little more. 13:36
pmichaud jnthn++ ruoso++ 13:37
That use of .isa() to "get list of parents" really bugged me.
jnthn Hey, you had a commit bit too. ;-)
pmichaud I find that my ideas for design improvements are ultimately rejected more often than not. :-| 13:38
at least, it seems that way. :-)
pugs_svn r26701 | wollmers++ | [t/spec] cosmetic cleaning: remove trailing space
jnthn btw, what are we going to call the new Parrot PMC for doing the proto-y object stuff? 13:39
ProtoObject is tempting now Perl 6 ain't using it. ;-)
pmichaud I'll let you decide that. Allison suggested Protoobject.
I'm find with Protoobject, but please don't capitalize the 'O'.
jnthn :-(
pmichaud *fine 13:40
jnthn I liked it much better with the capital O...
pmichaud but the word is consistently "protoobject", not "proto object"
jnthn hmm...true
pmichaud similarly we say "prototype" and not "proto type". I.e., we would say Prototype and not ProtoType 13:41
jnthn *nod*
I guess we can also be 1337
prot00bject!
w00t!
OK, Protoobject it is.
pmichaud When I was actively teaching (and serving as committee member on graduate projects), I often ran into an Over Eagerness to Capitalize Words that seemed Important. 13:42
jnthn And you Really didn't like It? ;-)
pmichaud I continually saw project proposals (and even some final reports) where the writer thought it was important to always capitalize things like "Web Browser" 13:43
as in:
"The user opens the Web Browser and navigates to the start page for the application."
jnthn Ugh.
moritz_ pmichaud: they were all Germans in their heart :-)
masak agrees with pmichaud
moritz_ we capitalize nouns, and it's hard to get rid of that habit
jnthn moritz_: I was thinking about German too. :-)
pmichaud It must have been German then.
jnthn "Mein Hund ist kaput."
masak UnNecessary CaPitaliSation Should Be AvoidEd. 13:44
jnthn ^ capital 'cus it's noun
uh, one too far over :-)
pmichaud But I did see it a lot. It also bugged me a bit that the committee chairpersons (who were supposed to have revised the draft before getting to me) didn't notice or think to correct the author on it first.
anyway, "Protoobject." :-) 13:45
13:45 clkao joined
jnthn rant win :-) 13:46
pmichaud although with prot00bject I can always aim my laser right between the "eyes".
masak :)
pmichaud (The zero in my current font has a dot in the middle, which makes them look like a miniature xeyes application :-)
13:47 ruoso left
pmichaud Ugh, I really don't want to do $otherjob today. 13:47
13:49 hcchien joined
masak don't do it, then! 13:49
pmichaud Well, if I don't, then I might not have $otherjob tomorrow. Plus I'm supposed to travel to $otherjob.location tomorrow to demonstrate the current state of the project, which is likely to be Nil() unless I do $otherjob.work today. 13:51
13:52 H1N1 is now known as H1N1[A]
jnthn Ugh. Un-fun. 13:52
pugs_svn r26702 | ruoso++ | [smop/src-s1p] add ^roles and ^parents... the implementation is yet very naive... thats probably where the mro gets in 13:53
pmichaud and yes, this also means I'm likely to be scarce tomorrow (travel, meetings, travel)
I should be back around on Fri.
masak pmichaud: good luck with $otherjob!
jnthn I expect I'll do Rakudo Day on Friday. 13:55
masak IO::Encoded.ins: "Returns the number of lines or records that have been input. Now with cleaned-up localization usage." what does the latter sentence mean?
also, why must getc have C<Int $chars = 1> as a param? nearly all other methods are content with being called repeatedly... 13:58
$*IN.getc for ^$n;
also, given that it _does_ have this param, isn't the C<Char> return type wrong?
or is the intention to only return the last char read?
moritz_ masak: because IO without proper buffering is painfully slow, perhaps 13:59
13:59 nemstep` is now known as nemstep
moritz_ masak: but yes, the return type looks wrong 13:59
masak moritz_: I question both the need for the param and the return type.
moritz_ masak: feel free :-)
masak in no other language is getc required to read several chars, AFAIK.
let's save experiments in overdesign for the external libraries as far as possible. 14:00
donaldh slightly off-topic question: how is pugs_svn implemented?
moritz_ donaldh: with Bot::BasicBot
masak: right. that's what read() is for :-)
masak moritz_: that, too.
donaldh moritz_: is it listening to commit emails? or triggers, or what?
14:01 ruoso joined
moritz_ donaldh: it uses SVN::Log to talk to the svn server directly... should I put the code somewhere? 14:01
donaldh moritz_: yes that would be great. I've already started tinkering with ilbot since I find it one of the prettiest IRC log bots I've seen. 14:02
moritz_: is it polling the svn server then? 14:03
ruoso jnthn, proto-y?
14:05 smtms joined
moritz_ donaldh: moritz.faui2k3.org/files/svnbot.pl.txt 14:05
donaldh: problem is, currently it polls too often
it should be ever 20s, but I fear it does it more often, actually
jnthn ruoso: ?
ruoso jnthn, sorry... I didn't backlog fully in the last days... what is the proto-y thing you mentioned? 14:06
jnthn ruoso: A PMC that gives us something a bit more prototype-oo based - basically to avoid some of the "faking" our P6object library does today. 14:08
And help get us more to the representation/meta-class separation point.
ruoso I see....
isn't that p6opaque? 14:09
donaldh moritz_: thanks, I'll take a look.
jnthn yes and no
p6opaque will be a subclass of it
ruoso ah... ok
jnthn Which adds some Perl 6 specific bits.
ruoso makes sense... 14:10
jnthn, wouldn't this get you really close to the ResponderInterface concept as in SMOP?
14:11 amoc joined
jnthn In my view, Parrot's never been so far away from it anyway. 14:12
Your ResponderInterface is essentially a list of things you expect a representation to be able to do, so far as I can tell.
Which isn't all that different from Parrot's vtable, apart from Parrot's v-tables provide a more extensive range of things, not just object-y ones. 14:13
ruoso jnthn, except that not all ResponderInterfaces need to implement the REPR api
but yes... I see what you mean
jnthn Not all Parrot PMCs implement all of the vtable either. 14:14
ruoso so the Protoobject is basically a PMC that will be able to implement the REPR api
jnthn Right. It'll basically subclass Object, and let you do things like add_method, etc.
Without having to explicitly deal with the fact that there's also a Class behind the scenes.
ruoso "also a class"? 14:15
jnthn So you'll be able to just deal with one object and not two. And that one object will handle the full repr API.
Right. Remember that Parrot is essentially class-based behind the scenes.
This will just help us ignore that a bit more. :-) 14:16
ruoso right... got that...
so you really could call it ResponderInterface instead ;)
jnthn (While still allowing everything else that thinks in a class-based way to continue inter-operating just fine with us.)
Well, things that go into Parrot should be not specific to Perl 6.
ruoso ResponderInterface is not
it's simply an abstraction that hides how the objects are laid out in low-level 14:17
jnthn I really don't think that name is going to fly. :-)
ruoso ok... not really important...
but it's cool that we're getting the same building blocks at this point
jnthn Yes.
It's nice that smop has come and investigated this area in more detail. 14:18
ruoso it does make me proud of it ;) 14:19
knowing that no matter what, SMOP was already worth the work ;)
jnthn Oh, I think it's for sure that SMOP has been helpful in fleshing out various areas of the spec and working out how to do stuff. 14:21
ruoso now I need to get my secret plan going... which is to make perl 5.12 to run Perl 6
by merging SMOP inside perl5
masak \o/ 14:23
moritz_ well, it's not secret anymore :-)
jnthn That's certainly a "rather you than me" task. :-)
masak ruoso: I've said it before: the moment that happens, smop will turn up on my radar like a frikkin' blimp.
ruoso I think as soon as this refactoring is over, I'll be able to start working on that 14:24
moritz_ how much work is left?
14:24 [particle]1 joined
pugs_svn r26703 | jnthn++ | [t/spec] Update an introspection test to be inline with the spec and correct a few other bits in it, and fudge it for Rakudo. 14:25
14:25 frettled joined
ruoso moritz_, we need to finish porting p6opaque and the low-level HOWs 14:25
and then get mildew passing all the tests it was passing before 14:26
PhatEddy I have been working on package RT tests and believe I have come up with a file of tests for 14 otherwise (afaict) untested package related RT's (19 tests). 14:27
14:27 [particle]1 is now known as [particle]nyc
jnthn PhatEddy++ # thanks! 14:27
14:27 frettled left
PhatEddy Anyone willing to look at the file before I decide what to do with it next. I could put up a gist link ... 14:27
moritz_ jnthn: are you planning to make :ver<1.0> parse in Rakudo? 14:28
PhatEddy: sure
TimToady gah, you guys talk too fast--how am I ever supposed to backlog all that? :)
14:28 frettled joined
PhatEddy link to package tests: gist.github.com/107520 14:29
[particle]nyc my $backlog is lazy;
TimToady is not allowed to be lazy when it comes to Perl 6
[particle]nyc for @need-to-know.lines { ... }
14:29 iblechbot joined
TimToady well, except vicariously lazy 14:29
moritz_ PhatEddy: the first one... shouldn't that stringify to AlsoEmpty() (with the parens)? 14:30
PhatEddy: the third one is wrong... it shouldn't live, it just shouldn't give a Null PMC access 14:31
PhatEddy: same with the next one 14:32
is eval('Simple::B1::a'), 1, 'enum in package' # - I don't see no B1 anywere - should that be just B? 14:33
PhatEddy first one gives invoke() not implemented error now ...
TimToady ruoso: we're going to have to figure out the relationship of p5 and p6 namespaces, and I'm afraid it ain't gonna be pretty...
masak gist.github.com/107541 # only in Perl...
ruoso pmichaud, if I use a PIR closure in the regex, will it be possible to see the Perl 6 lex from that closure?
TimToady, the cool thing in P6 is that the namespace is not global 14:34
so it makes it very easy to solve that
PhatEddy if you think the first one should be AlsoEmpty() I can test for that ...
moritz_ yes
PhatEddy working on third and fourth
TimToady it's just that dispatchers that want to interoperate will often need to look in two places
moritz_ #65022 - you can simply use an eval_lives_ok here
ruoso TimToady, why? if you're in P6, you look in the lexical scope only... 14:35
if you're in p5, you look in the package only
TimToady there are things that have to look in both p5 and p6 lexical scopes interleaved, if we can interleave code
moritz_ same with RT #64204
ruoso sure... but that's when doing lexical lookup... no problem there
masak oh my, pun is still un-updated after the prefix:<=> thing... :/ 14:36
ruoso the dispatcher only have to decide if it does a lexical or package lookup
the P6 dispatcher is lexically-based
TimToady is vaguely reassured
ruoso the P5 dispatcher is package-based
14:36 pmurias joined
jnthn masak: nice! 14:36
masak ♥e Perl 6 14:37
PhatEddy If you are asking to make the third and fourth like the tests for # 64204 I can do that and it would make sense.
masak er, s/e//
moritz_ PhatEddy: yes, I think so
TimToady well, lovee would be the opposite of the lover
and I think Perl 6 loves you too
pmurias jnthn: re HoH a representation doesn't have to store it that way it just needs to expose such an iterface 14:38
lambdabot pmurias: You have 1 new message. '/msg lambdabot @messages' to read it.
TimToady well, at least it ♥s you...
PhatEddy Ok and that B1 should be 'B' ...
pmurias jnthn: the other choice would be to have an Arrayish interface to the data 14:39
amoc also ♥s Perl
ruoso btw... for those wondering... "faz" comes from the verb "fazer" in portuguese, which is "to do".... but when used with the singular third person (and portuguese allows it to have the subject hidden)... 14:40
14:40 SamB joined
PhatEddy moritz_: Well once I fix the warts and it runs, will it be OK to name it basic.t in S10-packages and commit it? 14:40
moritz_ (to me "faz" is a big German newspaper :-)
masak TimToady: I think you're right. :)
moritz_ PhatEddy: yes 14:41
masak for those of you who haven't tried out pun yet, I can do this and you can't: perl6n -e '.say if .chars > 78' lib/Installer.pm
(neener neener)
PhatEddy moritz_: thank you so very much for spotting the problems. I don't really understand karma or I would try to play with it ... :) 14:42
moritz_ PhatEddy: there's not much to understand. If I type PhatEddy++, you get karma. That works for everyone here :-) 14:43
PhatEddy moritz_++ 14:44
masak karma++
14:44 ejs joined
TimToady everyone here++ 14:46
frettled That didn't work, someone needs to file a bug. ;) 14:47
TimToady
.oO(I know a bug who would not like to be filed...)
 14:48
skids
.oO(filing bugs is hard, given the unibody construction)
 14:49
moritz_ @karma here
lambdabot here has a karma of 1
moritz_ well, that worked. 14:50
TimToady everyone here++ 14:51
SamB don't expect lambdabot to remember your karma for long ;-P
TimToady @karma everyone here
lambdabot everyone􏿽xC2 has a karma of 0
TimToady stupid lambdabot
SamB what the HECK???
14:51 azawawi left
TimToady lambdabot is running in Latin-1 mode :( 14:52
SamB ah
you used some wierd UTF-8 space!
TimToady actually, I used a Latin-1 space, but encoded in utf-8
moritz_ the non-breaking space 14:53
SamB well, it doesn't look like it would have worked anyway
TimToady shoulda, if lambdabot were doing Unicode correctly
another good reason for writing our own...
SamB oh, wait ... 14:54
SamB remembers that the second byte of a utf-8 encoded latin-1 char isn't the same as the n
frettled @karma everyone here
lambdabot everyone has a karma of 3
TimToady sometimes it is, and sometimes it isn't the same
14:55 payload left, mberends joined
TimToady @karma ¥ 14:56
lambdabot ¥ has a karma of 0
14:56 nihiliad joined
frettled @karma lambdabot 14:56
lambdabot lambdabot has a karma of -1
TimToady Perl 6++ 14:57
@karma Perl 6
lambdabot Perl 6 has a karma of 1
TimToady now *that* one is a weird Unicode space 14:58
14:58 ejs left
TimToady @karma Perl 6 14:59
lambdabot Perl 6 has a karma of 0
frettled Ah, time for that Oslo.pm meet where Marcus will repeat his G(et)it talk for the benefit of those of us who didn't attend during NPW.
TimToady that's probably the one we *should* be using "NARROW NO BREAK SPACE"
frettled or just regular space :D 15:00
TimToady with a :nobreak adverb
:!break, excuse me
frettled \o/
TimToady must...back...log... 15:02
masak TimToady-eval: say +['foo', 'bar'].comb(/foobar/)
TimToady Nil 15:03
er, 0
masak TimToady: in what way, then, is a list treated as a string?
TimToady [] makes an array, not a list :P
masak er.
TimToady actually, comb should probably work on array too
masak rakudo: say Array ~~ List
p6eval rakudo 71c69d: OUTPUT«1␤»
masak stands arms akimbo 15:04
Liskov says "comb should work on Array too!" 15:05
moritz_ says "that's what we have >>. for"
TimToady well, if comb works for lists at all; is that specced anywhere?
masak TimToady: aye.
S32/Str
moritz_ that's a weird place for that :-) 15:06
masak we talked about it yesterday; I'm making the required additions in Container and IO right now.
moritz_ TimToady: what do you think, should .comb always return a List of Match objects?
or optionally strings?
masak btw, S32/Containters.pod contains a lot of signatures on the form 'our Junction multi method none( @values: )' -- I suppose that's not a valid way to supply a return type in Perl 6? 15:07
moritz_ why not? 15:08
jnthn It's valid.
masak oh. ok.
moritz_ std: our Junction multi method none( @values: ) { }
p6eval std 26703: OUTPUT«ok 00:02 37m␤»
15:08 wolverian left
jnthn rakudo: our Int multi foo() { }; say &foo.returns 15:08
p6eval rakudo 71c69d: OUTPUT«Object()␤»
jnthn rakudo: our Int sub foo() { }; say &foo.returns
masak for a moment, I thought it was someone's Java interference.
p6eval rakudo 71c69d: OUTPUT«Int()␤»
jnthn Ah, for multis heck knows what it returns. :-) 15:09
moritz_ and heck ~~ Object
masak good thing it doesn't blow up, at least.
jnthn rakudo: our Int sub foo() { }; say &foo.candidates[0].returns
p6eval rakudo 71c69d: OUTPUT«Method 'candidates' not found for invocant of class ''␤current instr.: 'parrot;P6metaclass;dispatch' pc 248543 (src/gen_actions.pir:23326)␤»
jnthn rakudo: our Int multi foo() { }; say &foo.candidates[0].returns
p6eval rakudo 71c69d: OUTPUT«Int()␤»
jnthn w00t 15:10
masak jnthn: maybe only subs should also have a .candidates?
jnthn masak: dunno
môže byť
masak TimToady: I'm now blocked on the .comb-on-lists commit. should I just scrap it for now?
I could adapt the comment in Str.pod acoordingly... 15:11
TimToady at the moment the spec seems to assume that strings should be returned by default. I didn't write that part...
moritz_ TimToady: I know, that's why I'm asking
masak does git svn blame
jnthn makes a cuppa and digs back in
moritz_ IMHO it would be sane to assume that Matches are returned by default
and that :str or so would be an optimization 15:12
TimToady well, if it's really cat mediated, then foobar should certainly match across the boundary
and I think Match objects should be the default, not string 15:13
since Match can autopromote to string at need, but not the other way
moritz_ aye 15:14
TimToady though the real sticking point for implementation is cat
if you don't cheat and do it eagerly
dalek kudo: 45fcaee | pmichaud++ | docs/spectest-progress.csv:
spectest-progress.csv update: 380 files, 11013 passing, 0 failing
 15:15
TimToady we've said that StrPos is opaque, but returning a match across an array boundary will certainly stress that
masak TimToady: git-svn claims you wrote the "You may also comb lists and filehandles" part. 15:16
TimToady and then there's the whole mutable/immutable problem, which we also have with returning string results
pmichaud s/you wrote/last edited/ # perhaps
masak pmichaud: nope. wrote.
I went all the way back. 15:17
TimToady svn claims that line is wayland
pmichaud even into S29?
masak aye.
at that time, it was called Functions.pod.
TimToady oh, S29
okay
masak TimToady: you're only on the first layer if you find wayland. :)
TimToady yes, I don't mind that part :)
pmichaud git-svn ftw
masak TimToady: then why did my array example eval to 0?
TimToady I already changed my mind on that 15:18
masak oh, ok.
I missed that.
so, 1, then.
TimToady yes, returns one match object 15:19
which, even if returned as a Capture and autoitemized, still returns 1
but I think comb can stay explicit List return rather than Capture
15:20 donaldh left
TimToady there being little use case for using comb to extract a single value 15:20
15:20 donaldh joined
pugs_svn r26704 | masak++ | [S32/IO] removed $chars param from .getc 15:20
r26704 | masak++ |
r26704 | masak++ | See <irclog.perlgeek.de/perl6/2009-05-06...122346> for the
r26704 | masak++ | discussion about this.
r26705 | masak++ | [S32/Containers] added .comb
r26705 | masak++ | [S32/IO] added .comb
15:25 payload joined
TimToady on the other hand, using comb(/./) will generate an awful lot of Match objects 15:28
looking at the getc discussion, I wonder if there wants to be something that says: whatever this is, return me its characters in order 15:29
and *that* could have a limiter
skids irclog.perlgeek.de/perl6/2009-04-17#i_1071956 <-- previous conversation on return value of comb
TimToady unfortunately, ..chars is taken
I like the "spec fight!" part :) 15:34
well, to argue it the other way for a bit, in most simple cases strings are indistinguishable from match objects except in boolean context 15:35
(and methods)
moritz_ maybe .chars and .bytes etc. just return lazy lists of characters? 15:38
TimToady I'm half inclined to leave it as strings, but change the default to . instead of \S+
moritz_ and they know their lengths, so that they can be used as numbers efficiently
TimToady (comb, that is)
moritz_ (don't know if that's feasible, and if it's a good idea)
TimToady setup of a lazy chars object just to get the length is probably rather heavyweight
moritz_ right
TimToady but if .comb is split(''), then it already applies to most of the things you might want the chars from, including filehandles 15:40
and already has the limiter we removed from getc
we could have a .words that was short for .comb('\S+') if we decided to go that way 15:41
er,
/\S+/
I think I just made a good argument for .words :)
(by accident) 15:42
so I think .comb() is the new split('')
so we can say .comb.mumble.join without arguments 15:43
masak words++
I've actually wanted this at times.
TimToady for words() {...}
dalek kudo: 871c1e3 | jnthn++ | src/ (2 files):
Get .^parents and .^methods a bit more in line with spec; update WALK.
kudo: bd1c1e9 | jnthn++ | t/spectest.data:
Add S12-introspection/meta-class.t.
masak \o/
kudo: 79eb819 | jnthn++ | src/parser/ (2 files):
Get us able to parse module names with colonpairs in (don't do anything with them for now though; this just brings us closer to STD.pm and lets us parse a spectest I want to be able to run).
TimToady like lines() that way
masak exactly. 15:44
TimToady though I wish we weren't overloading plural for two different meanings that way; maybe lazy iterators are okay for the longer units
moritz_ jnthn: you brought us a large step closer to running most of the modules in pugs/ext/ :-) 15:45
TimToady esp if the iterator can optimize when it finds itself in numeric context
jnthn moritz_: By...?
moritz_ jnthn: by the parsing commit
jnthn moritz_: Oh.
:-)
moritz_ jnthn: they all start with module Name:v<0.0.1> or so
jnthn I only did it so I could get away with parsing a spectest. :-)
TimToady: $object.*@candidates(@args); 15:46
And $object.@candidates(@args);
In the second one, IIUC, we call the first one, and if it just returns rather than defers we don't bother with the rest of the list.
With .* what would happen?
Would we call them all anyway?
15:47 Su-Shee joined
Su-Shee Hi. 15:47
TimToady masak: are you editing S32 then?
jnthn And if one of them did defer, so we consumed part of the candidate list, would we then resume calling the next unconsumed thing, rather than re-calling?
masak TimToady: I'll need to backlog and re-read, but sure.
TimToady thanks
masak TimToady: er, does this mean that IO and List will also have .words, in addition to .comb? 15:48
TimToady yes 15:49
masak gotcha.
TimToady and that there's a function words that reads from $*ARGFILES 15:50
or whatever we're calling it today
moritz_ $*MAGICARGS :-) 15:51
jnthn $*ARGH
TimToady std: for lines {...}
p6eval std 26705: OUTPUT«##### PARSE FAILED #####␤Missing block at /tmp/YK1AF1COtS line 0:␤------> ␤ expecting parameterized block␤FAILED 00:02 35m␤»
masak rakudo: for lines { ... }
p6eval rakudo bd1c1e: OUTPUT«No applicable candidates found to dispatch to for 'lines'␤current instr.: '_block14' pc 70 (EVAL_19:52)␤»
TimToady std: for lines {...}; 15:52
p6eval std 26705: OUTPUT«##### PARSE FAILED #####␤Missing block at /tmp/vuSQ7kzZfZ line 1:␤------> for lines {...};␤ expecting any of:␤ blockoid␤ parameterized block␤ standard stopper␤ terminator␤ whitespace␤FAILED 00:02 35m␤»
TimToady hmm, why isn't it...oh, right...
std: for foo {...}
p6eval std 26705: OUTPUT«##### PARSE FAILED #####␤Missing block (eaten by accidental listop?) at /tmp/DN5DFyJCUA line 0:␤------> ␤ expecting parameterized block␤Undeclared routine:␤ foo used at 1 ␤FAILED 00:02 35m␤»
Su-Shee moritz_: it seems your indispensable documentation in your blog has a little link mistake..
TimToady I was looking for 'eaten by accident' but lines is already defined 15:53
so my test didn't catch it
15:53 Psyche^ joined
moritz_ Su-Shee: that happens... which post, which link? 15:53
Su-Shee moritz_: all of the perl 6 basics series linked from "posts in category" in your left-side "menu"
15:54 justatheory joined
Su-Shee moritz_: 03-control-structures.html/03-control-structures.html for example 15:54
moritz_ Su-Shee: I see what you mean... I'll look into that tonight 15:55
Su-Shee: thanks for the notice
15:55 kidd joined
Su-Shee moritz_: thanks for writing it. :) 15:56
masak rakudo: my $x = "OH HAI"; "foo" ~~ / ^ foo {{ $P1 = find_lex␤say $P1␤}} $ / 15:59
p6eval rakudo bd1c1e: OUTPUT«The opcode 'find_lex_p' (find_lex<1>) was not found. Check the type and number of the arguments␤current instr.: 'parrot;PGE;Grammar;' pc 276 (EVAL_18:127)␤»
masak oops.
rakudo: my $x = "OH HAI"; "foo" ~~ / ^ foo {{ $P1 = find_lex "$x"␤say $P1␤}} $ /
p6eval rakudo bd1c1e: OUTPUT«Lexical '$x' not found␤current instr.: 'anon' pc 3 (EVAL_19:3)␤»
masak submits rakudobug
ruoso masak, pmichaud already told me PGE didn't catch up with perl 6 lexicals yet
pmichaud, but I wonder if there's any way to work-around it in the meanwhile... 16:00
masak yes, but did he tell RT? :)
this was supposed to be the workaround.
that's why it's important to complain when it doesn't work.
things such as variable access are half the utility of closures.
16:03 kaether joined
TimToady BLing, I'm fine with .^parents 16:03
I pretend not to care about anything behind .HOW in any case 16:04
ruoso one thing it would be interesting for you to care about HOW would be to define the different roles that a HOW could implement
which would define which OO features it supports 16:05
16:08 cdarroch joined, Psyche^ is now known as Patterner
masak just wrote C<a ? b ! c> in Java, uncomprehending about the syntax errors he got 16:09
TimToady should I show up for the parrot thingy before YAPC, or is that likely to be VM-only-ish?
jnthn Oh, is it defined as before YAPC now?
pmichaud TimToady: I think you could go either way.
The hackathon is post-YAPC.
jnthn Oh, are we talking about YAPC::EU or YAPC::NA?
pmichaud YAPC::NA
jnthn ah.
TimToady I'm talking about NA
pmichaud "the parrot thingy" == "Parrot Virtual Machine Workshop" 16:10
TimToady is that a hackathon?
pmichaud TimToady: I expect it to be more tutorial-like than hackathon-like.
TimToady k
jnthn yeah, as opposed the "the other parrot thingy" == "Parrot Developer summit"
TimToady and the other is after EU?
or NA?
pmichaud particle would be the person to ask about the PVMW at YAPC::NA
TimToady Glo seems to want to know where I'll be when :)
ruoso masak, btw... I think the fix to that bug would be the fix to accessing named closures from the Perl 6 lexical scope itself, which is how it is implemented now.... 16:11
jnthn TimToady: I've heard discussion of something Parrot-ish at YAPC::EU, but don't know before or after or if...
TimToady so there's a hackathon after NA?
pmichaud so, the current schedule (as I understand it): PVWM Jun 20-21, YAPC::NA Jun 22-24, Hackathon Jun 25-26
masak ruoso: ok.
jnthn ruoso: The fix is probably more about setting :outer correctly. 16:12
(I suspect the regex doesn't know its outer lexical scope.)
TimToady ok, Daniel didn't seem to know about the post-hackathon
pmichaud the regex does, its embedded closure doesn't.
jnthn Ah, OK.
ruoso pmichaud, the embedded doesn't do it properly tho
pmichaud the regex does, its embedded closure doesn't. 16:13
ruoso sorry
the regex doesn't
rakudo: sub foo { sub bar { }; return / foo <bar> /;}; "foo " ~~ foo();
p6eval rakudo bd1c1e: OUTPUT«No such attribute '$.pos'␤current instr.: 'parrot;PGE;Grammar;' pc 583 (EVAL_16:233)␤» 16:14
ruoso hmm... let me try to reproduce the specific problem a bit better
pmichaud bar isn't returning anything that PGE understands.
ruoso that wasn't the problem I was trying to reproduce
rakudo: sub foo { my sub bar { }; return / foo <bar> /;}; "foo " ~~ foo(); 16:15
p6eval rakudo bd1c1e: OUTPUT«Unable to find regex 'bar'␤Null PMC access in invoke()␤current instr.: 'parrot;PGE;Grammar;' pc 586 (EVAL_18:234)␤»
ruoso there it is
masak submits rakudobug
jnthn ruoso: Does making bar non-lexical change anything?
ruoso it stops solving my problem ;)
jnthn, but the above error was a sane error 16:16
jnthn, as it wasn't lexically scoped, it was found
(by above I mean the one before the last) 16:17
pmichaud rakudo: sub foo { my sub bar { 1 }; return { bar(); } }; foo();
p6eval rakudo bd1c1e: ( no output )
pmichaud rakudo: sub foo { my sub bar { 1 }; return { bar(); } }; foo().say; 16:18
p6eval rakudo bd1c1e: OUTPUT«_block38␤»
ruoso rakudo: sub foo { my sub bar { 1 }; return { bar(); } }; foo().().say;
p6eval rakudo bd1c1e: OUTPUT«1␤»
masak ah, that one is in RT already. 16:19
arnsholt Hmm. Rakudo doesn't want to build after I updated a few minutes ago. Anyone else? 16:20
ruoso pmichaud, does that mean it is a different problem than you initially thought?
jnthn arnsholt: details?
(I've got Rakudo building fine here... 16:21
)
moritz_ arnsholt: did you reconfigure + 'make parrot'?
arnsholt Says "Class 'Perl6MultiSub' not found", and a stack trace
ruoso pmichaud, as an absolutely misinformed bystander... it seems to be that PGE is taking the package as the namespace instead of the closer lexical scope...
pmichaud ruoso: not precisely
16:21 payload left
pmichaud ruoso: PGE uses parrot's find_name opcode, which first checks the lexical scope, then checks package scope. 16:22
moritz_ pmichaud: iirc you wanted to review the patch to the build system... any progress on that? or do you want me to do a first pass?
arnsholt moritz_: I reconfigured, and I think it does make parrot as part of the top-level rule. At any rate, it says that parrot is up to date now
TimToady jnthn: when does the code in a role block execute?
jnthn TimToady: When you create a particular parameterization of the role.
role Foo[::T] { };
ruoso pmichaud, oh right... so it really seems to be an :outer problem...
moritz_ arnsholt: you could also try a 'make clean' in rakudo
arnsholt Yeah, Trying that now 16:23
jnthn class C1 does Foo[Int] { }; # runs once here
TimToady I wonder if we need a COMPOSE {...} block
jnthn class C2 does Foo[Str] { }; # runs again here
TimToady to check things at compose time
jnthn class C3 does Foo[Int] { }; # doens't run again, we already had a Foo[Int] created
arnsholt Yeah, that worked 16:24
jnthn TimToady: So you can do things like, checking if another role has been composed too?
TimToady earlier wolverian was asking if there's a way to require another role in the composition
jnthn *nod*
TimToady (still bling)
so kind of COMPOSECHECK, really
or POSTCOMP
jnthn Right. post-compose
TimToady implying possibly PRECOMP
jnthn "is there anything about this class independent of what else is composed into it that I care about?" 16:25
Hmm. Perhaps.
I'd really want it post compose I guess because the order of composition is not meant to matter.
TimToady would probably be a mistake to overload PRE/POST
ruoso I'm not sure we need such blockx
16:26 M_o_C joined
TimToady ruoso: as a primitive for building other assertions about the composition, mostly 16:26
ruoso I think that's something that could be delegated to the HOW of the role
we could promote custom HOW instead of hook blocks
jnthn ruoso: Most users aren't going to write HOWs.
TimToady well, presumably most users wouldn't write POSTCOMP either, but something that desugars to it 16:27
ruoso a bit afraid of the huge amount of hook blocks we have now in Perl 6
jnthn True.
masak as long as they don't slow down people who don't use them... 16:28
TimToady waves hands frantically in the direction of the optimizer, then hides behind a tree
masak 休
pmichaud I think the problem is that our regexes aren't getting a capture_lex on them, so they aren't capturing the lexical environment. 16:29
oh, alas, that appears not to be the problem. 16:30
ruoso jnthn, TimToady, one thing I just realized however, is that compose_role is on the side of the target class at this point... 16:31
pmichaud oh, I see the problem. Yes, PGE bug. Its outer handler is properly lexical, but the inner one is not.
ruoso pmichaud, sounds easy to fix when you say it ;P
masak pmichaud++
ruoso jnthn, TimToady, maybe it should be done in the side of the role
TimToady it?
ruoso the role composition
pmichaud rakudo: sub foo { sub bar { }; return token / foo <bar > /; }; "foo " ~~ foo(); 16:32
p6eval rakudo bd1c1e: OUTPUT«Malformed regex definition at line 1, near "/ foo <bar"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)␤»
TimToady compose yourself...
pmichaud rakudo: sub foo { sub bar { }; return token { foo <bar > }; }; "foo " ~~ foo();
p6eval rakudo bd1c1e: OUTPUT«Syntax error at line 1, near "}; \"foo "␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)␤»
ruoso compose yourself into that class
TimToady I think the class and the role should each say "after you" a couple times first
pmichaud rakudo: sub foo { sub bar { }; return token { foo <bar> }; }; "foo " ~~ foo();
TimToady "no I insist"
p6eval rakudo bd1c1e: OUTPUT«No such attribute '$.pos'␤current instr.: '_block35' pc 547 (EVAL_17:220)␤»
pmichaud there.
ruoso what changed 16:33
?
pmichaud token versus regex
ruoso cooo
cool
pmichaud regex has to take a continuation -- that continuation didn't have its outer lexical context set properly
ruoso testing
pmichaud token doesn't have to take a continuation (no backtracking), so there's no inner continuation to have to worry about.
TimToady token can have explicit backtracking 16:34
.*? and .*! 16:35
pmichaud but we don't backtrack into the token
within the token, yes, there's backtracking.
TimToady trew
ruoso pmichaud, .clone is not implemented in Match...
is it?
pmichaud I don't think it is.
moritz_ isn't there some modifier to force backtracking?
ruoso right
pmichaud Match objects are not things that lend themselves to cloning.
ruoso what are the parameters to Match.new?
pmichaud an existing Match object will work. 16:36
Match.new($/)
16:39 kaether_ joined 16:40 sri_kraih left, AzureStone left, Kisu left, kane___ left, DemoFreak left, simcop2387 left, rewt left, pmichaud left, Helios left, allbery_b left, awwaiid left, yahooooo left, revdiablo left, kcwu left, pasteling left, kaether left, jrockway left, literal left, mtve left, hcchien left, eternaleye left, H1N1[A] left, xinming_ left, aindilis left, nemstep left, Tene left, cotto left, amoc left, SamB left, wollmers left, cxreg left, Aisling left 16:41 kaether_ is now known as kaether
PhatEddy rakudo: package A {} ; try { A::nonesuch }; say "hi"; A::nonesuch 16:41
p6eval rakudo bd1c1e: OUTPUT«hi␤Null PMC access in invoke()␤current instr.: '_block14' pc 110 (EVAL_20:66)␤» 16:42
16:42 Eevee joined
PhatEddy should try catch the Null PMC there? 16:42
16:42 Aisling joined 16:43 jnthn joined, literal joined, frew|wor1 joined, hcchien joined, xinming_ joined, aindilis joined, nemstep joined, eternaleye joined, H1N1[A] joined, jan_ joined, hatseflats joined, [particle] joined, japhb joined, buu joined, Maddingue joined, Tene joined, cotto joined, irc.freenode.net sets mode: +oo jnthn Tene
PhatEddy I guess otherwise how would you catch it? 16:43
masak PhatEddy: no. a Null PMC access should never occur in the first place. 16:44
It Is Simply Wrong.
16:45 rafl joined, awwaiid joined, yahooooo joined, revdiablo joined, kcwu joined, pasteling joined
PhatEddy TimToady: I happened to notice that my proposed spec tests sometimes check for ~$! ~~ /:s Null PMC/ which is rakudo specific ... rationalizing the question I guess. 16:46
16:46 SamB joined, wollmers joined, cxreg joined, phenny joined, diakopter joined, Woody2143 joined 16:47 pmichaud joined, cls_bsd joined, DemoFreak joined, simcop2387 joined, rewt joined, avar joined, Helios joined, allbery_b joined, irc.freenode.net sets mode: +o pmichaud, sri_kraih joined, gfldex joined, scook0 joined, breinbaas joined, rhr joined, ascent_ joined, ingy joined, bacek joined, broquaint joined, kane___ joined, Kisu joined, AzureStone joined, dalek joined, agentzh joined, bacek left, rhr left, ingy left, Kisu left, AzureStone left, broquaint left, ascent_ left, kane___ left, AzureStone joined, Kisu joined, kane___ joined, broquaint joined, bacek joined, ingy joined, ascent_ joined, rhr joined, kaether_ joined, amoc joined, Infinoid joined, silug joined, sephee joined 16:48 kaether left, kaether_ is now known as kaether
TimToady
.oO(netsplits are Simply Wrong too)
 16:48
pmichaud spectests should probably not do anything Rakudo specific.
(or any-other-implementation-specific)
TimToady no "probably" about it 16:49
though generally the attempt should indicate the specs aren't clear enough at some point 16:50
so what *should* the generic error "My platform blew up" look like to Perl 6? 16:51
16:51 eternaleye left
jnthn segmentation fault 16:51
;-)
16:51 ALFONSO joined, kaether_ joined
TimToady well, then, is it $! ~~ ESIGSEGV? 16:52
ruoso pmichaud, does the match contain the full string being matched? how do I get it?
TimToady for extra credit, make $! ~~ ESIGKILL work :) 16:53
PhatEddy Please bear in mind someone could do a 'die "lost in space"' from a setting/prelude some day ...
pmichaud ruoso: the match does contain (a reference to) the full string being matched... I'm not sure there's a method to get to it yet, though.
ruoso hmm... it seems it is not copied using Match.new($/) 16:54
pmichaud well, it would take the same original target that $/ has.
TimToady well, culturally we can prohibit string exceptions where we decide they're detrimental, or at least classify them all as X_RANDOM or some such
pmichaud Assuming that $/ is a Match object.
ruoso it is...
but $/.perl doesn't seem to be very usefull 16:55
16:55 hudnix joined
pmichaud I'll add .orig to PGE 16:55
ruoso pmichaud, it currently dies with null pmc at get_string when I try to coerce it to bool
16:55 kaether left, kaether_ is now known as kaether
pmichaud it? 16:56
ruoso rakudo
pmichaud rakudo dies with null pmc at get_string when you try to coerce ?? to bool
ruoso the match...yes...
pmichaud which match?
ruoso pmichaud, I'm pushing the code to github... 16:57
pmichaud okay.
ruoso github.com/ruoso/faz/blob/40aca7f7c...patcher.pm
16:57 IRSeekBot joined
ruoso it dies between 'subrx 4' and 'subrx 4.1' 16:58
pmichaud because $_ isn't set.
at least, that's my guess.
ruoso what should it have?
pmichaud well, wiat.
I misread the rx(...) part.
you're passing the Match object to a regex? 16:59
you need to either send 'pos', or set the Match object's .to()
ruoso in that same call
?
pmichaud (that still doesn't quite explain the null pmc)
17:00 km2 joined
pmichaud creating a new Match object creates a false match unless you do something to turn it into a valid one. 17:00
ruoso but the submatches are actually a pge token directly
pmichaud I'm not quite sure what you're wanting $submatch = $/.new($/); .... $result = rx($submatch); to do 17:01
PhatEddy Just noting that we do string "die" from places in setting as of now ...
pmichaud PhatEddy: "die" is standard Perl 6. There's no problem with that.
ruoso pmichaud, I thought it would be needed to have a new match
pmichaud regexes will create + return a new match
you don't have to create one for the regex to populate.
ruoso right... 17:02
PhatEddy Could someone then explain: (TimToady) well, culturally we can prohibit string exceptions 17:03
TimToady culturally: if you do that, you will be taken out and shot
ruoso cool
pugs_svn r26706 | masak++ | [S32/Str] comb now defaults to /./, use words for /\S+/ 17:04
pmichaud I took 'string exceptions' to mean "keying off of specific strings returned by an exception"
TimToady but more likely it means we need a better way of implicitly typing our exceptions in various scopes
so anything thrown in the compiler defaults to a syntax error type
and anything thrown in the setting defaults to some other exception type
17:05 mizioumt joined
TimToady alternately, it's not default types, but mixins of who has seen this exception go past 17:05
[particle]nyc and anything thrown in a string defaults to a string exception type? same for regex, expression, and other dsl's?
TimToady so an exception gains a pedigree as it passes up the stack
[particle]nyc sounds like swine flu
bird-pig-human
17:06 masak left, ALFONSO left
ruoso interesting... 17:07
TimToady but probably in general we need to have helper routines that die with an appropriately typed exception, panic, parsefail, settingfail, whatever
[particle]nyc so !!! in a setting gives a different typed exception than in a rule 17:08
TimToady or we go through at some point and change every die "foo" to die new XYZ: "foo"
[particle]nyc brr
PhatEddy there aren't that many in the rakudo setting right now I think
[particle]nyc die XYZ("foo") would do it, too 17:09
pmichaud ...and many places that are currently "die" may need to become "fail".
TimToady I think perhaps, since any type defines a coercion, we can just say XYZ("foo")
usually fail XYZ("foo")
17:10 azawawi joined
[particle]nyc XYZ("foo").fail ? 17:10
ruoso pmichaud, hmm.... now the program just go nowhere... it says "before" but not "after" nor "fail".... I pushed the last version to github...
[particle]nyc does invoking an exception throw it?
dalek kudo: 2e2f74b | jnthn++ | src/pmc/p6invocation.pmc:
P6Invocation should act more like a mutable iterator; we'll only clone it if we have to.
kudo: cab8784 | pmichaud++ | build/PARROT_REVISION:
Bump PARROT_REVISION so we get fixes to PCT and .orig in PGE.
moritz_ adn that would automatically gather information about the source location etc?
dalek kudo: c38a2ea | pmichaud++ | :
Merge branch 'master' of [email@hidden.address]
TimToady certainly not
at least, not if you mean invoking the type object 17:11
it's currently undefined on an actual exception object
17:11 japhb left
TimToady afaik 17:11
and I think it's bad huffman to use .() for control flow 17:12
even .fail is a bit bogus, from the standpoint of keeping the control flow out front where it can be easily seen 17:13
[particle]nyc i like your reasoning on that.
17:13 mico joined
pmichaud what does 'return' mean in this context? do we 'return' from pointy subs? 17:13
ruoso pmichaud, ah... sorry... 17:14
TimToady and I still think that if we want to encourage people to avoid string exceptions, we should huffmanize exception types as much as possible
[particle]nyc you LEAVE them iiuc
TimToady Exception:: is completely bogus in that regard
and I've long been asking for something like X:: instead
or even XFOO to go with EFOO 17:15
pmichaud TimToady: I'd be very happy if you defined some basic exception types for us implementors to use as a model.
or gave us a model to work from.
[particle]nyc the X! sigil/twigil pair :)
TimToady OH NOES 17:16
Tene that would be really great.
TimToady X-Files
[particle]nyc X-Org
TimToady an exceptional organization, to be sure 17:17
skids Exceptions seem to be the exception to huffmanization. IDoNotKnowHowToHuffmanizeException. 17:18
[particle]nyc X'Cept
TimToady X'I-DIDN'T-LIKE-IT
I actually kinda like X' as a productive prefix unlikely to run into other uses of X- 17:20
pmichaud although it does look like a metaop.
X~ X+ X-cetrera.
TimToady I expect the rest of the word to be in Klingon
pmichaud *-cetera
Tene TimToady: if you could spec out some core exception classes or roles, I'd be very pleased.
TimToady X'GBOR'TH!!! 17:21
Tene klingon doesn't use X
TimToady does now :)
pmichaud Gesundheith.
Grrr.
Can't type today.
(or any other day lately)
TimToady I can'th thype eithher.
skids ><
pmichaud pbthbhthth
see also: Ack Phhth 17:22
PhatEddy I don't know if I followed this well enough to do a detailed RT ticket - but I think I could open a somewhat vague one on 'typed exceptions for prelude'
TimToady probably not worth a ticket till specced
pmichaud Perhaps what we need is to find the current instances of die/fail and decide what they should be.
then from that we can start to understand the exception types we want. 17:23
TimToady what, build the sidewalks where there are already paths in the grass?
pmichaud Oh, right. Never mind, that will never work. :-)
moritz_ sounds not perl6ish enough
azawawi pmichaud: im trying to build rakudo on win32 and it keeps failing. sial.org/pbot/36440. Any ideas how to fix it?
pmichaud we'll probably violate some language zoning ordinance or something.
TimToady but, but, but, think of all the prior art we could mangle into unrecognizability 17:24
pmichaud azawawi: unfortunately, I don't have win32 to be able to test much (more)
azawawi tries it on feather1... 17:25
TimToady I do kinda like the notion of exception types as mixins though
pmichaud azawawi: that said, I know that some recent Parrot commits removed PARROT_EXPORT from a number of functions, and it's possible that doing so has caused dynpmcs to stop working in certain Windows compilers. 17:26
TimToady especially if an exception handler can mix in a new type and rethrow
pmichaud well, the Failure role seems to me as though it really wants to be used as a mixin
TimToady that'd be F'd now... 17:27
pmichaud for example, assuming that infix:</> should return a Failure instead of immediately dying... what should that look like in Perl 6? 17:28
TimToady return Num but F'd; :)
pmichaud return Num but F'd("Divide by zero"); # ?
TimToady well, s/return/fail
jnthn azawawi: It works under the MS VC++ compiler; it may be something specific to mingw
pmichaud and is it "return" or.... right
jnthn F'd? 17:29
oh, Failed...
17:29 mizioumt left
jnthn not...right... :-) 17:29
pmichaud And X'd can be "X-rated" :-)
jnthn
.oO( Failed wasn't the first word that came to mind )
TimToady to the pure all things are pure :P
pmichaud well, Perl 6 can certainly be colorful. We're moving out of simply using punctuation for swearing to actually having it in the identifiers. :-) 17:30
PerlJam greetings.
pmichaud it seems odd to me that we would do fail Num but F'd(...); though. Why two "fail"s?
TimToady my F'ing $F'er is F'd; 17:31
azawawi jnthn: I am going to investigate it then. :)
pmichaud azawawi: did this just start showing up?
17:32 mico left
azawawi pmichaud: after today's parrot svn update... 17:32
ruoso pmichaud, cool... it parses already... now I'm struggling to get the captures...
[particle]nyc how do you access the return type of a sub inside the sub? 17:33
17:33 M_o_C left
PhatEddy azawawi: I am getting the same think with mingw and Activestate - am working with the latest pull on cygwin ... 17:33
TimToady fail Num: "divide by zero"; # maybe
jnthn std: ::F'ing := Failure; my F'ing $F'er is F'd; 17:34
p6eval std 26706: OUTPUT«##### PARSE FAILED #####␤In "my" declaration, typename F'ing must be predeclared (or marked as declarative with :: prefix) at /tmp/QEM2pH3BOX line 1:␤------> ::F'ing := Failure; my F'ing $F'er is F'd;␤FAILED 00:02 35m␤»
[particle]nyc seems you'd want to "return {whatever-type-i-should-be-returning} but F'd"
...frequently, so it may deserve a shortcut
azawawi PhatEddy: strawberry + mingw here
moritz_ std: ::F'ing ::= Failure; my F'ing $F'er is F'd;
p6eval std 26706: OUTPUT«##### PARSE FAILED #####␤In "my" declaration, typename F'ing must be predeclared (or marked as declarative with :: prefix) at /tmp/gakuGGANGC line 1:␤------> ::F'ing ::= Failure; my F'ing $F'er is F'd;␤FAILED 00:02 35m␤»
TimToady std: subset F'ing of Failure; my F'ing $F'er is F'd; 17:35
p6eval std 26706: OUTPUT«ok 00:02 36m␤»
pmichaud That's just F'd.
[particle]nyc Mo'Fo, the new exception-based web framework for perl 6 17:36
pmichaud we might be careful with that -- Morrison and Forrestor might come after us.
*Forrester (grrr!)
jnthn Or the Mozilla Foundation. ;-)
17:36 kaether left
dalek kudo: ee7d456 | jnthn++ | (2 files):
P6Invocation should become more Iterator-ish.
 17:36
kudo: 663cde0 | jnthn++ | src/ (2 files):
Refactor indirect method dispatch, and make $x.@foo form do something along the lines of what we'll eventually want it to.
[particle]nyc my $typing is F'd; 17:37
skids E'pic F'
moritz_ skids: that's a parse error ;-)
TimToady std: E'pic F'
p6eval std 26706: OUTPUT«##### PARSE FAILED #####␤Syntax error (two terms in a row?) at /tmp/2H2iX1Azy8 line 1:␤------> E'pic F'␤ expecting any of:␤ POST␤ argument list␤ infix or meta-infix␤ infix stopper␤ postfix␤ postfix_prefix_meta_operator␤ standard
..stopper␤ statement modifier loop␤ ...
ruoso continues to pester pmichaud... 17:38
pmichaud, I've just pushed a last version... 17:39
it already parses...
but i'm failing to get the captures...
pmichaud azawawi/others: I suspect that the point where PARROT_REVISION was bumped in 9a1cacdc is the beginning of compilation problems under Win32 (more)
TimToady sigh, *still* backlogging... 17:40
[particle]nyc pmichaud: i agree. i think jonathan F'd up there.
17:40 literal left
TimToady I believe in victorian capitalization, Gentle Reader 17:40
pmichaud my suggestion is to get a known version of rakudo before that (I suggest 4d7fe560), but try building it with the latest Parrot.
I suspect you'll get the same error.
17:40 literal joined
TimToady fouled up? 17:41
pmichaud "failed up"
[particle]nyc there's trouble with the parrot makefile when using nmake
moritz_ followed up?
17:41 literal left
jnthn What most likely happened is that the Parrot change happened, but we didn't bump our Parrot revision for a couple of days later. 17:41
[particle]nyc ooh, "failing up" sounds like a good self-help book title
jnthn I suspect it's whoever did the dynpmc changes that really failed up. 17:42
TimToady What The Failure?
pmichaud correct.
I also suspect there aren't any direct "dynpmc" changes.
[particle]nyc cotto did dynpmc, i believe they were tested on strawberry
i tested fine on msvc 17:43
pmichaud [particle]: I think those changes are in a branch, ye?
*yes ?
[particle]nyc no, merged to trunk
jnthn [particle]nyc: It seems it's just mingw
[particle]nyc maybe that affects rakudo, if it wasn't modified to match something new?
strawberry uses mingw, no?
TimToady std: sub infix:<the> {...}; *.WHAT the *
pmichaud [particle]: I think PARROT_EXPORT has been removed recently.
p6eval std 26706: OUTPUT«ok 00:03 36m␤»
pmichaud (from a variety of functions) 17:44
jnthn [particle]nyc: I saw fperrad asking about the patch in #parrot recently IIRC, and he builds on mingw
(IIRC)
azawawi [particle]nyc: yes, strawberry+mingw
17:44 literal joined 17:45 kaether joined
pmichaud does Parrot's "make test" do any testing of dynpmc building and linking? 17:45
[particle]nyc pmichaud: it was removed from pmc2c
so vtable functions are no longer exported in pmc or dynpmc.
jnthn They musta done a bit more than just rip that out.
Otherwise it wouldn't work on MS VC++.
pmichaud ISTR that one of the reasons we had to leave PARROT_EXPORT in was because otherwise things wouldn't link properly on certain Win32 compilers 17:46
jnthn I don't know enough about mingw to know why it'd break that and not MS VC++ though.
[particle]nyc there's runtime resolution of which vtable functions are needed, rather than compile time
jnthn pmichaud: I believe they look up the vtable indirectly now.
[particle]nyc ...when building pmc's.
pmichaud jnthn: okay.
jnthn pmichaud: Which would be a legit change, IMO.
pmichaud Sure, I'm fine with that.
skids X'pt, F'd, E'vt, C'tl, S'gl, N'n <-- just trying stuff on 17:47
jnthn MS VC++ was certainly one of the compilers that did need PARROT_EXPORT before.
pmichaud azawawi: just to verify, you've done "make realclean" in both the rakudo and parrot trees and then built from that?
azawawi: or even better, start from a completely fresh checkout...?
azawawi pmichaud: yup i did that...
[particle]nyc ...with no installed parrot 17:48
azawawi pmichaud: im going to pull it out again on another directory
[particle]nyc you're failing during link time, so it could be trying to link with an old libparrot
pmichaud in particular, the link failures all appear (to me) to be the vtable functions. 17:49
[particle]nyc yes
TimToady maybe we should reserve X' for builtin types for X in A..Z 17:50
pmichaud +1
skids thinks maybe spec-reserving idents { <[A-Z]> ' <ident> } ...
[particle]nyc X'kcd
why not <upper>?
or whatever the real unicode type is 17:51
s/type/class/
TimToady Your code is Σ'd
pmichaud I wonder if S'task, T'Pring, or any of those characters might complain, though. 17:52
skids Sorry A..Z or is that S'ry?
[particle]nyc I'm U'r M'n 17:54
17:54 azawawi left 17:55 azawawi joined, azawawi left
TimToady I'd think O'Reilly would really like Perl 6, except for the fact that we reserved his name... 17:56
revenge for the camel :)
pmichaud Otoh, at least O'Reilly is a valid identifier in Perl 6. That's more than can be said for many languages. 17:57
skids O'right common usage F'
pmichaud and we could reserve the O' section of the namespace for "really good documentation"
TimToady std: F*d
p6eval std 26706: OUTPUT«Undeclared name:␤ F used at 1 ␤Undeclared routine:␤ d used at 1 ␤ok 00:02 35m␤»
17:58 scook0_ joined
pmichaud ruoso: you're having difficulty getting the captures? 17:59
17:59 azawawi joined, Aisling left
pmichaud It looks to me as though they should simply be in $<subrx> 17:59
ruoso pmichaud, I got a little further with that...
now I need to be able to use a match as a capture
17:59 Aisling joined
pmichaud "use a match as a capture..."? 17:59
matches are already captures. 18:00
18:00 scook0 left
ruoso when I try to do self.run-action($<subrx>.ast, |$<subrx><rx>) 18:00
it dies
moritz_ how so? 18:01
pmichaud rakudo has no clue how to do |$<subrx><rx>, afaik
ruoso ah... right...
pmichaud I know rakudo can do |@array and |%hash... but I think that's it for now.
ruoso ok..
so if I
my %a = $<subrx><rx>; my @a = $<subrx><rx>; foo(|@a, |%a); 18:02
would that work?
18:02 japhb joined
pmichaud it might -- I'm not sure what will happen when $<subrx><rx> gets copied into %a. 18:02
my %a := ... might work in that case.
moritz_ you'll need %(...) and @(...) before the assignment
(presumably)
pmichaud and yes, what moritz++ said.
time for me to grab some lunch, and then really focus on $otherjob for a while. 18:04
18:16 abra left
PhatEddy Given our discussion on catching the Null PMCs, might eval_perl6_lives_ok and eval_platform_lives_ok (no dies) be plausible additions to Test.pm. 18:19
We could fake with string pattern matching for now ... 18:20
(sorry about the m'ss'ng p'nct'ti'n)
azawawi pmichaud, jnthn: ok, rakudo worked now on a fresh directory. must be something missing in realclean/clean. thx
pugs_svn r26707 | lwall++ | slight braino 18:21
18:21 LadyLunacy joined 18:22 mizioumt joined
PerlJam TimToady: heh, I was just about to make that change. 18:27
szabgab azawawi, fixed some Padre code in your Plugin 18:28
azawawi szabgab: hi there :) 18:32
szabgab yeah , hi me :-)
when are we going to have a perl 6 debugger in Padre ? 18:33
ruoso rakudo: sub bar { }; sub foo (*@_, *%_) { bar(|@_, |%_) }; foo(); 18:34
moritz_ probably two days after we have the first Perl 6 debugger at all
p6eval rakudo ee7d45: ( no output )
18:35 ZuLuuuuuu left
azawawi pmichaud: after investigating the win32 build problem, it seems that neither clean/realclean delete generated files in rakudo pmc/ 18:39
18:39 mizioumt left
azawawi pmichaud: after removing them; rakudo win32 build worked again. 18:40
18:43 kaether left 18:46 mtve joined, desertm4x joined 18:49 amoc left
ruoso considerign I have a token like token { <a> <b> } 18:49
is there a way for me to expand the captures of <b> as if it was inlined, instead of being a subtoken? 18:50
PhatEddy I counted about 40 perl6 rts, either open or new, with Null PMC in the title. find/grep through t/spec turned up only one of the rt#s ...
moritz_ ruoso: $/ = $<b>
or maybe $/ := $<b> 18:51
ruoso but that would override <a> wouldn't it?
(and rakudo doesn't like it, anyway)
moritz_ ah, now I know what you mean
don't think you can 18:52
ruoso hmm...
moritz_ (though I've wished for that functionality as well)
<|b>
ruoso maybe I could change the match in a closure after that... 18:53
moritz_ matches are immutable, notionally 18:54
18:54 pmurias left 18:58 Kisu left
pmichaud azawawi: perhaps you could do a "make realclean" and let us know which files are getting left behind...? 18:59
azawawi pmichaud: sure... but im new to git; how do i get a list of ignored files? 19:02
pmichaud from a built copy of rakudo, just do 'make realclean' and then let us know which files are in the directory. 19:05
There are only six that are part of the repository.
mutablevar.pmc p6invocation.pmc perl6scalar.pmc
objectref_pmc.template perl6multisub.pmc perl6str.pmc
so any files outside of those six ought to be removed by 'make realclean'
Infinoid That realclean issue sounds like rt.perl.org/rt3/Public/Bug/Display.html?id=65006 19:07
Does the patch attached to that ticket help?
19:08 r0bby joined
ruoso does rakudo support slurpy named arguments? 19:08
pmichaud yes.
(slurpy named arguments: yes.)
(patch attached to ticket: probably yes.)
I'll apply the patch in 65006, just a sec. 19:09
Infinoid LylePerl++ for providing the patch
I had initially balked at how some lines had double-backslashes but others don't, but it turns out it's only needed for wildcards
pmichaud same here. 19:10
azawawi pmichaud: it seems that it is the issue im talking about. i am seeing *.o all over rakudo src/...
pmichaud azawawi: agreed.
ruoso pmichaud, what about mixing explicit named args with |%named?
pmichaud ruoso: I'm pretty sure that works.
ruoso: it might not work if there are duplicate named args, though.
19:11 mizioumt joined
ruoso not the case... althought still not working... 19:11
moritz_ rakudo: sub a(:$b){say $x }; my %h = b => 5; say a(|%h)
p6eval rakudo ee7d45: OUTPUT«Scope not found for PAST::Var '$x' in a␤current instr.: 'parrot;PCT;HLLCompiler;panic' pc 146 (src/PCT/HLLCompiler.pir:105)␤»
moritz_ rakudo: sub a(:$b){say $b }; my %h = b => 5; say a(|%h)
p6eval rakudo ee7d45: OUTPUT«5␤1␤»
pmichaud looks like it works :)
ruoso rakudo: sub a(*%_) { say %_<a> }; my %h = a => 5; say a(|%h); 19:12
p6eval rakudo ee7d45: OUTPUT«5␤1␤»
ruoso rakudo: sub a(*@_, *%_) { say %_<a>; say @_[2]; }; my @a = 1,2,3; my %h = a => 5; say a(|@a, |%h);
p6eval rakudo ee7d45: OUTPUT«5␤3␤1␤» 19:13
ruoso hmm....
rakudo: class A { method a(*@_, *%_) { say %_<a>; say @_[2]; } }; my $o = A.new; my @a = 1,2,3; my %h = a => 5; $o.a(|@a, |%h); 19:15
p6eval rakudo ee7d45: OUTPUT«5␤3␤»
ruoso rakudo: class A { method a($b, *@_, *%_) { say %_<a>; say @_[2]; } }; my $o = A.new; my @a = 1,2,3; my %h = a => 5; $o.a(9, |@a, |%h); 19:16
p6eval rakudo ee7d45: OUTPUT«5␤3␤»
ruoso I don't get it... it looks very much like this in the code I'm trying...
azawawi ruoso: hi
ruoso pushing to github
azawawi, hi
azawawi ruoso: is development on mildew on github or something since im not seeing any activity here in pugs repo... 19:17
ruoso azawawi, there has been a stall this week... but it's still in the pugs repo
pmichaud, github.com/ruoso/faz/blob/14156c09b...patcher.pm
about line 66 19:18
azawawi ruoso: cool :)
19:18 nihiliad left
pmichaud maybe try it with a binding for %named and @pos (instead of assignment) ? 19:18
jnthn takes a break from the -Opain task of refactoring p6object
ruoso pmichaud, changing to assignment was already me trying something else ;) 19:19
most importantly...
the hash is there
in the say before the call
but is empty in the say inside the method
19:20 donaldh left, donaldh joined
pmichaud did you try defining run-action without the explicit *%_ ? 19:20
(since it gets one implicitly) 19:21
ruoso will try now
no deal
pmichaud are the positionals appearing as expected?
and just the named arguments missing?
ruoso the @pos array is empty
both sides
but the explicit positional arg is there 19:22
let me add a positional capture in the outermost token
positionals failing too 19:24
I think the problem is with the slurpy 19:25
getting the expected items in the run-action signature gets the value
both named and positional... 19:26
pmichaud checking... I think I might know what the problem is.
ruoso hah
got it
if I use the slurpy with a name different than @_ and %_ it works 19:27
pmichaud I think that rakudo currently thinks that any block with @_ and %_ gets those as arguments.
which is wrong -- it should be only blocks that otherwise have placeholder vars. 19:29
19:30 desertm4x left
pmichaud yes, looks like rakudo puts @_ or %_ as parameters to any block. 19:30
(in which they are used)
TimToady: is it only blocks containing placeholders that get the implicit @_/%_ when used? 19:32
for example:
sub foo($x) { say @_.perl; } # does sub foo get a *@_ parameter?
clearly: sub foo { say $^a.perl, @_.perl; } # sub foo($a, *@_) { ... } 19:33
ruoso pmichaud, yes... this issue was raised before once... 19:34
hmmm... just got into an unexpected issue... 19:36
my action class has an "end" method
and in order to allow customization via composition 19:37
I call it by doing $action.*end
it turns out that Any seems to have an "end" method as well (apparently from the list api)
so it fails as it tries to call end there as well
baest pmichaud: hello, in the tests in parrot/t/compilers/pge/perl6regex/rx*, is it possible to test for exceptions, like parse errors (thrown exceptions) 19:38
pmichaud baest: yes -- see rx_syntax for some examples
ruoso why does Any has a end method?
pmichaud we look for a pattern instead of 'y' or 'n'
baest pmichaud: cool, thanks
pmichaud ruoso: are you asking why it's defined in S32? 19:39
ruoso pmichaud, no... I'm asking why rakudo implements it 19:40
pmichaud ...because it's specced in S32. :-)
ruoso is it?
where?
19:40 desertm4x joined
pmichaud item end 19:40
our Any method end (@array: ) is export
Returns the final subscript of the first dimension; for a one-dimensional
array this simply the index of the final element.
ruoso er... that is in the Array section... 19:41
pmichaud (Containers.pod)
so is .elems, which we know is defined on Any.
19:41 mizioumt1 joined
pmichaud as are .keys, .values, .kv, .pairs, etc. 19:41
ruoso I strongly feel we should rethink that 19:42
pmichaud, but really... in Containers.pod it mention as part of the Array API, not Any
pmichaud Yes, but we already know that some things mentioned as part of Array/List/Str/Num are really defined on Any. 19:43
so Containers.pod is known to be wrong in a few areas already. 19:44
ruoso pardon me for my memory... but why do we know that?
moritz_ ruoso: you can't rely on type information in S29 and S32*
pmichaud Because TimToady has declared it so
ruoso TimToady, around?
pmichaud specific known cases are .grep, .elems, and .join
1.join(' ') # "1" 19:45
19:45 goksie joined
pmichaud 3.elems # 1 19:45
ruoso btw... Faz works nwo
baest pmichaud: I'll send the quote word literal patch for pge to you tomorrow. I have some time in the weekend, do you have any other stuff in the pipeline that I could do? 19:46
ruoso it does the action processing completely... there are some issues that need to be solved in rakudo yet... (mostly contextual variables)
pmichaud I'm not arguing that .end _should_ remain part of Any; I'm just saying that that's how Rakudo has been consistently interpreting the spec, except in places where it obviously makes little sense
19:46 ungluecksfee joined
pmichaud I started working on contextuals this morning... should have it done on Friday. 19:46
19:46 mizioumt left
pmichaud (Would be done sooner but $otherjob commands my time today/tomorrow) 19:46
ruoso pmichaud, I see... but I've just got at this point in a behavior that really scared me out 19:47
$foo.*bar is a good way of implementing plugins by composition
moritz_ I know one more (small) PGE bug... ~ reports nonsense in the error message if the terminator is not a literal
19:47 hercynium joined
moritz_ rakudo: 'a' ~~ /~ a [b]/ 19:47
p6eval rakudo ee7d45: ( no output )
moritz_ rakudo: 'a' ~~ /~ a [b]/ || say $! 19:48
p6eval rakudo ee7d45: OUTPUT«Use of uninitialized value␤␤»
moritz_ uhm, where can I find the error message?
ruoso I realised we've been too tolerant on what should be part of the Any API... 19:49
and $foo.*bar will suffer a lot from that
pmichaud ruoso: I'm open for changes or clarifications to the spec, that's for sure.
ruoso I think that's something we weren't taking into perspective
pmichaud more particularly, we really ought to start updating S32 with the things we do "know" already. 19:50
like eliminating type constraints when they really shouldn't be there.
moritz_ is also scared by the shear number of methods in Any
ruoso renaming end to finalize in Faz in the meanwhile
ruoso s/finalize/finish/ 19:51
baest moritz_: looks something I should be able to do, but I'm not sure that I understand what the problem is 19:54
ruoso it's cool how Perl 6 makes some things so damn easy...
pmichaud baest: I don't know of many tasks off the top of my head
ruoso Faz almost reimplements the core of the Catalyst framework
moritz_ baest: and I can't demonstrate it because I don't see how the error message can be made available
19:54 nihiliad joined
pmichaud moritz_: you're seeing an error message somewhere, yes? 19:55
so a sample rakudo program would demonstrate it, even if p6eval won't show it.
baest pmichaud: ok, it might be you had something. I'll find something
moritz_ pmichaud: no. The point of ~ in regexes is that it should provide some good error messages...
pmichaud: since it doesn't throw an exception anymore, I don't know how to access that error message 19:56
pmichaud baest: I'll be eager to look at/apply the double-quote patch.
moritz_: oh, I thought you were referring to a parsing error in the regex.
moritz_ pmichaud: iirc TimToady said it should be in $!, if it's not a thrown exception
baest pmichaud: cool, you have it, right?
pmichaud rakudo: 'a' ~~ /~ a [b] [c]/
p6eval rakudo ee7d45: ( no output )
pmichaud rakudo: 'ab' ~~ /~ a [b] [c]/
p6eval rakudo ee7d45: OUTPUT«Unable to parse , couldn't find final a␤current instr.: 'parrot;PGE;Match;FAILGOAL' pc 3008 (compilers/pge/PGE/Regex.pir:477)␤»
moritz_ ah 19:57
I got that wrong somhow
pmichaud it needed a 'b' to match.
moritz_ rakudo: 'ab' ~~ / ~ [a] b c /
pmichaud it's easier to see with 19:58
p6eval rakudo ee7d45: OUTPUT«Unable to parse , couldn't find final [␤current instr.: 'parrot;PGE;Match;FAILGOAL' pc 3008 (compilers/pge/PGE/Regex.pir:477)␤»
pmichaud rakudo: 'abc' ~~ / a ~ c b /
p6eval rakudo ee7d45: ( no output )
pmichaud rakudo: 'ab' ~~ / a ~ c b /
moritz_ baest: that's the error message I mean
p6eval rakudo ee7d45: OUTPUT«Unable to parse , couldn't find final c␤current instr.: 'parrot;PGE;Match;FAILGOAL' pc 3008 (compilers/pge/PGE/Regex.pir:477)␤»
moritz_ baest: it says couldn't find final [
baest pmichaud: the double-quote patch. I think I did send it to you, but I'm not sure
pmichaud baest: oh yes, you did. I meant to say I'm eager to see the quote word literal patch (sorry!) 19:59
moritz_ ie the first character of the regex to be matched, even if it's not a literal
19:59 azawawi left
pmichaud yes, we ought to be able to fix that error message somehow. 19:59
moritz_ if it's not easy to extract a good target for the error message, it would be nicer to just say "couldn't find delimiter" 20:00
or so
baest pmichaud: oh, yes that was more difficult (and fun) :)
pmichaud rakudo: 'ab' ~~ / a ~ 'c' b /
TimToady phone
p6eval rakudo ee7d45: OUTPUT«Unable to parse , couldn't find final 'c'␤current instr.: 'parrot;PGE;Match;FAILGOAL' pc 3008 (compilers/pge/PGE/Regex.pir:477)␤»
pmichaud I'm guessing that [b] doesn't set its .to and .from properly.
baest moritz_: ah, the wrong char is reported 20:01
moritz_ baest: exactly
ruoso pmichaud, is there some way to Dump a regex in rakudo?
pmichaud ideally I'd expect it to say "...couldn't find final [b]..."
ruoso: _dumper($/)
baest but shouldn't [b] just be 'b', that should be the same
pmichaud oh wait
you said regex
no.
a regex is just a sub.
at execution time, it's already been turned into code. 20:02
ruoso right...
TimToady nick and I are getting lonely
pmichaud on my way
having to locate telephone apparti
apparati
moritz_ rakudo: 'ab' ~~ / ~ 'b' a c / 20:03
p6eval rakudo ee7d45: OUTPUT«Unable to parse , couldn't find final 'b'␤current instr.: 'parrot;PGE;Match;FAILGOAL' pc 3008 (compilers/pge/PGE/Regex.pir:477)␤»
moritz_ rakudo: 'ab' ~~ / ~ (b) a c /
p6eval rakudo ee7d45: OUTPUT«Unable to parse , couldn't find final (␤current instr.: 'parrot;PGE;Match;FAILGOAL' pc 3008 (compilers/pge/PGE/Regex.pir:477)␤»
baest weird it says final. It wonder if it know it is the last one 20:04
s/I/It/
20:04 LadyLunacy left
pugs_svn r26708 | lwall++ | [STD] better error message for 'for lines {...}' 20:05
20:05 mizioumt1 left 20:08 masak joined
masak I won't stay long. 20:10
jnthn oh noes! 20:11
masak just thought I'd update Rakudo to chase the spec changes I introduced today.
jnthn yay :-)
masak (change the signature of .comb, implement .words)
in fact, I did this already. :)
took just under a minute.
running spectests now.
20:13 ejs joined, kaether joined
masak ok, t/spec/S32-str/comb.t is going to break badly. 20:13
maybe I should just rename it words.t :P
jnthn -> nom, bbs
baest moritz_: it looks like the ~ actually upsets the PGE parsing of the regex. I'll be happy to take a look at it in the weekend if nobody beats me to it 20:14
20:14 mizioumt joined
baest is off 20:15
ruoso masak, hmm... you're not on #november-wiki... 20:20
I just sent you a message there ;) 20:21
20:21 pmurias joined
masak ruoso: I'm only partially here, that's why... 20:21
pmurias ruoso: do you use a debugger when working on smop? 20:22
ruoso pmurias, ddd
or gdb if the problem is not so hard
Tene pmichaud: will i be stepping on any toes if i work on refactoring rakudo's trait handling to get it up to spec? or did you do that already? 20:31
Tene git log
moritz_ Tene: more likely jnthn will do 20:32
Tene nods.
20:35 goksie left 20:38 Jedai joined
pmichaud Tene: I think it's not too far away from spec now. We have to consider how we want things like 'readonly', 'rw', etc. to work. 20:40
but I'm fine with aiming to bring it closer to spec.
pugs_svn r26709 | masak++ | [S32-str] 20:41
r26709 | masak++ | * changed comb.t to new specced signature
r26709 | masak++ | * added words.t which partially takes comb's place
jnthn We need some parsing improvements to be able to parse multi trait_auxiliary:is
pmichaud that may actually happen soonish (like, within ~10 days)
jnthn I wasn't planning to hack on traits stuff for a copule of weeks myself...
pmichaud I'm rapidly wanting to define builtins in p6.
jnthn Sure.
That'd be nice. :-)
pmichaud and parsing things like "trait_auxiliary:is" would be part of that.
20:42 bacek left
moritz_ rakudo: multi trait:auxiliary:<is> { }; say "alive"; 20:43
p6eval rakudo ee7d45: OUTPUT«Malformed routine definition at line 1, near "trait:auxi"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)␤»
dalek kudo: a40974a | jnthn++ | src/ (2 files):
Use getinterp instead of new ParrotInterpreter to improve performance.
kudo: 45ddf59 | jnthn++ | src/builtins/ (2 files):
Use getinterp instead of new ParrotInterpreter to improve performance (part 2...missed some!)
moritz_ rakudo: multi trait_auxiliary:<is> { }; say "alive";
p6eval rakudo ee7d45: OUTPUT«Malformed routine definition at line 1, near "trait_auxi"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)␤»
moritz_ masak: it's nice to write builtins in Perl 6, no? ;-) 20:44
masak moritz_: surprisingly painless. :) 20:45
moritz_ aye ;-)
pmichaud squirrels away that quote for future use. :-)
ruoso writing everything in Perl 6 is surprisingly painless
pmichaud painless for *some* :-) 20:46
ruoso I'm still surprised how easy it was to get Faz running from some sketches...
masak ruoso: but it takes some time to write everything.
std: macro f { '1 + 1' }; say f()
p6eval std 26709: OUTPUT«ok 00:02 36m␤»
literal Faz?
ruoso github.com/ruoso/faz/tree/master 20:47
masak std: quasi { say "foo" };
p6eval std 26709: ( no output )
masak TimToady: ^ 20:48
why didn't it say "ok"?
mberends grabs the proto pumpkin while masak's not looking
masak mberends: just make sure to pull, 's all.
Su-Shee how do I say perl 5's wantarray in perl 6?
jnthn You don't. 20:49
masak mberends: I made a commit minutes ago.
dalek kudo: 35c6746 | masak++ | src/setting/Any-str.pm:
[setting/Any-str] implemented .words
kudo: 6b962c2 | masak++ | src/setting/Any-str.pm:
[setting/Any-str] changed signature of .comb
TimToady std: ;
kudo: 615936e | masak++ | t/spectest.data:
[spectest.data] added S32-str/words.t
p6eval std 26709: OUTPUT«ok 00:02 34m␤»
masak my work here is done.
good night, peeps.
pmichaud masak++
jnthn Su-Shee: There's no wantarray kinda thing in Perl 6.
Su-Shee: Instead, you return something that behaves the right way in different contexts.
masak dalek reports commits out of order!
dalek--
20:50 masak left
TimToady hmm, works okay here 20:50
20:50 desertm4x left
jnthn brb 20:50
TimToady std: quasi { say "foo" }; 20:51
p6eval std 26709: OUTPUT«ok 00:04 40m␤»
TimToady must've been recompiling or something
ruoso TimToady, I've just (about an hour ago) hit a design issue in the API 20:52
moritz_ TimToady: it did report it as a pass
TimToady: just obscured by dalek
TimToady no, that was me
his explicitly said (no output) 20:53
moritz_ ah
before the ;
ruoso TimToady, basically... the Any API is too big, and causing unexpected effects when using $foo.*bar (backlog for more details)
ruoso throws the issue in the airs and throws himself out of the chair in order to go home... 20:54
pmichaud start of relevant backlog: item end
our Any method end (@array: ) is export
Returns the final subscript of the first dimension; for a one-dimensional
argggh
sorry
start of relevant backlog: irclog.perlgeek.de/perl6/2009-05-06#i_1123715 20:55
TimToady I think C<hides> was invented to stop .* propagation
ruoso hmmm... that's a new one to me
TimToady classes intended to be used with .* have to be specially written anyway 20:56
ruoso so all methods in Any have this "hides" thing? 20:58
pmichaud I would think that a derived class would use <hides> to obscure the ones in Any. (Haven't read the spec... reading now.)
TimToady or Any is marked "is Hidden" 20:59
20:59 iblechbot left
TimToady afk & 21:00
ruoso class Action hides Any ?
pmichaud Yes.
ruoso but what if Action is a Role
is role Action hides Any still valid?
21:01 km2 left
ruoso really decommute & 21:01
21:01 ruoso left 21:08 ungluecksfee left 21:15 Whiteknight joined 21:20 skids left 21:22 donaldh_ joined 21:23 pmurias left 21:33 ejs left 21:39 hercynium left 21:40 Su-Shee left
TimToady in any case the base class for a .* method should probably use 'lastcall' in that final method, which would prevent propagation into Any 21:41
(it's possible that should be "is lastcall" as a trait on the method so it can return normally)
jnthn TimToady: Did you see my question earlier about .* and an explicit candidate list? 21:42
TimToady it just calls them all till "lastcall" suppresses the rest 21:47
jnthn OK, but suppose I have 21:49
class A { method m { } }; class B is A { method m { nextsame } }
my $x = B.new; my @cands = $x.WALK(:name<m>); 21:50
$x.*@cands();
How many times is m in class A called?
That is, does the "nextsame" in B's method m consume one of the candidaets in the "iterator"
And thus when we fall back into the .* dispatcher it sees that there is nothing left? 21:51
TimToady it's erroneous to mix .* and "next" 21:52
jnthn Oh?
OK.
TimToady as I said earlier, you have to write your methods specially to use .* 21:53
it's not transparent
jnthn OK, so .* does not give nextsame etc a candidate list to work with, so it's an error.
TimToady it could still mean something 21:54
but it would mean "skip this iteration and don't return a value in the list"
next means "I'm not here"
jnthn Hmm.
TimToady so it's not erroneous in that sense, I guess 21:55
jnthn That's a tad trickier to handle, but do-able I guess.
What about if I do
$x.m()
TimToady it's just a different dispatcher loop without a last
jnthn In the above example.
I didn't say there was a list of candidates to call here. 21:56
TimToady has a last in the dispatcher's loop after first success
there's always potentially a list of candidates for a method call
if there are parent classes
jnthn OK, that's what I'm getting at
So with class A { method m { } }; class B is A { method m { nextsame } }
TimToady .@foo overrides that list
jnthn my $x = B.new; $x.m;
Then the nextsame will call m in A? 21:57
Or rather, defer to...
TimToady yes
jnthn OK, I understand that bit then. :-)
TimToady but if you call it .@foo, it defs to the next one in @foo
jnthn Yes.
TimToady whatever dispatcher it's in the middle of, that's the one it iterates 21:58
jnthn And .$foo (where $foo is a Callable) just calls that, and a nextsame in that is an error because there's no next candidate?
TimToady likely
jnthn OK. 21:59
TimToady arguably it's just a function call written as single dispatch
but ."$foo" is symbolic, so calls A
jnthn That just interpolates $foo as a string, right?
TimToady yes
pmichaud open question from scrollback, before I close my window where I was working on it.... 22:00
19:32 <pmichaud> TimToady: is it only blocks containing placeholders that get the implicit @_/%_ when used?
TimToady which would you rather? 22:03
pmichaud from a rakudo implementation perspective, it's -slightly- easier if we allow it for any signatured block. 22:04
22:04 mikehh_ joined
pmichaud but only very slightly. User-level concerns would weigh stronger. 22:04
TimToady I can see that { @_ } might be surprising occasionally, but it seems more consistent 22:05
so I think I'm okay with that
pmichaud well, I'm presuming the block has to have a signature of some sort before we get the magical @_
otherwise it's a simple lexical
TimToady but does it autogen a sig like a placeholder?
pmichaud I hope not. 22:06
Otherwise we end up with an inability to get to a sub's @_ from a nested block.
TimToady that's what I meant when I said "surprising"
pmichaud Oh.
My original question was presuming the block had some sort of other signature generator already in place.
TimToady well, only placeholders do that so far, right? 22:07
pmichaud Correct.
TimToady but we could automatically allow autoslurpy of @_ and %_ on any existing sig 22:08
if there isn't already a slurpy there
-> { @_ } would work then
but it still seems a bit chancey
*@_ = OUTER::<@_> maybe 22:09
pmichaud actually, I think that's the case that decides it for me (more)
moritz_ doesn't like -> { @_ }
pmichaud -> $a, $b { @_ }
TimToady wait, you can't default a slurpy :)
pmichaud I'd expect the @_ to be the OUTER @_, and not an implicit *@_ on the pointy block.
TimToady well, I guess we can restrict it to placeholders then 22:10
pmichaud because then the pointy block would slurp up all of the values of the loop
if we had
for @list -> $a, $b { @_ }
we'd only get one iteration.
oh, maybe not.
but you kinda get the idea.
@_ becomes a bit too magical in that case.
lambdabot Maybe you meant: . ? @ v
TimToady given our other talks about $a, $b, $c? I'd think it'd slurp all the args 22:11
pmichaud right
TimToady well, I gotto run away for a while
pmichaud I'm wondering if a sub with a slurpy positional ends up with .count == Inf
moritz_ if I explicitly write I signature, I wouldn't like the idea of it getting messed with by some internal magic
TimToady anything else hanging fire?
moritz_: agreed
pmichaud no, I just needed to know that one so I can add the correct fix to Rakudo.
TimToady probably be back online later tonight
pmichaud any other fires from me will have to wait until Friday, I suspect. 22:12
TimToady driving to Santa Rosa & 22:13
jnthn pmichaud: We have a test that looks like this: 22:14
(for P6object)
obj = proto.'new'()
isa_nok(obj, 'P6Protoobject', objname)
22:14 exodist left
jnthn Oh, hmm, let me think how to ask this... :-) 22:15
pmichaud okay to deprecate/change.
jnthn Ok.
I mean, everything is an instance of the Protoobject PMC now.
pmichaud I don't think anything depends on that. But it is nice to have some way to ask "are you a protoobject"
jnthn $thingy === $thing.WHAT
pmichaud that's not exactly the same. 22:16
jnthn or $thingy =:= $thing.WHAT
No, true.
pmichaud there's not ever just one protoobject
Dog{ :name<fido> } is still a "protoobject"
jnthn Yeah, of course.
pmichaud but it's not =:= Dog.
jnthn True. 22:17
moritz_ uhm, =:= only queries containers, not types?
jnthn I mean, I know how to make the information available.
erm, rephrase
pmichaud That's fine. As long as there's a way to determine protoobject-ness, I'm happy.
jnthn I know how to determine it cheaply (in the PMC)
I just don't know the best way to make it available.
pmichaud presumably we could expose it as a method or via some obscure vtable. 22:18
jnthn not as a method, but as a vtable perhaps, yes.
inspect_str is the obvious one.
pmichaud yes, except "inspect" really worries me at times.
jnthn Or get_pointer. You can do anything with get_pointer. ;-)
pmichaud because of its putative "cloning semantics".
jnthn That might be something we choose to fix in Protoobject. 22:19
pmichaud Also, if what we have *isn't* a protoobject, I'd be a bit worried about running inspect_str on it.
jnthn True.
That may not work out so well.
Hmm.
22:19 mikehh left
jnthn It needs to be some vtable thing I guess rather than a method. 22:20
pmichaud So, is the Protoobject PMC analogous to P6protoobject or to P6opaque ?
jnthn p6opaque
There is no p6protoobject as such after this though.
pmichaud I think I'd prefer it to _not_ be named Protoobject, then.
because Protoobject really meant something else.
jnthn parrotopaque made most sense to me. 22:21
pmichaud I hadn't quite made the connection that the PMC replacing P6protoobject would be the base class for all objects.
I like Parrotopaque
you can even capitalize the 'O' in that case, if you want.
jnthn :-P
pmichaud Parr0t0paque.
jnthn Yeah, there's no separate proto-object thingy any more.
There's just one thingy that is a proto by virtue of having no instance storage. 22:22
pmichaud so how do we implement .ACCEPTS on proto-object thingies, then?
same way we do .defines and .Str ?
that feels.... messy.
jnthn I can only see that we'd be able to intercept them. 22:23
pmichaud I don't quite grok "intercept" in this case.
jnthn It is, but I guess it's a dispatcher decision.
Currently in the defined vtable method I do
pmichaud You mean on every method dispatch to every object we'd be looking for special instances of protoobject thingies?
jnthn Well, we'd be saying "is this a proto" first and only then doing it. 22:24
pmichaud what would be saying that?
jnthn The dispatcher in the meta-class probably should be. I'm fully expecting we'll cheat in the common case. 22:25
pmichaud I'm having trouble seeing even the cheat.
But essentially you're saying the dispatcher has to be aware of all proto-object specific methods?
jnthn Yeah. I'm not entirely sure I like that, but I don't see another way. 22:26
Of course
pmichaud I definitely don't like it.
jnthn We don't have to look at the method name.
It looks more like how I have defined.
Parrot_Protoobject_attributes *po = (Parrot_Protoobject_attributes*)PMC_data(SELF);
if (!PMC_IS_NULL(po->attrib_store))
return SUPER();
else
return 0; 22:27
22:27 AzureStone left
jnthn That is, if we have storage allocated, we're an instance, do whatever an Object would do. 22:27
pmichaud I understand it at the vtable level, yes.
But I'm asking specifically about .ACCEPTS
jnthn Otherwise, we're undefined.
OK, in this case our find_method would looke like the above.
pmichaud except we wouldn't return 0 22:28
jnthn Apart from instead of "return 0" we'd say something like "look in this set of proto-specific methods"
Plus in p6opaque we won't call SUPER() but instead we'll have the Perl 6 dispatcher right there (in the p6opaque/default metaclass version).
pmichaud could those methods come from a Role or something like that?
jnthn We could have a way of saying "use the methods from this role" 22:29
(as the proto-fallback ones)
pmichaud I'm a bit more comfortable with that.
jnthn And it'd basically steal the role's methods hash.
Essentially it comes the "proto-overrides" list.
pmichaud Does Perl 6 have a way to ask an object "are you a type object"?
other than simply checking it for definedness? 22:30
jnthn It's hand-wavey. There's the Abstraction role.
pmichaud oh, type objects do Abstraction, yes?
jnthn But sicne the proto-object is just meant to be, well, a not-so-defined normal object...
pmichaud well, clearly the normal objects don't "do Abstraction"
jnthn How exactly Abstraction fits in isn't so clear to me. 22:31
pmichaud I still think that type objects (protoobjects) differ by a role of some sort.
Abstraction being the obvious case.
jnthn They can lie about doing the Abstraction role easy enough. And that's probably the easiest way.
We can't just go in-place mixing in to a proto-object though. 22:32
The proto-object and the real objects today are different types so far as Parrot is concerned, and I don't think that's really the way things should be.
22:33 mberends left
jnthn (where real object = the instance of) 22:33
pmichaud it sounds to me like a type object is ultimately something that .does(Abstraction) -- no matter how we internally cheat to get it that way.
jnthn Not least because S12 says otherwise, but because even in writing the meta-class extra bits recently I've seen places where we're going to have to cheat elsewhere if we don'thave it like that. 22:34
Yes, I agree.
.does(Abstraction) being true is The Way.
pmichaud it doesn't bother me that they're different types -- since they actually respond differently.
jnthn It does bother me.
I think it's going to cause us problems.
pmichaud in the Parrot version of this, I think it should be .does("abstraction") 22:35
jnthn I can make it that way.
pmichaud in the same sense that Parrot currently has a "scalar" type
jnthn *nod*
pmichaud (and "array" and "hash")
then in the Rakudo version, we promote to checking for the Abstraction role 22:36
jnthn *nod*
pmichaud and the Abstraction/abstraction role(s) can then do the cheating to say "okay, are you _really_ an abstract object?" and provide the alternate method definitions.
s/abstract object/type object/
jnthn Yes. 22:37
pmichaud because we have to overload not only .ACCEPTS, but also various postcircumfixes and lots of other things that are type-object-dependent
jnthn Right.
As you can imagine, this isn't the most -Ofun refactor. I'd not be doing this if I didn't think we really needed to. 22:38
pmichaud anyway, for now I'm happy with $I0 = does $P0, "abstraction" as being the "are you a type object" test.
jnthn OK, great.
pmichaud for backwards compatibility, we might temporarily overload VTABLE_isa to explicitly check for "P6protoobject" until we can get through a deprecation cycle. 22:39
(since non-protoobjects would obviously return false in that case.)
i.e., we can have a false P6protoobject class that hangs around for backwards compability until we can fully deprecate it.
I _hope_ not too many people have installed methods in P6protoobject.
(if they have, we could have our dispatcher check for those also for a while) 22:40
jnthn I find it quite unlikely.
Heh, we could go look up the methods in P6protoobject and use them as our abstraction ones...
pmichaud Right
That might be a nice intermediate.
But deprecate it in favor of some other better-named role for the future. 22:41
(or alias it somehow)
jnthn OK 22:43
22:56 AzureStone joined
mizioumt help /me 22:56
oops, sorry 22:58
22:59 ruoso joined 23:01 frew|work joined 23:03 frew|wor1 left
jnthn rakudo: say Int === Int 23:03
p6eval rakudo 615936: OUTPUT«1␤»
jnthn rakudo: say Int{ :Int(42) } === Int 23:04
p6eval rakudo 615936: OUTPUT«1␤»
jnthn Hmm.
rakudo: say Int{ :Int(42) } =:= Int
pmichaud I'm gone for the rest of the evening and most of tomorrow -- gotta finish my $otherjob stuff.
p6eval rakudo 615936: OUTPUT«0␤»
pmichaud See you all late tomorrow night, or Friday. 23:05
jnthn pmichaud: OK, hope it goes smoothly.
pmichaud Me too, although I think it will all go fine.
jnthn :-)
pmichaud later
jnthn cu
rakudo: say Int{ :Int(42) }.WHAT =:= Int.WHAT
p6eval rakudo 615936: OUTPUT«1␤»
23:08 donaldh_ left 23:15 samlh joined, payload joined
Infinoid aa 23:16
(oops)
meppl good night 23:17
23:19 meppl left 23:20 donaldh left, donaldh joined 23:26 hercynium joined 23:30 Kyosuke_Kiryu joined 23:33 Kyosuke_Kiryu left 23:36 DemoFreak left
ruoso jnthn, btw... the idea for the object to lie about .does(Abstraction) is interesting... I'm still unsure on how that relates to uninitialized values and autovivifiable objects... 23:53
23:54 skids joined
ruoso since an uninitialized object is a type of Failure, and failures are undefined as well 23:54
which mean they don't have instance storage
jnthn ruoso: definedness is not just about "doesn't have instane storage" 23:55
ruoso what else is it about?
jnthn It's up to the HOW how it chooses to dispatch .defined (if I remember my S12 correctly) 23:56
Though at a more basic level I guess you can write a .defined method.
ruoso yes... Object implements .defiend
.defined
jnthn Sure, but you can override it to do what you like.
But this means .defined isn't a way really to say "is this a proto-object or not". 23:57
ruoso that's my point...
how do you say that?
23:57 justatheory left
jnthn Well, that's what the discussion was about. But I'm guessing "says is does the Abstraction" role could be an answer... 23:57
ruoso right... you're just moving the question to someone else ;) 23:58
jnthn Since type-y things are meant to.
ruoso how do a object knows if it is a type object or not
how does it decide to say true to .does(Abstraction)
jnthn Oh, for *that* I'm just going off "does it have instance storage" for type objects.
ruoso heh... 23:59
jnthn I'm curious though how in smop you plan to handle methods that dispatch differently in the type object?
e.g. 42 ~~ Int is very different from 42 ~~ 1

Powered by Cro and the Raku® Programming Language.

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

Thank you!