Check your feather email | spec.pugscode.org | paste: sial.org/pbot/perl6 | pugs.blogs.com | www.treehugger.com/files/th_images/paradigm.jpg
Set by rodi on 29 September 2006.
00:03 spoop joined 00:08 lisppaste3 joined 00:13 vel joined 00:21 nekokak joined 00:26 putter joined
TimToady is that ねこかき or ねこかく or ねこかっ？ 00:26
nekokak: is that ねこかき or ねこかく or ねこかっ？ 00:31
or is it finnish?
or is it mojibake on your terminal? 00:33
mugwump TreyHarris: I'm doing a design for O/R mapping, and trying to figure out how to tell that attributes are parts of the primary key 00:35
TreyHarris i wonder if there's some way to tell colabti.de that #perl6 should be considered to be a utf8 channel, i always have to change the encoding from western-1
00:36 etzel joined
TreyHarris mugwump: hmm... wouldn't doing it using .WHICH, even if it worked that way, require doing it experimentally? (i.e., if changing this attribute results in a new object, i must have changed something in the primary key?) 00:36
mugwump that's why I think WHICH is a macro, not a function 00:37
TreyHarris what does the macro do?
mugwump but that can't work, because you have runtime type hopping. hmm. 00:38
ok, here's the application;
say I've got a class which I'm expected to be automatically backed by a database;
class Dog { has Str $.tag; has Country $.registered_country; has Num $.weight; method WHICH of (Str, Country) { $.tag, $registered_country } } 00:39
from that, I can tell I need to create a table with two columns, but I have to guess which properties are a part of the key 00:40
TreyHarris you mean, from outside the class you can't tell 00:43
TimToady this sounds like the sort of stuff you should be talking over with dduncan 00:44
TreyHarris but can WHICH return any arbitrary data, including a Seq like that?
TimToady TreyHarris: yes, or we should just rewrite colabti's Python script in Perl... 00:45
WHICH is required to return something that is comparable with ===
mugwump Yeah, I ran a paper past dduncan on this, and he liked it
he's been the one to comment on it the most so far
TimToady we let people override WHICH to make value types, but using the override to make flapping-in-the-wind types would be erroneous 00:46
TreyHarris eh, I thought === was defined in terms of WHICH?
TimToady they are defined in terms of each other.
mugwump turtles all the way down and all that :)
TreyHarris define "flapping-in-the-wind types"? sorry, haven't heard that term before
mugwump that's variant WHICH return values 00:47
TimToady objects that change their identity on the fly
would drive a sane hash nuts.
WHICH is supposed to be immutable
mugwump ok, in my paper [utsl.gen.nz/talks/dbic-tangram.pdf], I'm going to say that the behaviour if changing attributes that affect .WHICH is undefined and implementation-specific 00:48
TreyHarris i think that was the bone of contention earlier. .WHICH is immutable, or its mutating indicates that you no longer have the same object?
TimToady I do love the term "erroneous". :)
mugwump Well, I think to dictate either behaviour would hurt the VM that doesn't work like that. 00:49
TreyHarris mugwump: thinking of it in terms of Tangram does make it clearer what you're asking though :)
lumi I love the phrase "So don't do that." 00:50
mugwump One other small thing, I'm using 'of' incorrectly. I need a new preposition
Seq of (Int) doesn't mean a sequence of exactly one int as specced
Or does it?
Seq of Int sure means a sequence of integers of any length, is Seq of (Int) different enough? 00:51
TimToady not necessarily, the Seq role can interpret its parameter however it likes.
it doesn't have to distribute the "of" like Array necessarily 00:52
TreyHarris or "of Seq[Int]" versus "of Seq[Int, Int]" versus "of Seq[Int, Int, Str, Int]"
TimToady so maybe Seq of (Int*) is the distributed form
mugwump is that a new use of postfix:<*> ?
TimToady the "slurpy" form.
no 00:53
mugwump oh, like a method sig
TimToady maybe it has to be written Seq of :(Int*)
or Seq of :(Int) 00:54
seems a little inside out
TreyHarris yeah... if a sig has Type as a parameter, it would be nice if Int as an argument would be interpreted as ::Int, and not as Int()
TimToady Seq[:of(:(Int))]
mugwump heh
TreyHarris then just Seq[Int] could work, no? 00:55
TimToady well, that's what Seq of :(Int) means.
depends on how the sig of Seq is declared I imagine
mugwump I'd write it Seq[ of => :(Int) ] to avoid the very close use of two different meanings of :
TimToady if role Seq is splatting the Type param already, then Seq of (Int,Int) could work 00:56
if it's expecting only a single arg which is a sig, then no 00:57
the latter is probably more consistent 00:58
variadic types outside of signatures is a bit suspect, notationally.
I would expect List of Int to distribute though. 00:59
mugwump yeah, I think I'd rather use a parametric collection type than have variable length parts of a Seq
TimToady but Seq is more like Scalar, except with more bits in it
TreyHarris yes.... but TreeNode[::X = Whatever] would be legal, though, if you're never going to take variadic args, yes?
mugwump hmm, variadic args are more like TreeNode[\$capture] 01:00
s/args/types/
TimToady I can well imagine other uses for the variadicness of such an argument list, so maybe it's better to limit types to a single parameter 01:01
especially since :of is named. 01:02
TreyHarris i forgot, do we write that TreeNode[::X = Whatever] or TreeNode[::X = *] ?
mugwump Well, in the Seq case we want more than one, for our Seq of :(Int, Str, Int) 01:03
TimToady I think use * as rvalue and Whatever to match against it in a sig.
but the default is an rvalue
mugwump But I could fully forgive a role implementation for not supporting certain classes of Signature
TimToady sure
TreyHarris TimToady: ok. so TreeNode[::X = *] but multi dwim (Whatever $x) { ... } 01:04
mugwump hmm, perhaps the role requires a hook during role or class composition, too
TimToady In general the $x is not necessary
since Whatever is presumably a singleton 01:05
TreyHarris sorry, I was using ... as "fill in here" rather than "yada-yada"
so I was saying "I'll take an arg of any type and use it as $x below"
TimToady doesn't matter, you still don't need the $x
that's not what Whatever does
Whatever is a very specific type 01:06
it matches '*' and only *
if you want any type, use Any
mugwump unless you also want to match junctions, in which case use Any|Junction :)
TimToady whatever is for default case selected from the caller's end.
TreyHarris right, I remember that now. 01:07
TimToady so you'll have a lot of specific multis and then a special Whatever case
1..10 never matches Whatever, but 1..* does
the right arg, that is 01:08
TreyHarris mugwump: Any doesn't match a Junction? and (3|4|5) !~~ Int? 01:09
mugwump ?eval sub foo(Any $x) { say "I got $x" }; foo(1&2)
evalbot_r13794 OUTPUT[I got 1 I got 2 ] (Bool::True)
TimToady autothreads 01:10
mugwump ?eval sub foo(Any|Junction $x) { say "I got $x" }; foo(1&2)
evalbot_r13794 OUTPUT[all(VStr "I got 1",VStr "I got 2") ] Bool::True
TreyHarris ?eval (3|4|5) ~~ Int
evalbot_r13794 (Bool::True)
01:10 weinig|bbl is now known as weinig
TreyHarris ?eval (3|4|"moose") ~~ Int 01:10
evalbot_r13794 (Bool::False | Bool::True)
TimToady we have to be careful because passing an unexpected junction into a sub would make both legs of an if/else run.
TreyHarris gotcha.
TimToady (arguably, that's what autothreading does) 01:11
but the naïve implementation it would do the if and skip the else
TreyHarris iff the value is both true and false, you mean?
TimToady despite the fact the the junction is both true and false.
yes
TreyHarris an all-false junction would just run the else, and an all-true would just run the if regardless 01:12
TimToady presumably
mugwump I'll ask questions about how Junction is defined that gives it this special behaviour some other day, but for now I'm curious as to whether there's a BUILD equivalent for role composition
Actually I'll ask Damian that Junction question 01:13
TimToady yes, it's called BUILD.
TreyHarris mugwump: according to audreyt, BUILD :)
TimToady all the role BUILDs are magically gathered into the class BUILD at composition time.
<insert handwaving character here>
mugwump but that's an object construction hook, I want a class composition hook 01:14
eg, to generate some methods in the role based on the type parameters
s/roles/class/
bah
eg, to generate some methods in the class based on the role type parameters
TimToady that would be in the province of the MRP 01:15
mugwump in a sense, a sub definition that uses a type from the role's type argument does this
MRP?
TimToady just as a class body runs as a MOP object instance, the role body is run by some kind of MRP object instance.
01:15 Qiang joined
TimToady (I'm making this up on the fly.) 01:15
anyway, somebody's methods are getting called when that closure runs. 01:16
mugwump sure. It might be good to have syntax for that sometime so that you don't have to subclass your metaclass to do this, but I think we need parametric roles prototyped first
TimToady It's possible stevan has already thought this part rather more thoroughly than me. 01:17
TreyHarris mugwump: so "role Antlered { submethod COMPOSE { foo() } }; class Moose does Antlered { #[ ... ] }; class Elk is Moose { #[ ... ] }; my Str $baz = "moose!" but Antlered; 01:18
you'd expect COMPOSE to be called three times at compiletime and once at runtime?
or four times at compile time?
er... s/two/; s/four/three/; # can't count 01:19
s/three/two/; # can't write syntactically correct substitutions either
TimToady and perhaps COMPOSE is a block trait rather than a submethod... 01:20
or maybe it's a macro
hmm.
role macros. yum.
installs new declarators, for instance. 01:21
TreyHarris maybe it just gets called once at compile time for Moose; Elk is a Moose and so it defers to Moose its Antleredhood. so you call just twice. (maybe it can infer from My Str $baz = "moose!" but Antlered at compile time it needs a Str that does Antlered, but in the general case, you're sometimes not going to know what to compose until runtime) 01:22
TimToady it would be good to avoid keywords that look like they want to be generic when you mean something non-generic.
mugwump what's an example of a "declarator"? I should understand that term, but I don't
TimToady "my", "has"
"requires_method 01:23
mugwump also "method", etc ?
TimToady "must"
yes
a role could have special declarators for requirements.
if the yadas don't suffice.
much like in ruby, but cooler
mugwump the yadas are runtime anyway, if written method X { ... } 01:24
TimToady as macros you're not limited to one syntax.
I think S12 allows the class to assume that all role ... should be resolved at composition time by default.
assuming that classes are elaborated in derivation order... 01:25
so you know whether your parent classes supply the method.
01:25 spoop joined
mugwump I would have thought that just writing the method signature without a block was enough 01:26
TimToady but I'm just talking through my hat here
mugwump like a forward declaration of a sub
TimToady {...} is intended to fail at runtime, but that doesn't mean the compiler can't fail earlier if it can figure it out. 01:27
it's there for the compiler, not just the human.
mugwump true, but all those extra characters. the Portugese will hate it
:)
TimToady {¡*!} might make the Spanish happy 01:29
mugwump What's that, the Siesta operator?
TimToady dunno, I'll tell you tomorrow.
my wife is picking me up in about 1 minute, so gotta dash...
biab & 01:30
mugwump seeya!
01:30 mako132_ joined 01:35 buu joined 01:40 buubot joined 01:42 lollan joined
svnbot6 r13795 | mugwump++ | Add test for specifying memoized instances via WHICH 01:52
r13795 | mugwump++ | This also allows for singletons.
putter ?eval my @x = <a b c>; scalar(@a) 01:55
01:55 evalbot_r13794 is now known as evalbot_r13795
evalbot_r13795 Error: Undeclared variable: "@a" 01:55
putter ?eval my @x = <a b c>; scalar(@x) 01:56
evalbot_r13795 Error: No compatible subroutine found: "&scalar"
putter didn't there use to be a scalar()?
mugwump ?eval my @x=^4; +@x 02:04
evalbot_r13795 4
mugwump slaps evalbot_r13795
?eval my @x=<0 1 2 3 4>; +@x 02:05
evalbot_r13795 5
mugwump ?eval ^4
evalbot_r13795 (0.0, 1.0, 2.0, 3.0)
mugwump ?eval my @x=<0 1 2 3 4>; item(@x) 02:06
evalbot_r13795 ["0", "1", "2", "3", "4"]
svnbot6 r13796 | putter++ | [t/xx-uncategorized/lexical_variable_in_eval.t] 02:10
r13796 | putter++ | Created. eval_ok isn't seeing lexical variables.
02:14 weinig is now known as weinig|sleep 02:23 chaoslawful joined
mugwump TimToady: ok, I'm assuming that Seq requires a Capture passed to its :of role parameter to get the per-element typing I use in my paper... 02:23
ie, Seq of Int works like Array of Int, Seq of :(Int) means a single integer Seq 02:26
TreyHarris like in Haskell, the difference between [Int] and (Int) 02:28
putter ?eval "abc\c[HEBREW POINT HIRIQ]def" 02:29
02:29 evalbot_r13795 is now known as evalbot_r13796
evalbot_r13796 "abcÖ´def" 02:29
putter ?eval "abc\c[LINE FEED]def" 02:30
evalbot_r13796 Error: Invalid unicode character name: LINE FEED
stevan putter: !
putter stevan! :)
mugwump ?eval "abc\c[LINE FEED (LF)]def"
evalbot_r13796 "abc\ndef"
mugwump :>
stevan putter: sorry I missed the Boston.pm thing,.. $work kept me busy until too late in the day
mugwump thinks he might know which file that is using ... :)
putter ?eval "abc\c[CARRIAGE RETURN]def" 02:31
evalbot_r13796 Error: Invalid unicode character name: CARRIAGE RETURN
stevan mugwump: re: what you were talking about with TimToady eaerlier
putter stevan: so no problem. I was going to tell you "sorry I missed the Boston.pm thing,..."
TreyHarris mugwump: "LINE FEED (LF)" isn't in my unicode name table...
stevan putter: :)
mugwump putter: get utsl.gen.nz/scripts/unicode
stevan mugwump: are you looking to create different sets of methods based on the role generic parameter? 02:32
or specific type based stuff?
(for Array vs. Str, etc etc)
mugwump well, this is one possible way to implement any of the parametric role features
putter mugwump: what is this binary file and why am I getting it? 02:33
mugwump er, it's a script. sorry, my web server doesn't set mime types for files with no extension
then you can go, eg, unicode 'hot springs' to see 'hot springs' chars or unicode '' 20 to see page 20 02:34
putter ah, I see... 02:35
?eval "abc\c[CARRIAGE RETURN (CR)]def"
evalbot_r13796 "abc\rdef"
mugwump stevan: so, eg, when you write:
role Foo[::T] { sub blah(T $x) { ... } }
s/sub/method/
the "sub" declarator is generating a new signature, composing the role parameter into the class being built 02:36
s/sub/method/ again
anyway, those declarators are methods on the role metaclass 02:37
putter mugwump: thanks!
stevan mugwump: I always saw Roles as being kind of lazy, but parametric roles as being *really* lazy 02:38
putter so \N{} in p5 tolerates the " (LF)" being missing, but pugs currently doesn't.
mugwump yeah, they're real louts. 02:39
TreyHarris mugwump: it's generating a new signature? i would think it's generating the signature :(Foo self: T $x where { self.does(Foo[::T]) }) or something like that. but my mind may be too polluted with Haskell's type inference.... 02:40
mugwump you mean GADTs ?
stevan TreyHarris: I saw it that way as well 02:41
TreyHarris mugwump: yes, exactly
mugwump ok, so when you write a parametric role definition like that, you need a concrete type when you use it with "does"
then, it composes the method into the class, resolving types that are generic 02:42
you can't have parametric classes
stevan mugwump: it might help to look at the role body as being deferred
TreyHarris my $foo = Foo[Int]; class Blah does Foo[Int]; class Baz[::T] does Foo[T];
mugwump what's a role body?
TreyHarris you can't have parametric classes?
mugwump TreyHarris: no. 02:43
TreyHarris how does Array work then?
or Seq?
stevan body == role { # this is the body }
mugwump they're roles
stevan mugwump: where as a normal (non parameterized) role's body would execute at compile time
a parameterized role's body is executed each time you pass in a new parameter
mugwump hmm. that makes sense. 02:44
stevan mugwump: in one of the MM prototypes I had parameterized roles as being like this: 02:45
mugwump I think the parametric :) role is also run a first time at compile time
stevan class 'Foo' => sub { my $T = shift; ... bunch of meta operations ... };
s/class/role/
mugwump right, so $T is the composing class? 02:46
stevan the $T was then in the closure and would generate a new package named 'Foo[T]'
mugwump oh
ok
stevan no,.. $T was the type parameter
it was a quick hack,.. but it seemed to work quite well
mugwump then, Foo[T] is composed?
putter oh, that's cute. a new way to segfault ghc. :/
TreyHarris is there anything a class can do that a role *can't* do then? 02:47
stevan mugwump: let me find the test in svn, it might make more sense to see it
mugwump ok.
TreyHarris: sure. be instantiated, for instance
TreyHarris because it's clear that roles have abilities--being abstract, having parameters--that classes do not
stevan mugwump: see this test - svn.openfoundry.org/pugs/perl5/Perl..._classes.t
lambdabot tinyurl.com/jj6ob 02:48
stevan and this one - svn.openfoundry.org/pugs/perl5/Perl..._classes.t
lambdabot tinyurl.com/z8edq
mugwump interesting. I note that's using classes, not roles 02:50
stevan yes, it should conceptually work for roles as well 02:51
at least in that metamodel
mugwump sure
stevan we are getting close to a solid role implementation in Moose, at which point adding parameters will probably be a next step
mugwump what's still slightly gelatinous? 02:52
stevan the internals
TreyHarris this very test is what I based my thinking that classes could be parameterized from. so that's not true, it's only true in Perl6::MetaModel?
svnbot6 r13797 | putter++ | [t/xx-uncategorized/unicode_aliases.t]
r13797 | putter++ | Created. "\c[LINE FEED (LF)]" works, but not "\c[LINE FEED]" or "\c[LF]".
r13797 | putter++ | And the third test segfaults my pugs.
stevan they are pretty bad IMO
TreyHarris: I am not sure why @larry does not want parameterized classes 02:53
mugwump: I am refactoring the composition part using Set::Object actually
mugwump A nice module, that one. jll++ 02:54
stevan TreyHarris: Perl6::MetaModel bends a number of "rules" on the meta-level,.. which may not be available on the user level
02:54 mako132_ joined, Qiang_ joined
stevan has to run, his ice cream is melting :) 02:55
TreyHarris ok. the only problem i see, and I'll have to consider it some more, is that there are patterns in which you have the following inheritance line: Abstract -> Concrete -> Abstract -> Concrete. this seems to be unsupported currently, as a class can do a role, but a role cannot inherit from a class. i don't think you can even fake it out with delegation, since there's no way to say a role is or does a class.... 02:56
you'd have to fake it out with a runtime exception if you instantiate the "abstract class" 02:59
putter ?eval sub ok(*@a){} ok(!defined $/[0], "Correctly didn't capture 2"); 03:00
03:00 evalbot_r13796 is now known as evalbot_r13797
evalbot_r13797 Error: No compatible subroutine found: "&defined" 03:00
putter bug?
TreyHarris ?eval sub ok(*@a){} ok(!defined($/[0]), "Correctly didn't capture 2"); 03:01
03:01 mauke_ joined
evalbot_r13797 undef 03:01
TreyHarris or more properly:
?eval sub ok(*@a){} ok(! $/[0].defined, "Correctly didn't capture 2");
evalbot_r13797 undef
putter my impression is defined() is arity 1, suggesting a misparse. 03:02
TreyHarris putter: See S29. defined is arity 1-or-2 03:03
but arg 2 is a role
and "Correctly didn't capture 2" isn't
putter ahh! my thanks. TreyHarris++ 03:04
03:05 SubStack joined
TreyHarris putter: another case where TimToady's current pet peeve would help. 03:06
"no subroutine compatible with *what*?"
putter :) 03:09
putter goes to see if exists() has similar issues...
TreyHarris stevan: the classic example of the exclusionary role has been (to make up syntax) "role Pregnancy requires Mammal excludes Platypus"
TreyHarris will bbiab 03:11
putter exists (@array : Int *@indices ), weeee... 03:12
but not quite that bad. a unary form is said to exist 03:14
03:16 mauke_ is now known as mauke
putter oh, no. exists(%h,<key>) 03:16
obra 'evening
putter wonders if there is a recursive exists(), so one can ask something like exists(%h<a><b>[3]<c>) 03:17
'evening obra
exists %h<a><b> looks so much cleaner than exists(%h<a>,<b>) or %h<a>.exists(<b>). :/ 03:28
" a b\tc" ~~ m/%<chars>:=( \s+ \S+ )/; defined($/<chars>{' a'}) 03:34
?eval " a b\tc" ~~ m/%<chars>:=( \s+ \S+ )/; defined($/<chars>{' a'})
evalbot_r13797 Error: Can't modify constant item: VUndef
putter bug? 03:35
03:39 spoop joined
mugwump ?eval my %h=<1,2,3,4>;say "yes: {%h.exists(1)} no: {%h.exists(2)}" 03:40
evalbot_r13797 OUTPUT[yes: no: ] Bool::True
mugwump ?eval my %h=(1,2,3,4);say "yes: {%h.exists(1)} no: {%h.exists(2)}"
evalbot_r13797 OUTPUT[yes: 1 no: ] Bool::True
putter bah, my mistake 03:41
the bug is the regex is failing.
mugwump gah! damned readline bug in pugs 03:46
03:48 vel joined
mugwump putter: is reduce :{$^a&&$^a.exists($^b)&&($^a~~Array??$^a[$^b]!!$^a{$^b})} too much to type? 03:49
?eval sub _exists(Item $x, *@x){ reduce :{$^a&&$^a.exists($^b)&&($^a~~Array??$^a[$^b]!!$^a{$^b})} $x, @x }; _exists(%h, <a b 3 c>) 03:51
evalbot_r13797 Error: Undeclared variable: "%h"
mugwump grr
?eval my%h=(a=>{b=>[1,2,3,{c=>4}]}); sub _exists(Item $x, *@x){ reduce :{$^a&&$^a.exists($^b)&&($^a~~Array??$^a[$^b]!!$^a{$^b})} $x, @x }; _exists(%h, <a b 3 c>)
evalbot_r13797 4
mugwump ?eval my%h=(a=>{b=>[1,2,3,{c=>4}]}); sub _exists(Item $x, *@x){ reduce :{$^a&&$^a.exists($^b)&&($^a~~Array??$^a[$^b]!!$^a{$^b})} $x, @x }; _exists(%h, <a b 3 d>)
evalbot_r13797 Bool::False
mugwump ?eval my%h=(a=>{b=>[1,2,3,{c=>4}]}); sub _exists(Item $x, *@x){ reduce :{$^a&&$^a.exists($^b)&&($^a~~Array??$^a[$^b]!!$^a{$^b})} $x, @x }; _exists(%h, <a b>)
evalbot_r13797 [1, 2, 3, {("c" => 4),}]
mugwump ?eval my%h=(a=>{b=>[1,2,3,{c=>4}]}); sub _exists(Item $x, *@x){ reduce :{$^a&&$^a.exists($^b)&&($^a~~Array??$^a[$^b]!!$^a{$^b})} $x, @x }; _exists(%h, <a d>)
evalbot_r13797 Bool::False
mugwump actually, that's not right, is it? the last value could be Bool::False 03:52
04:04 SubStack joined
putter :) 04:13
buu Hrmph. I need the lexical pad from an eval
TreyHarris mugwump: what's the readline bug? 04:17
mugwump if the line is very long, the display doesn't work 04:25
TreyHarris pugs interactive, or a program using readline? 04:26
mugwump pugs
TreyHarris the lines above are too long for your pugs? 04:27
mugwump the cursor ends up in the wrong place 04:29
and after hitting enter, the output starts on the wrong line
TreyHarris oh... what OS are you running under? 04:30
mugwump Ubuntu Dapper Drake 04:31
TreyHarris on os x, both ppc and intel, it works fine... i wonder what the bug is 04:32
mugwump sure, blame the platform 04:35
TreyHarris *grin* no, i'm just making an empirical observation :) 04:38
mugwump ghci seems to be fine 04:39
04:45 snowstalker joined
TimToady looks like both of the infinite loops are dispatches going to the multi with a slurpy when it was expected to go to the other case. 04:45
though in the URI case the "other" one is a single scalar parameter, while in quicksort it's the null parameter list that isn't getting called. 04:47
04:52 baest joined 04:55 nothingmuch joined
putter bitesized project idea: something which takes test output and produces a compact (ie, uses ranges) force_todo list. might even be a one-liner. 05:11
05:28 xinming joined
masak TimToday: do you have some thoughts on audrey's "Motivation for /<alpha>+/ set Array not Match?"? nntp.perl.org/group/perl.perl6.language/26284 05:28
I like it, and I'd feel sorry if it was really a good idea but got warnocked due to other things taking precedence 05:29
05:35 DHGE joined 05:38 MacVince joined
MacVince ?eval "foo" x 2 x 2 05:38
evalbot_r13797 "foofoofoofoo"
MacVince ?eval "foo" x 2 * 2 05:39
evalbot_r13797 0.0
MacVince ?eval "foo" x (2 * 2)
evalbot_r13797 "foofoofoofoo"
svnbot6 r13798 | putter++ | t/regex/from_perl6_rules/ 05:44
r13798 | putter++ | Revised the 15 recently undeleted test files.
r13798 | putter++ | They should now all be spec (hopefully), be proper test files, and pass smoke (untested!).
r13798 | putter++ | Added many eval_*'s, and force_todo's.
r13798 | putter++ | The "disable entire file with a comment" kludges are all gone.
r13798 | putter++ | The tests correctly indicate a great many missing features.
r13798 | putter++ | propcharset.t and properties.t take a while - about 10 minutes each on my machine.
r13798 | putter++ | It would be nice for someone to raise the other test files in the directory to the same standard. capture.t and subrule.t still have "silently eliminate lots of tests with a comment" kludges.
putter ponders that he almost never runs with -Ofun. Usually -Oneeds-to-be-done instead. :/ 05:47
good night all &
05:51 kanru joined
obra seen audreyt 05:53
jabbot obra: audreyt was seen 14 hours 31 minutes 10 seconds ago
06:19 Eidolos joined 06:46 iblechbot joined, dduncan joined 06:48 dduncan left 07:06 _vytautas joined 07:07 vytautas joined 07:18 avarab joined 07:19 kane-xs joined 07:20 avarab is now known as avar 07:24 zakharyas joined 07:25 elmex joined 07:31 jferrero joined 07:40 Psyche^ joined, avarab joined 07:55 Psyche^ is now known as Patterner 07:57 penk joined 08:01 marmic joined 08:15 drrho joined 08:19 mdiep_ joined
nothingmuch so what did I miss? 08:23
08:39 H2S04 joined 09:11 carrumba joined, carrumba left 09:34 j0sephi joined
svnbot6 r13799 | audreyt++ | * Implement quantified method calls: $obj.*meth, $obj.+meth, $obj.?meth, 09:40
r13799 | audreyt++ | as well as quantified dynamic method calls: $obj.[+*?]$meth.
r13800 | audreyt++ | * User-defined macros now take effect even in prefix unary
r13800 | audreyt++ | and optional unary. (Previously only listop worked.)
gaal nothingmuch: a fun ride to tlv :)
svnbot6 r13801 | audreyt++ | * Move typeOfParam into Pugs.AST. 09:41
r13801 | audreyt++ | This is used in the previous commits to implement class-wide
r13801 | audreyt++ | attributes with accessors, namely "class C { my $.x }".
r13802 | audreyt++ | * Adjust possiblyApplyMacro such that we can support 09:44
r13802 | audreyt++ | non-prefix macros if needed. (May affect parsing
r13802 | audreyt++ | speed significantly; still need to benchmark before
r13802 | audreyt++ | checking that in.)
r13803 | audreyt++ | * Refactor list comprehension tests from t/junction/ into
r13803 | audreyt++ | t/syntax/list_comprehension.t.
r13804 | audreyt++ | * Fix magickal line numbers in t/magicals/. 09:48
09:48 ruoso joined
svnbot6 r13805 | audreyt++ | * Correct multi invocation data set in t/oo/methods/multi.t. 09:51
r13806 | audreyt++ | * t/oo/attributes/array.t: minor indent fix; no functional changes.
r13807 | audreyt++ | * t/oo/methods/chaining.t: 09:55
r13807 | audreyt++ | [===] ($foo, $_foo2, $_foo1)
r13807 | audreyt++ | should be written as:
r13807 | audreyt++ | [===]($foo, $_foo2, $_foo1)
r13808 | audreyt++ | * Finish triaging of t/oo/methods/. 09:58
r13809 | audreyt++ | * Now that ::?CLASS support is checked in (although it's lexical 10:07
r13809 | audreyt++ | as specced, not dyamic as the test implies), triage the text.
r13810 | audreyt++ | * Fix instance.t: compile-time errors can't be caught by try{},
r13810 | audreyt++ | so use eval'' instead. Also unTODO the now-passing parts about
r13810 | audreyt++ | default attribute values.
10:08 chris2 joined
nothingmuch gaal: you missed a fun walk =) 10:36
audreyt @tell putter uhm, I don't think I've advocated pugs not testing regex and leave all that to parrot; I was consistently advocating a shared set of spec-based tests. 10:52
lambdabot Consider it noted.
nothingmuch audreyt: wrt MO - i need to catch up with reality, $work, etc
you're free-ish now, right? 10:53
audreyt right
nothingmuch ok... so i'll try to hunt you down when i am too
audreyt @tell putter deleting the Perl6::Rules tests was more of a local cleanup on disabled tests; I didn't mean to push it through, but when it's pushed I didn't rollback it in time. I'm glad that you reanimated them. 10:54
lambdabot Consider it noted.
audreyt @tell putter to qualify "locally": I was release engineering, and "3810 failed" really doesn't help me tracking down the number of failing tests; I don't see a way to easily make the 300 failing subtests pass across two regex engines; however, if you can take care of todo/skip triaging those 3000 subtests such that they reflect the default PCR engine correctly, I'd be very happy; until then I'll still delete them locally but not push back. 11:00
lambdabot Consider it noted.
audreyt @tell putter also on macbook here, I get random segfaults after ~500 invocations to PCR with embedded perl (it used to exhibit the same thing with PGE), but putting "sleep" between them will "fix" it. I'll look into that bug also. I guess pushing the delete through without reverting it in time is inexcusable; I'm very, very sorry that I've done it, and thanks for noticing. 11:06
lambdabot Consider it noted.
audreyt @tell putter also, eval was seeing lexical vars just fine; it's eval_ok that couldn't see it. we can make eval_ok a macro, but that makes Test.pm's barrier of adoption by v6.pm and parrot/perl6 much, much higher. if you still think lexical_variable_in_eval.t warrants existence, maybe move it into ext/Test/ instead of on the main perl6 test suite? 11:13
lambdabot Consider it noted.
11:17 buetow joined
svnbot6 r13811 | audreyt++ | * Charnames.hs: Allow charnames.pm aliases such as \N[LF]. 11:29
r13812 | audreyt++ | * Now that t/xx-uncategorized/unicode_aliases.t is passing, we merge them 11:32
r13812 | audreyt++ | back into char_by_name.t.
r13813 | fglock++ | v6.pm - simplified 'operator' ast 11:35
r13814 | audreyt++ | * TODO out object_id.t. 11:38
r13815 | audreyt++ | * golfex.t: (=<>).reverse is now the same as =<>.reverse, so we 11:47
r13815 | audreyt++ | must qualify it as list(=<>).reverse to get the list context reverse.
11:57 jsiracusa joined 11:59 fglock joined 12:03 weinig joined
svnbot6 r13816 | audreyt++ | * Add List::reverse to handle the [1,2,3].reverse case. 12:06
r13817 | audreyt++ | * Pugs.Eval.Var: Slury scalar in param lists should count as 12:09
r13817 | audreyt++ | one required argument level.
12:12 Limbic_Region joined
svnbot6 r13818 | audreyt++ | * Fix expected outut of head-neck-tail in examples/: 12:12
r13818 | audreyt++ | [1,2,3].say should print "1 2 3\n", not "123\n".
Limbic_Region salutations all 12:13
12:15 frankg joined
svnbot6 r13819 | audreyt++ | * catch_type_cast_mismatch.t: %ordered_hash[0] should be a legit use 12:18
r13819 | audreyt++ | to get the first pair from an ordered hash, I think...
r13820 | audreyt++ | * pair.t: TODOize.
12:21 fglock joined, jsiracusa left
svnbot6 r13821 | audreyt++ | * Triage of t/data_types/ finished... 12:33
audreyt 44 to go... 12:38
fglock clkao: re PC::Runtime own version of insideout - Class::Insideout methods were in the top-ten profiler times; Match performance is very critical for v6.pm
12:39 nothingmuch joined
Limbic_Region audreyt - would you mind expounding on the 44 to go? Do you have a specific set of tests you want to get passing before the next release and all other failing tests can be todo'd? 12:44
fglock does ' $a=(1,2,3) ' makes a capture operation? (captures the list to the first positional) 12:45
since $a is supposed to not be a reference 12:46
ok, I think I figured it out 12:52
12:54 mj41_ joined 12:58 count_leto joined
svnbot6 r13822 | fglock++ | pX/v6 - moved tests to t/00-compile-time/ 13:03
audreyt Limbic_Region: 44 files, not 44 subtests 13:09
Limbic_Region: that's all the failing ones modullo t/regex/
fglock: no, $a is simply a Seq object autovivifed into an Array 13:10
the thing doing the vivify is the non-listop infix:<=>
fglock audreyt: what is the S\d\d for Seq? 13:12
audreyt 06, but 02/03/04 too 13:13
Seq is what was known as Tuple
nothingmuch: I'm afraid I need to sleep -- maybe some 20 hours from now, or the day after 13:14
fglock right - thanks!
nothingmuch it'll take me as long 13:15
maybe more
Limbic_Region audreyt - got it 13:18
Limbic_Region begins a smoke 13:19
audreyt er, please ^C it :) 13:20
Limbic_Region k
more test triaging on its way?
audreyt no, i accidentally checked in a bogus fix 13:21
r13823 should be good to smoke now
svnbot6 r13823 | audreyt++ | * oops, the slurpy-param count fix accidentally broke named args.
Limbic_Region oh, ok
fwiw, I think triage is being used incorrectly - it means to sort not to fix ;-) 13:22
audreyt it's to sort into todo and tofix :)
Limbic_Region true enough
beginning smoke 13:23
hrm, I wonder if bin sorting could benefit from triage sorting 13:24
13:29 vel joined 13:30 agentzh joined
agentzh @messages 13:30
lambdabot audreyt said 3d 4h 36m 17s ago: all of operators/quoting.t now passes except for pair inside <<>> -- I don't feel comfortable with that feature, so will discuss more beore impl
lanny said 2d 23h 12m 17s ago: Smoke Syns are acting up again. :/
svnbot6 r13824 | fglock++ | pX/v6 added HACKING.CompileTime (still empty) 13:33
fglock does this makes sense (new plan for v6.pm emitter): The compile-time system builds a Program object. ' Program.print ' emits the compiled code. 13:35
svnbot6 r13825 | fglock++ | pX/v6 - updated HACKING.CompileTime
___particle___ why not Program.emit()? 13:36
by whatever name, building an object and calling a method on it seems sane
PerlJam fglock: makes sense to me
___particle___ fglock: do you call methods on the intermediate trees to transform them? 13:37
fglock ___particle___: because ' print Program ' would work too; and ' print Program.perl ' would show it's internal data
13:38 ___particle___ is now known as [particle]
fglock ___particle___: yes, that's the idea 13:38
[particle] ok, wondering if you could do Program.emit('yaml') or something
Program.yaml.print?
fglock particle: example: svn.openfoundry.org/pugs/misc/pX/Co.../01-bool.t 13:39
lambdabot tinyurl.com/eo3jj
fglock yes, Program.yaml.print
13:40 avar joined
fglock at this time, Program is backend-dependent 13:41
[particle] i like this syntax.
what is $b->_61__61_() ?
fglock it is the ascii representation of '==' 13:42
[particle] ah, i see. the comment didn't make sense until now
13:46 weinig is now known as weinig|away, cjeris joined 13:47 BooK joined
fglock [particle]: in order to emit Parrot, one would write the Compile-Time classes in pir 13:50
[particle] i'd assume a perl5 module 13:51
13:51 BooK joined
[particle] Pugs::Emitter::Perl6::PIR::* 13:51
fglock or a v6-alpha module :) 13:52
[particle] what's the diff between perl5/PCP6 and misc/pX/PCP6 ? 13:55
fglock pX is experimental - if it works, it can be merged back 13:56
[particle] ah, ok. so perl5 is stable, misc is head
fglock more like: perl5 is unstable, misc/pX is throw-away :) 13:57
cpan is stable 13:58
[particle] ok, you got me there :)
14:00 buetow joined
[particle] any docs other than source on what methods a PEP6::PIR would have to implement? 14:00
fglock sorry, not yet - HACKING.CompileTime will be it 14:02
[particle]: it would be nice if you could add some files to misc/pX/PCP6/lib/Pugs/Emitter/PIR - I could help you work it out 14:05
[particle] i've just started that
fglock woot
[particle] i figure i'll copy 01-bool.t 14:06
to pir-01-bool.t
fglock how about a subdir t/01-pir/01-bool.t - so we could mirror the p5 stuff into it 14:07
[particle] sure, great. i wondered how you wanted to do it. t/pir/01-bool.t 14:08
fglock it's nice that it doesn't depend on the AST format, so we can even use PIL as input (not just v6) 14:09
[particle] hrmm, could we eventually use PASM?
err PAST
fglock sure - you just have to write a tree walker to build the output object 14:10
Limbic_Region audreyt - 96.59% on my Win32 14:11
fglock [particle]: then you can have PAST->perl5 too :) 14:12
[particle] precisely! 14:13
fglock what's a good name for the 'objectification' operation? such as int->Int 14:14
rgs boxing ? 14:15
fglock 'enbox'? :)
rgs "embox" reads better
fglock rgs: thanks!
[particle] why not box/unbox 14:21
14:21 lanny joined
fglock [particle]: sure 14:22
lanny @tell TreyHarris role Pregnancy; class Mammal does Pregnancy; class Platypus is Mammal hides Pregnancy; ... S12 doesn't make clear if you can hide roles though.
lambdabot Consider it noted.
svnbot6 r13826 | fglock++ | pX/v6 - updated HACKING.CompileTime 14:23
fglock [particle]: i added some of the method names
[particle] great! 14:24
audreyt Limbic_Region: uploaded the smoke?
svnbot6 r13827 | audreyt++ | * Further tweaks to the slurpy-param model. The quicksort 14:25
r13827 | audreyt++ | example was broken all along, and this commit removes the
r13827 | audreyt++ | previous made-it-work kluge:
r13827 | audreyt++ | multi quicksort () { () }
r13827 | audreyt++ | multi quicksort (*$x, *@xs) { ... }
r13827 | audreyt++ | my @x;
r13827 | audreyt++ | quicksort(@x); # this should not have succeeded!
r13827 | audreyt++ | It was binding to variant #1, but there's really nothing
r13827 | audreyt++ | there that would cause it to bind that way. I think rewriting
r13827 | audreyt++ | the example to "quicksort(|@x)" is the sanest approach, but
r13827 | audreyt++ | that'll have to wait till tomorrow as |@x doesn't work at this
r13827 | audreyt++ | moment...
14:25 ofer1 joined 14:27 penk joined
lanny quicksort(|@x) isn't going to be very intuitive syntax for sorting some type of @thing. 14:27
audreyt another fix is 14:28
lanny audreyt++
audreyt multi quicksort (*[]) {...}
but I'm not sure at all that syntax works, or is specced
PerlJam it's easily inferred from the existing spec :) 14:29
Though it could be that *[...] could be considered an error (in judgement at least ;-)
audreyt perhaps, but there's no mention that slurpy arrays are unpackable
would make some sense though. 14:30
lanny The workaround is to check $x in the second multi declaration and return if it's unset but I know that's not the point of multi
audreyt exactly.
|@x will work Correctly
gaal didn't multis-as-guards need a rethink anyway?
audreyt but as you said it wasn't very intuitive. 14:31
svnbot6 r13828 | audreyt++ | * unTODO the now-passing slurpy_param.t and oo/attributes/class.t.
gaal (hello!)
audreyt hi!
PerlJam still isn't used to the unary |
the |-as-junction has become quite hardwired
svnbot6 r13829 | audreyt++ | * next.t: fix plan number. 14:32
audreyt PerlJam: & is a precedent though
actually, so is ^
all junctiona infix has an absolutely unrelated prefix meaning
PerlJam yes, I don't know why my brain is having trouble with unary |
fglock lunch & 14:33
svnbot6 r13830 | fglock++ | pX/v6 - updated HACKING.CompileTime
audreyt gaal: how's life/work/tuits these days?
lanny slurpy empty seems to handle quicksort() as well. and if you are used to haskell... :) 14:34
audreyt yeah. but the current OldParam has no way to express it 14:35
newland SigParam can express it quite elegantly though
gaal audreyt: new project starting at $job tomorrow, so I don't know exactly 14:36
lanny Anyone have any idea what "applying a Signature to a value would look like and extracting the positional parameters"? Needed for sort and the signature will be held in a Signature variable. 14:37
gaal audreyt: you're itching for a release I take it...
audreyt parsefail 14:38
PerlJam lanny: I don't understand "applying a signature to a value" 14:39
lanny PerlJam: Neither do I. :) It's in the spec for sort() in S29. Signature got added to SortCriterion so you can do stuff like: sort :(Num, Str), @values; and sort will write the KeyExtractor and Comparator for you. 14:40
audreyt gaal: that is so :)
see y'all tomorrow :) *waves and goes sleep* 14:41
gaal audreyt: night! next two days are probably not free but hopegully the weekend (unless hiking w/nothingmuch)
PerlJam lanny: then it sounds like it'll treat each value as a Capture in a way
lanny Or worse sort :(Num but canonicalized(&abs), Str), @values; 14:42
Agreed.
I woudl accuse TimToady of having a fevered imagination on this kind of thing but everyone already knows that's true. 14:44
Limbic_Region yes audreyt - smoke uploaded 14:47
about to run another 14:49
14:56 etzel joined 15:06 hexmode joined
lanny Is .yaml what the AST looks like of a given expression? 15:19
15:19 jferrero joined
lanny er... given statement. 15:19
kolibrie lanny: .yaml and .perl are representations of a data structure in those serialization formats 15:25
lanny Thanks, Kolibrie. Finally realized what I wanted was interactive pugs to look at the syntax tree 15:26
gaal lanny: another way is Pugs::Internals::emit_yaml, but that's very verbose 15:30
clkao fglock: oh ok. why is it using inside out at all? ;) 15:31
15:31 buetow joined
audreyt clkao: when you overload ${} @{} %{}, what else can you do? 15:36
I guess you can bless CVs. that feels very slow though 15:37
15:37 chris2 joined
lanny I wonder if audreyt will be the first recognized medical case of "Sleep Developing" 15:37
svnbot6 r13831 | lanny++ | * split.t - fixed bug in test wrapper that was causing failing tests 15:41
clkao audreyt: heh 15:43
fglock: so i m tking the bits to mke data::capture. but i will be using refs for the values. do you think it makes sense for PC::Match?
not having a full copy of matched things might be a good idea - but the thing in ::match are not ref'ing other things i guess... 15:44
fglock clkao: hi 15:50
svnbot6 r13832 | lanny++ | * readline_chomped.t -- todo on 'is chomped' working
clkao fglock: hi
fglock insideout is being used so that we can overload all of $ % @ 15:51
clkao *nod*
fglock the reason to use ref-ref-thing is so that we can use ref-variables as cheap accessors 15:52
clkao uhm ? are you using ref-ref?
fglock like in: $match = { thing => \$thing }; $$thing = 42; # sets $match->thing 15:53
it's a bit cheaper than method calls 15:54
actually: $match = { thing => \$thing }; $thing = 42; # sets $match->thing
clkao oh y, that's good and consistent with what i am doing anyway 15:55
do you do tht for positional as well ?
and i noticed your keys also have the array index
fglock yes: $match = { array => \@array, hash \%hash }; 15:56
yes, that's specced
clkao no, i mean the values of the array and hash 15:57
{ array => [\1, \'foo'], hash => { key => \'value' } }
need that for data::bind
fglock sorry? I mean as in S05: "The numbered captures may be treated as named, so $<0 1 2> is equivalent to $/[0,1,2]" 15:58
Captures don't need that
clkao oh ok. 15:59
only in match.
16:03 Psyche^ joined
fglock though in some places you turn positional arguments into named arguments, and named into positionals 16:05
clkao ok. but what i cared more is the value of the named. because i need to have them as ref in capture 16:07
16:11 Psyche^ is now known as Patterner
clkao fglock: so what happens if you feed a match object as a capture to a subroutine? do you alsao have the index keys? 16:18
fglock clkao: you mean when used as a hash? yes, it was implemented like this 16:22
lanny ?eval my @ps = 1=>'a',2=>'b'; @ps.fmt("%d->%s", "\n") 16:23
16:23 evalbot_r13797 is now known as evalbot_r13832
evalbot_r13832 "1->a\n2->b" 16:23
16:24 iblechbot joined
clkao fglock: uhm, do you mean about the value ref? 16:24
fglock $($match) is whatever you set it to - it doesn't depend on @($match) or %($match) 16:25
lanny Why does that work? fmt(List $obj, Str $fmt, Str $comma) { join($comma, map -> $v { sprintf($v, $fmt); }, @$obj; }
And sprintf doesn't understand :(Pair, Str) 16:26
16:26 putter joined
lanny oops. Reverse the parameters on sprintf() above. 16:27
fglock lanny: maybe it is handling the pairs in Array context? 16:28
lanny They seem to be Pair objects as far as I can trace them. 16:29
16:29 lollan joined
fglock ?eval @( 1 => 'a' ) 16:29
evalbot_r13832 (1 => "a")
fglock ?eval @( 1 => 'a' ,)
evalbot_r13832 [(1 => "a"),]
lanny What it breaks is this: 16:30
?eval (1=>'a').fmt("%d->%s","\n")
evalbot_r13832 Error: Insufficient arguments to sprintf
fglock ?eval (1=>'a').kv.fmt("%d->%s","\n") 16:31
evalbot_r13832 Error: Insufficient arguments to sprintf
fglock ?eval (1=>'a').kv
evalbot_r13832 (1, "a")
lanny Yeah. Fix is something like map -> $v { $v~~Pair ?? sprintf($fmt, $v.kv) !! sprintf($fmt, $v) 16:32
putter audreyt: re lexical_variable_in_eval.t, I'll add a warning to Test's pod. and look at the tests. :/
problem is failing regex tests produce matches on which hash access is deadly.
re eval_ok, I'm certainly never going to advocate Test.pm use even more language features ;) 16:34
though basic text macros aren't actually _that_ daunting. compared to say junctions.
bah, wrong channel again ;) 16:36
16:36 araujo joined
putter fglock: does v6 actually use the "real" Test.pm? 16:36
audreyt: if so, it may be the first non-hs impl which did. pilrun pil2js et al used tailored simplified ones.
though my fuzzy recollection is Test dropped its junctions at parrot's request, so maybe they did too. 16:37
fglock putter: yes - it uses a copy that is in the /t directory - you can use perl -It -Ilib ... in order to include it
re hash access is deadly - this is fixable 16:39
16:39 rindolf joined
rindolf Hi all. 16:39
16:39 avar joined
rindolf Hi avar 16:39
putter modified copy, or identical to ext/Test/lib/Test.pm? I've wondered off and on if we should create a simplified Test for bootstrappers. but each time it is easier to just do another custom one, and "well, no one will be doing another bootstrap any time soon". ;) 16:40
rindolf ingy: here?
16:40 avar joined
fglock putter: it is a plain copy 16:40
putter re fixable, yes?!? :) problem is $m = ... ~~ /.../; $m<foo><bar> goes boom trying to make the $m<foo> exist. 16:41
rindolf Hmmm... avar is bouncing.
Hi fglock
putter re plain copy, neat :). another infinite todo list item bites the dust. yay 16:42
err, re goes boom, "when the match fails".
fglock rindolf: hi
svnbot6 r13833 | fglock++ | pX/v6 - HACKING.CompileTime - added more Classes, Methods; special Classes 16:43
fglock putter: would you add a test to PCR?
clkao fglock: no, i meant values of the array and the hash 16:45
putter @tell audreyt "shared set of spec-based tests"++. second regex test deletion I've noticed lately, so was wondering if there was a pattern. none- great.
lanny aha
fglock putter: re Test.pm - we should start using Versioning (as in S11)
lanny ?eval my @p = (1=>'a'); @p
evalbot_r13832 [(1 => "a"),]
lanny ?eval my @p = (1=>'a'); map -> $v { $v.WHAT }, @p 16:46
16:46 evalbot_r13832 is now known as evalbot_r13833
evalbot_r13833 (::Int, ::Str) 16:46
lambdabot Consider it noted.
lanny A singe Pair gets stripped down to it's .key and .value for some reason.
16:46 avar joined
lanny ?eval @p = 1=>'a',2=>'b'; map -> $v { $v.WHAT }, @p 16:47
evalbot_r13833 Error: Undeclared variable: "@p"
lanny ?eval my @p = 1=>'a',2=>'b'; map -> $v { $v.WHAT }, @p
evalbot_r13833 (::Pair, ::Pair)
lanny Time to make a test. 16:48
fglock clkao: it depends on which context the Match will find inside the subroutine - if it gets used as a Hash, it will show positionals as keyed by number
clkao no, i don't mean the keys
the values.
in capture they should be references
ingy hi clkao 16:49
lambdabot ingy: You have 1 new message. '/msg lambdabot @messages' to read it.
putter audreyt: (re "advocated"... ah well, searching the irc log without better tools is hopeless. it was somehing vaguely like "we're debugging pge second hand; create unified tests?; let's just give them all to parrot!". I countered with "what about my regex stuff in prelude and alternate regex engines?".) 16:50
fglock clkao: would you give an example? 16:51
putter fglock: re versioning, intriguing idea
fglock: re adding tests to PCR, I think it's more of a pugs-side issue. and quite possibly the correct behavior, regardless of how inconvenient for regex tests ;) 16:53
?eval my $m = "" ~~ /a/; defined($m<foo><b>)
evalbot_r13833 *** Cannot parse regex: a *** Error: Error: 'require' trapped by operation mask at /home/audreyt/pugs/perl5/Pugs-Compiler-Rule/lib/Pugs/Runtime/Match/HsBridge.pm line 23. Error: Can't modify constant item: VUndef
fglock ?eval my @p = (1=>'a',); map -> $v { $v.WHAT }, @p
evalbot_r13833 (::Int, ::Str)
putter ok, so $m has a match failure other than the one intended, but it's the Error: Can't modify constant item: VUndef which bites. 16:54
@tell audreyt re lots of failing&toxic tests unhelpful... yeah, no-one has yet done the next-gen test infrastructure wanted long ago. ... 16:56
lambdabot Consider it noted.
putter @tell audreyt ...our test failure firewall is the test file. we directly modify test cases to protect the test file, and encode expectations. alternate backends, or here component of backend, have different fragilities and expectations. they basically have to live with doing their own fragility workarounds, working around pugs' to see the tests, and ignoring expectations. 16:57
lambdabot Consider it noted.
putter the wild and wacky world of alternate backend testing
maybe when we start doing generative testing we can do a "firewall which is not a file" infrastructure.
anyone looking for a project? :) 16:58
lanny fglock: the loss of pairness probably tied up with pairs as named parameters (somehow) 17:03
rindolf Does anyone know when the slides for YAPC::Europe will be available? 17:05
The Birmingham YAPC::EU that is.
clkao fglock: 11:57 < clkao> { array => [\1, \'foo'], hash => { key => \'value' } } 17:06
putter re testing problem... silly me... "If you need context, use a normal C<eval> inside a C<ok> or C<is> or C<dies_ok>." 17:09
17:11 justatheory joined
svnbot6 r13834 | fglock++ | pX/v6 - added Native Class (int/num/str/bool) 17:11
fglock clkao: yes, in that case, the array values would appear twice, unless we add a Match.capture coercing method that returned a proper formatted Capture 17:12
clkao fglock: i am not talking about array in hash anymore... i am asking about using refs in the array and hash values, as i need to do so in capture
17:16 count_leto joined
fglock clkao: what's the question again? (sorry) 17:16
17:16 hexmode joined
clkao fglock: are you using refs or just values 17:17
in ::Match
fglock clkao: one sec - checking 17:18
17:18 avarab joined
svnbot6 r13835 | putter++ | [ext/Test/lib/Test.pm] 17:18
r13835 | putter++ | Noted in the pod that eval_{ok,is,etc} do not eval() in the context of the caller. When context is needed, ok(eval()) should be used.
putter audreyt: re segfault, segfaulting after 3 is easier to debug then after 500. But I'm running an old 6.4.1, on 64bit. so the causes may be different. how much of a priority do you think this? 17:19
fglock clkao: Match array/hashes store either Match (which is blessed ref) or ARRAY (which is ref) 17:21
clkao: never just values 17:22
clkao oh cool. 17:24
svnbot6 r13836 | putter++ | [xx-uncategorized/lexical_variable_in_eval.t]
r13836 | putter++ | Deleted. Test.pm's pod now documents that eval_ok() does not eval() in the context of the caller.
clkao anyway, i will make my data::capture and we can see if that can be superclass of match
fglock clkao: it can be cheated to be a superclass, even if the code is not reused 17:26
svnbot6 r13837 | lanny++ | [t/xx-uncategorized/pair_in_array.t] 17:27
r13837 | lanny++ | * demonstrate bug where (::Pair) becomes (.key, .value)
17:29 kanru joined
rindolf Hi clkao 17:30
svnbot6 r13838 | lanny++ | * fmt.t - todo :depends<list of single pair> on failing test 17:33
17:38 Ikarus joined, vel joined
svnbot6 r13839 | putter++ | [t/regex/from_perl6_rules/] 17:48
r13839 | putter++ | Instances of eval_ok('something which depends on caller context',...) changed to ok(eval('...'),...). Also eval_is().
putter yipes, there are 200+ plus tests which _may_ be hitting the same eval_ok isn't eval problem...
gaal putter: iirc using context vars can help with that 17:54
fglock yay! # pX/v6 emitter does ' 1+1 ' 18:00
svnbot6 r13840 | fglock++ | pX/v6 - passes minimal emitter tests
putter ok, topic for discussion: 18:04
almost all the uses of eval_ok and eval_is in t/ are incorrect. that is, the vast majority depend upon seeing non-global lexicals, or being in the same package. 18:05
oops, that's not quite true, 18:07
1/3 of eval_ uses don't have a [$@%] in the line. thus likely don't depend on lexicals, but may still depend on package - haven't checked. 18:09
of the remaining 2/3, about 4/5 are broken. 18:10
fglock very simple test for the new emitter: svn.openfoundry.org/pugs/misc/pX/Co...e/02-int.t 18:11
lambdabot tinyurl.com/h26g6
fglock coffee &
putter the ok(eval()) is(eval()) idiom outnumbers the use of eval_ok/eval_is. 18:13
strawman proposal: remove eval_ok/eval_is/eval_dies_ok from Test.pm. it is not what people think it is. it is pervasively being used incorrectly. the alternatives are already in widespread use. 18:17
comments? thoughts?
gaal ?eval my $x is context; eval '$x++'; $x 18:21
18:21 evalbot_r13833 is now known as evalbot_r13840
evalbot_r13840 \1 18:21
gaal ?eval my $x; eval '$x++'; $x 18:22
evalbot_r13840 \1
gaal heh
hm, putter, remoose me, the problem is that the eval takes place deep in Test, right? 18:23
putter yes
lol (re remoose)
TimToady ?eval my $x is context; eval '$+x++'; $x
evalbot_r13840 \1
gaal ?eval my $x is context; Moo::evinc('$x++'); say $x; class Moo { sub evinc ($s) { eval $s } } 18:24
evalbot_r13840 OUTPUT[1 ] undef
TimToady still in lexical scope of "my"
gaal ?eval { my $x is context; Moo::evinc('$x++'); say $x;} class Moo { sub evinc ($s) { eval $s } }
evalbot_r13840 Error: Unexpected "class" expecting comment, operator, statement modifier, ";" or end of input 18:25
gaal ?eval { my $x is context; Moo::evinc('$x++'); say $x;}; class Moo { sub evinc ($s) { eval $s } }
evalbot_r13840 OUTPUT[ ] undef
putter a couple of eval_* uses in ext/, 300-ish in t/, 200-ish of which are obviously broken. roughly. 400-ish instances of {ok,is}(eval()).
gaal ?eval { my $x is context; Moo::evinc('$+x++'); say $x;}; class Moo { sub evinc ($s) { eval $s } }
evalbot_r13840 OUTPUT[1 ] undef
TimToady yes
gaal whew.
putter: does this look like an approach for a fix?
I mean, there's obviously many places to apply it - but is this an acceptable solution IYO? 18:26
TimToady removing eval_ok seems simpler
gaal or making it a macro (is there a way we can wing that?)
TimToady it's not like ok eval is all that much longer
*cough* source filters *cough* 18:27
gaal a var has to be declared contextual, right?
putter "source filters"++. but here, eval_ok() is so little improvement over ok(eval()), that it doesnt seem worth any pain. 18:28
gaal: I'm not sure I understand your idea. Use "is context" to pass lexicals, and... how does the "eval() happens in the wrong package" get handled? 18:29
and is the idea dealing with the "people get it wrong", or just with "once they realize they have gotten it wrong, they can work around it by doing X". 18:30
?
gaal the idea is to make it convenient to write tests :) 18:31
if ok eval is good enough, great. I can't believe I started a sentence with four consecutive Perl 6 keywords. 18:32
18:32 Yappoo joined
putter lol 18:32
gaal (I didn't really though)
putter foresees a vast vista of executable poetry in p6... 18:33
gaal perl -pi -e 's/eval_ok/eval ok/g' will probably break things though
phone, bbl
putter re -pi -e, yeah, a bit more work ;) 18:34
18:34 larsen joined
putter hmm... instead of the usual arity parser rules, when you use Haiku;... 18:35
18:36 count_leto left
fglock re backends, macros are actually easier to implement than most other tricks 18:36
putter that's my impression as well. 18:37
err, wait.
fglock putter: would you review the pX/v6 emitter idea?
putter phone bbl
18:38 vel joined
putter back 18:40
re "easier", um, what do backends have to do for macros? frontends do most all the work, no? 18:42
fglock right - it is easier because it is not backend-specific 18:43
putter so while my only attempt at a frontend briefly had string macros, the whole parser was a kludge, so I can only hypothesize how difficult they would be to integrate with something more sane. 18:45
re right...
but v6, doing it's own parsing, would have to deal.
or am I missing something? 18:46
18:47 weinig joined
fglock putter: v6 uses something very close to macros internally: svn.openfoundry.org/pugs/perl5/Pugs...Control.pm 18:47
lambdabot tinyurl.com/kvn9s
18:48 weinig is now known as weinig|away
putter re emitter idea, 02-int.t ? 18:49
fglock yes, that's an example
putter re "very close to macros", indeed. so the question is how much effort it would be to get to the point that Test could safely use macros to implement eval_is. that I am unsure of. while thinking about Parser.hs and other attempts, I don't see a problem, but I've also been burned enough around here to have a "beware gotchas contributing orders-of-magnitude of pain" hesitancy. 18:53
re 02-int.t, hmm... so emit() is figuring out what kind of node it is holding by checking for the existance of some fields? 18:56
fglock the most likely problem is that we don't have a common AST that could be use inside macros
putter so not a problem for string macros. :) :) and that's all which is needed.
s/so/soooo/
fglock putter: emit() is creating compile-time objects (this ast format is just an example) 18:57
parsing into objects would be possible too, but that's not the point here 18:58
putter the usual compiler ast story is you grovel over the ast accumulating non-local information, mutate the ast, and then emit. the usual lossage is that your mutation was lossy, and you find pass 13 needs information discarded in pass 3. "non-lossy ast modification"++. 18:59
could you summarize "the point here"? 19:00
i guess i am unsure what aspects to comment on 19:01
fglock you have a syntax tree to walk - you walk the tree creating objects that represent the current program; then you stringify the object and you get source code
putter my first reaction was "why aren't the ast nodes typed, and emit a multimethod (or at least doing type based dispatch)". but I don't know if that's the kind of thing you are looking for... 19:02
fglock the ast nodes belong to classes that represent the Perl6 types - int, str, Array, ... 19:03
19:03 ashleyb joined
ajs_work ?eval $_=1; say "{$_} {$^a}" for 1..4; 19:03
evalbot_r13840 OUTPUT[1 2 3 4 ] undef
ajs_work Is that intentional, or should the $^a prevent the $_ from sucking/creating a positional parameter? 19:04
fglock ajs_work: you mean $_[0] ?
ajs_work fglock: I don't think so, no. 19:05
fglock ?eval my $out; $_=1; $out .= "{$_} {$^a}" for 1..4; $out
evalbot_r13840 Error: Unexpected "\"{$" expecting ".", operator name, qualified identifier, variable name, "...", "++", "--", array subscript, hash subscript or code subscript
fglock ?eval my $out; $_=1; $out ~= "{$_} {$^a}" for 1..4; $out 19:06
evalbot_r13840 \"1 23 4"
ajs_work You see, it's treating $_ and $^a as if they were $^a $^b
which obscures the old $_
and also implies :by<2> which isn't really what I think I wanted there 19:07
fglock putter: the way it works is not too different from what you said 19:08
?eval my $out; $_=1; $out ~= " _ {$_} ^ {$^a}" for 1..4; $out 19:09
evalbot_r13840 \" _ 1 \^ 2 _ 3 \^ 4"
ajs_work I was also wondering if AST nodes could be typed for multi dispatch. I really think that would be handy for differentiating certain types of circumstances in user macros. 19:12
I gave an example in the macro API doc
Of course, it's compile-time multi-dispatch which has some interesting quirks, but it's not unreasonable, I think 19:13
fglock ajs_work: link?
ajs_work docs/Perl6/API/macros.pod
search for "multi" 19:14
It's not a finished document by any stretch, but was just an attempt to see what docs/notes/p6ast_draft meant for the user API 19:15
19:15 bernhard joined
putter end of "lunch" for me. need to plan a jobsearch. really ;) 19:17
perhaps bash on eval_ok->ok(eval()) tomorrow, if no-one beats me to it.
feedback on the depreciating/disappearing eval_x continues to be welcome.
fglock: do you have a english rather than code sketch of your plan for feedback?
especially when beginning a new subproject, writing one can be useful. and helps others think about your core ideas without the challenge/flakeyness of inferring them from code...
fglock ajs_work: yes, I think it works
putter: ' HACKING.CompileTime ' in the module root 19:18
putter @tell audreyt Thoughts on the proposal to eliminate eval_ok eval_is eval_dies_ok and convert t/ to use ok(eval()) etc? 19:19
lambdabot Consider it noted.
putter fglock: awesome. that's the other thing I'll try to do tomorrow. :) 19:20
fglock putter: in order to add an eval_ok macro to v6.pm, you can do %Pugs::Grammar::Term<eval_ok> = '...rule...' # low-level access to the parser tables :) 19:21
putter lol 19:22
19:22 smash joined
svnbot6 r13841 | putter++ | [t/xx-uncategorized/namespaces.t] 19:22
r13841 | putter++ | [t/blocks/pointy.t]
r13841 | putter++ | Incorrect eval_ok()'s converted to ok(eval())'s. 2 files down, lots to go. Volunteers?
putter fglock: "I promise to use this power only for good..." :) 19:23
those were the two files I converted before noticing the scale of the problem, and bumping it to discussion.
changing the rest really isnt a big deal. the right pugs/anarchistic/just-go-for-it model would likely be to right now quickly convert the rest of the t/ files. but I'm out of time for today. feel free. 19:25
have fun & 19:26
19:31 justatheory joined
mugwump fglock: isn't that written as macro eval_ok is parsed /...rule.../ ? 19:32
fglock mugwump: yes - the 'Term' thing is just an example of how it could be implemented 19:46
19:46 BooK_ joined 19:51 hexmode joined
holoway /window 12 19:52
gack
silly spaces
20:01 fglock joined
fglock putter: re "the point here" - the "$ast" and "emit" in 02-int.t are just an example, they are not part of the project. 20:03
putter: the project is about emitting code from an expression like ' 1.int.infix:<+>( 1.int ) '
s/expression/perl6 program/ 20:04
20:11 hexmode joined 20:13 hexmode left 20:18 weinig|away is now known as weinig 20:19 mdiep joined 20:31 jferrero joined
fglock ?eval my $a; my $b=1; 'scalar ' ~ $a.WHAT ~ ' int ' ~ $b.WHAT 20:35
20:35 evalbot_r13840 is now known as evalbot_r13841
evalbot_r13841 "scalar Scalar int Int" 20:35
20:38 larsen joined 21:02 justatheory joined 21:03 theorbtwo joined
svnbot6 r13842 | fglock++ | pX/v6 - added TODO.CompileTime; several fixes 21:06
21:17 fglock left, avarab is now known as avar 21:26 Limbic_Region joined 21:45 weinig joined 22:06 weinig is now known as weinig|bbl 22:08 cjeris left 22:11 mauke joined 22:24 spoop joined 22:26 pfenwick joined 22:27 mako132_ joined 22:31 spoop joined 22:33 weinig joined 22:51 polettix joined 22:57 frankg joined 23:14 Aankhen`` joined 23:17 avarab joined 23:24 mako132_ joined 23:36 Aankhen`` joined
svnbot6 r13843 | lanny++ | * t/unspecced/sort.t - remove eval_ok 23:49

Powered by Cro and the Raku® Programming Language.

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

Thank you!