Pugs t-shirts www.cafepress.com/pugscode | Pugs 6.2.9 released | pugscode.org | pugs.kwiki.org | paste: sial.org/pbot/perl6 | www.geeksunite.net
Set by stevan on 15 August 2005.
rjbs Does Perl 6 have a hashy hash slice notation? That is, return a list of pairs for the given keys? 01:17
svnbot6 r6528 | fglock++ | * perl5/ Array 02:33
r6528 | fglock++ | - Fixed fetch/store single elements from a lazy slice
r6528 | fglock++ | - Fixed fetch/store of whole lazy slice -
r6528 | fglock++ | supports syntax:
r6528 | fglock++ | @a = (0..Inf); @a[1,10,100..10000] = @a[500..30000]
r6528 | fglock++ | (needs optimization)
r6528 | fglock++ | - New method tied()
r6528 | fglock++ | - Fixed binding of fetched result
r6528 | fglock++ | - Fixed stringification of unboxed values
r6528 | fglock++ | - New parameter 'max' in perl() and str() methods -
r6528 | fglock++ | controls how many elements of the lazy array are
r6528 | fglock++ | stringified.
r6528 | fglock++ | - Array is stringified using parenthesis.
r6528 | fglock++ | * perl5/ List
r6528 | fglock++ | - new methods start() end()
luqui ?eval (a => 1).keys 05:30
evalbot_6528 ('a')
luqui ?eval (a => 1)<a>
evalbot_6528 \1
luqui ?eval (a => 1)<a> = 2
evalbot_6528 Error: Can't modify constant item: VUndef
luqui ?eval my $x = 1; (a => $x)<a>
evalbot_6528 \\1
luqui ?eval my $x = 1; (a => $x)<a> = 4
evalbot_6528 Error: Can't modify constant item: VUndef
ods15 odd 05:48
whats with the <>, why the change 05:49
revdiablo {} subscripts don't auto-quote 05:51
? eval (a => 1){'a'} 05:52
?eval (a => 1){'a'}
evalbot_6528 \1
revdiablo without those quotes, it would take it as code.
?eval (a => 1){a}
evalbot_6528 Error: No compatible subroutine found: "&a"
revdiablo so there's <> that does quoting 05:53
ods15 yeah that was a bit inconsistent in perl5
like, if '{a}' work, and '{$bla}', what does '{func()}' do...
or '{$a b}' 05:54
revdiablo so yeah, perl6 disambiguates the situation
luqui Ods, {$bla} does not autoquote 05:55
the only thing that autoquotes in p5 is /[\w-]+/
but perl6 autoquotes nothing, which is nice in a way
(though it's a situation that I resisted for a while)
ods15 heh 05:56
luqui does anybody here know anything about PIL2
?
revdiablo ?eval (a => 1)«a» 05:57
evalbot_6528 Error: unexpected "\a" expecting term postfix, operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
revdiablo hmm, I dunno if the utf8 went through right 05:58
luqui ?eval (a => 1)<<a>> 06:07
evalbot_6528 \1
luqui I'm guessing it's a utf8 problem
svnbot6 r6529 | putter++ | lib6/Prelude/JS/Rules.pm: unfinished sketch of :perl5 regex support for PIL2JS. 06:10
luqui nothingmuch, do you know anything about PIL2? 06:25
nothingmuch no 06:26
luqui darn
nothingmuch sorry
luqui what are you working on these days?
nothingmuch work and rennovating the house
luqui ahh, so not perl 6
:-)
nothingmuch no =( 06:27
luqui too bad. well, I hope you're having fun betraying us...
luqui should talk, having the longest absences of anyone :-)
anyway, I do hope you're having fun
you're working with stevan right? 06:28
nothingmuch huh? 06:33
on what?
and no, i'm not having fun... Dust and insanity mostly
then it's also summer vacation, so the car is tied up
and I have to rely on rides to work
luqui eep 06:34
nothingmuch which sometimes don't get there
luqui why is the car tied up during summer?
nothingmuch sister needs enterntainment (summer camps all over)
luqui oh. I'm glad I have no siblings...
nothingmuch since parents don't have to teach in the summer, they can work from home 06:35
luqui that sucks though
nothingmuch so one works, while the other entertains sister
in normal times parents and I share the car to work
work needs to lease me a small car 06:36
luqui who are you working for then?
luqui guesses it's not stevan
nothingmuch www.hyperroll.com/
luqui for some reason I thought thtat it was
nothingmuch well i did do one project for stevan, that's what got me to YAPC (monetarily)
luqui ahh, that was it. it was a one-shot thing. that's what confused me 06:37
hyperroll.com looks like a nice weasel-wordy site
nothingmuch yup
weasel-wordy? 06:38
it's the "new website", since ca last year or so
some company does it for us, I think
luqui "HyperRoll delivers previously unattainable levels of speed, data volume, and *business dimensions*"
nothingmuch ah yes
luqui wtf does that mean? :-) 06:39
nothingmuch but then againa all of OLAP is weasel wordy
luqui OLAP?
nothingmuch speed means we're faster than the competition
data volume means we handle more data and take less space than the competition
and business dimensions are OLAP nonsense made possible by efficient speed and data volume
luqui heh 06:40
what is OLAP?
nothingmuch dimentionality in OLAP determines a certain slice by which you query your tables
luqui learns to google
nothingmuch i must admit i don't really know 06:41
anyway, ride issue fixed, must get ready for it
luqui good
adionara
nothingmuch morning 07:52
luqui quiet in here 07:57
nothingmuch indeed
ods15 hya nothingmuch 08:06
nothingmuch hi hop 08:07
s/p$/;
grr
ods15 yawn
nothingmuch ods15: just getting up, or got up too early? 08:08
heh, someone at work just told me that his computer now says: 08:12
"CPU not workable, press F1 to continue" during boot
when he presses F1 the machine resumes normally
nothingmuch kind of knew intel was scamming us with the gigahertz myth and all that, but I never thought you don't actually need a CPU to run windows 08:13
ods15 nothingmuch: hmm, niether really.. 08:20
i woke up at about 6pm yesterday after about 3 hours of sleep
nothingmuch odd 08:23
nothingmuch has a new idea
($by_name{$action->name} ||= Set::Object->new)->insert($action);
user defined autovivification would be nice here
my %by_name is autovivifying(sub { Set::Object->new }); 08:24
uh, perl6ify that
luqui class Foo will autovivify { Foo.new } {...}
nothingmuch uh, what does that mean? that I can say my %hash is shape (... contains a Foo ...) "? 08:25
luqui probably
nothingmuch which syn is that?
luqui how do you know what to vivify otherwise?
nothingmuch the hash has a subroutine
like in my example 08:26
or a trait saying which class to use
my %by_name will autovify(Foo);
luqui Yeah, I think my Foo %hash is the way to go there 08:27
nothingmuch i agree
what syn is that?
luqui uh
S06?
nothingmuch that's subs, doesn't make sense
luqui or S09 08:28
nothingmuch nope
luqui what are you looking for?
nothingmuch /autov/
luqui oh, autovivi
do the syns mention that even?
nothingmuch i don't find 'autov' so autovivi woon't turn up either 08:29
s 06 mentions it WRT to parameters and 'is ref'
luqui A06 is the only place that autovivi is mentioned too 08:30
maybe it hasn't been covered
ods15 nothingmuch: whats odd? my sleeping pattern?
nothingmuch ods15: yes
ods15 army guarding isn't very permissive regarding sleeping :/
nothingmuch i thought the army makes your sleeping pattern more rigid
oh, i see... =) 08:31
ods15 usualy yes, except when you guard
nothingmuch right
what is the effect of : sub foo (*@args) { ... }; foo(gather { ... });
does it delazify?
luqui certainly not 08:32
nothingmuch good
luqui *@ is allowed to be lazy
nothingmuch **@ flattens?
luqui er, eagerifies, yes
nothingmuch what about foo($scalar, @array)? do you get [ $scalar, [ @array ] ] or [ $scalar, @array ]
luqui I expect you get the latter, but a lazy proxy so @array can stay lazy 08:33
nothingmuch yeahg
luqui loves monads much now, and is entertaining the idea of making the hyper operators into monadic binding 08:35
-<< @foo # @foo.bind(&prefix:<->)
ods15 nothingmuch: wtf is lazy
nothingmuch ods15: you have 30 seconds to make a guess 08:36
ods15 not a clue
don't read/do anything until actually called?
luqui s/called/used/ yep
ods15 as opposed to non lazy?!?!?
nothingmuch "strict" or "eager" 08:37
ods15 wtf kind of thing does something without being used
luqui for instance: @foo = <>; print @foo; # input is read on "print", not the assignment
ods15 the whole beauty of doing 'if (NULL && *(int*)NULL) {}' is that it can be done and it's safe...
luqui: that's lazy? 08:38
luqui not in perl 5... I don't think it is in perl 6 either
but: @foo = lazy { <> }; print @foo; # is
nothingmuch ods15: explain why that is beautiful
luqui and what it has to do with laziness...
ods15 nothingmuch: if (ptr && ptr->value) { do something } 08:39
luqui: i mis understood laziness
nothingmuch uh, not all of the language is lazy
just certain things
my @array = 1 .. 1000000000000000; # doesn't explode your computer;
ods15 i thought you meant, since 'ptr->value' is not used, it's not evaluated..
nothingmuch more interestingly: 'my @array = 1 ...'; # to infinity
ods15 "not lazy" in that definition would be insane
coral (x|y) appears to be lazy
nothingmuch ods15: go learn haskell
coral but i think i'm just misunderstanding
nothingmuch ods15: you'll hate it.
coral also, must learn haskell :)
luqui coral, how is x|y lazy? 08:40
I'm just asking, not making a point
nothingmuch coral: 'x || y' maybe? WRT the y?
coral well, in my experiments with evalbot and (x|y) i've noticed that it tends to avoid picking one of the choices until the very last moment possible, if at all
ods15 anyway... that's pretty insane then? you can't keep track of when things are actually DONE, you can end up with crazy race conditions and deadlocks and not understand where/why
coral which seems to fit the definitionish of lazy
luqui coral, evalbot and junctions don't mix
coral ok 08:41
luqui ?eval 1|2|3|4
evalbot_6528 1
luqui ?eval 1|2|3|4
evalbot_6528 1
luqui ?eval 1|2|3|4
evalbot_6528 1
luqui ?eval (1|2|3|4).perl
evalbot_6528 '(1 | 2 | 3 | 4)'
luqui that's more like it :-)
ods15 whats |, not or i'm guessing
luqui it's a special kind of "value or"
?eval my $x = 3; if $x == 1|2|3|4 { "yes" } else { "no" } 08:42
evalbot_6528 'yes'
nothingmuch ods15: junctive or
?eval my $x = 2|5; ($x * 2).perl
evalbot_6528 '(4 | 10)'
ods15 then, it's not a value of any kind
nothingmuch it is
ods15 as in, 1|2|3|4 doesn't really have a value
it's more a list
nothingmuch it does
no it isn't 08:43
revdiablo 1|2|3|4 creates a junction
nothingmuch it's a junction =)
revdiablo which is decidedly a value
ods15 what it's value?
luqui ods15, it multiple values masquerading as a single one
nothingmuch ods15: (1|2|3|4)
coral ods15: have you seen the perl6 module Quantum::Superposition?
ods15 nothingmuch: that's "kind of" a list
nothingmuch ods15: you can passs that around, apply arithmetic to it, and so forth
coral or thereabouts.. it's a dconway thing, in any case.
nothingmuch coral: that is perl 5
coral er perl6
nothingmuch ods15: no it isn't, it's a scalar
coral ARGH
ods15 coral: a better question would be, "have you seen perl6?"
luqui wants to make junctions simply syntactic shorthand though
nothingmuch a list is a list of values with order
coral tries for the 5 twice and misses.
luqui because they so horribly defy the laws of logic
ods15 nothingmuch: ok, i get it 08:44
nothingmuch a junction is a nested collection of values with logical combinators
luqui--; # NONONONO
luqui nothingmuch, why
nothingmuch i've used junctions as values several times, I haven't gotten burned, and it was very nice
ods15 ?eval my $x = 5|2; ($x * 2).perl
evalbot_6528 '(4 | 10)'
nothingmuch this is the "i'm not an idiot" issue all over again
ods15 ?eval my $x = 5|2|2; ($x * 2).perl
evalbot_6528 '(4 | 10)'
ods15 ?eval my $x = 5|2|2.5; ($x * 2).perl 08:45
evalbot_6528 '(4 | 10 | 5)'
coral is there a way to turn a list into a junction?
luqui nothingmuch, okay, I understand that angle
ods15 ?eval my $x = 5|2|2.53; ($x * 2).perl
evalbot_6528 '(4 | 10 | 5.06)'
nothingmuch coral: any(@list)
luqui: another thing I don't like about perl 6 is the "this is not for the children... let's make it a pragma" approach
coral ?eval my $x = 1...; any($x).perl
evalbot_6528 pugs: out of memory (requested 1048576 bytes)
revdiablo ?eval any(1,2,3).perl
evalbot_6528 '(1 | 2 | 3)'
nothingmuch there are already enough daqngerous features to make the children scream and have nightmares
autothreading, off the top of my head, is one that is on by defau.t
ods15 ?eval my $x = 5; $x|=2; ($x * 2).perl 08:46
evalbot_6528 '(4 | 10)'
ods15 ?eval my $x = 5; $x|=2; $x&=3; ($x * 2).perl
evalbot_6528 '(6 & (4 | 10))'
nothingmuch and the way to make it safe is not to make it harder to give it a proper value
ods15 wtf is that
nothingmuch s/proper/; improper
ods15 ?eval my $x = 5; $x|=2; $x^=3; ($x * 2).perl
evalbot_6528 '(6 ^ (4 | 10))'
revdiablo ods15: & means all, | means any
nothingmuch ods15: that means it's 6 and (4 or 10)
coral hmmm. is there a way to take an infinite iterator like 1... and turn it into a junction without evaluating it for 1..inf first?
ods15 nothingmuch: what do you compare it with, another junction?
nothingmuch so if you say (6 & (4 | 10)) < 8
it's ture 08:47
luqui nothingmuch, autothreading is one of those features that enforces "this is not for children"
ods15 ay
thats scary
nothingmuch luqui: but it's on by default
luqui it enforces that you don't give junctions to functions that aren't expecting them
ods15 bbl hungry
nothingmuch ods15: i find it bleasant
pleasant
luqui yeah, and if it were off, then all subs would have to worry about junctions
nothingmuch luqui: if junctions were treated normally unless autothreading, things would not be so clumsy wrt to this safety issue
'sub foo is autothreading' # enough
or 'use autothreading'; # global for this scope 08:48
luqui so you want to be able to say foo(1|2) and let foo assume that its argument behaves normally when it doesn't?
nothingmuch or 'sub foo is pure' # maybe autothreading, maybe not - depends on runtime, #cpus, available memory, and so forth
luqui: i want (1|2) to propagate downwards
luqui oh, autothreading btw has no tie to cpu threads
nothingmuch and let the junction be resolved at the lowest level, by the builtin operators
pure operators will take junctions and return junctions 08:49
luqui that's a bad approach. it's the same reason that we had to kill junctive types and replace them with union/intersection types
nothingmuch things like $fh.seek(1|2); # death
luqui junctions do not obey the laws of logic
so you can't metaprogram if they're around
nothingmuch i argue that you can:
junctions are dealt with at the lowest level in the call stack
and created at the highest level
and anything in the middle doesn't have to care 08:50
luqui but it does
nothingmuch why does it?
luqui if $x == 1 { say "foo" } if $x != 1 { say "bar" }
if $x is a junction, both clauses may be executed
nothingmuch okay
luqui which is contrary to an assumption that the programmer may have made
nothingmuch so what about:
sub foo ($place) { $fh.seek($place) }; 08:51
luqui the writer of this module may not even know that junctions exist (this is that thing -- you don't have to know the whole language to write good code)
nothingmuch foo(1|2); # silently seeks to either 1 or 2?
luqui: yeah, i agree it's dangerous, but IMHO autothreading is still more dangerous
luqui depending on evaluation order :-)
nothingmuch, they're both dangerous
that's why I want to demote junctions 08:52
nothingmuch what I want junctions for is things like:
coral having unlink() or an equivalent in the core function set of any language by default is incredibly dangerous
and yet
nothingmuch is($got, 1|2, "foo");
luqui the same way I want to demote pairs: to make it very visible exactly where you are using them
nothingmuch, hmmm
hmmmmmmmmmmmm
nothingmuch this is where junctions shine
they are sooooo good when you know what the thing you call is doing
they are not useful for big nested calls 08:53
but lots of time control flow is encapsulated behind polymorphism
luqui and that wouldn't work if is were autothreaded
nothingmuch ?eval use Test; is(1, 1|2);
evalbot_6528 pugs: *** No compatible subroutine found: "&use" at tmp-Prelude.pm line 38, column 30-59
nothingmuch uh, crap
anyway it says "ok 1\nnot ok 2\n"
unless we made Test explicitly accept junctions
luqui right
nothingmuch but since it didn't work to start with, no one touched it with a ten foot pole ever since 08:54
i really really really like encapsulating logic
especially when the input is similar, the output is the same, but the path is different
i like to ask my objects 'are you happy? can you do this?' and not deal with it myself at the top level 08:55
this sometimes leads me to overly abusing multiple inheritence in perl 5
but roles and delegation should help me solve that in perl 6
so i don't want to retrain myself just yet
luqui oh, example: 08:56
sub is ($test, $output, $name) { if $test == $output { say "ok 1 # $name" } say "Debug 1"; if $test != $output { say "not ok 1 # $name" } } 08:57
unfolded from if {} else {}
for debugging, and suddenly you're debugging a problem that wasn't there to begin with
nothingmuch luqui: is there any middle grounds? 08:58
luqui So I think junctions are wrong, however I think locally encapsulated junctions (because of your examples) are also probably wrong
how can we fix that
nothingmuch i think that i'd rather just poke and see if something works
and then use it
instead of not being able to use it at all, just because the author didn't think of it at the time 08:59
luqui I'm not sure what you mean by that
nothingmuch well, if i'm not allowed or can't use junctions, then i'm screwed either way
if i can use them, but sometimes i'll cry because of it, i can use them in one liners 09:00
or with code that I wrote
and I won't have to say in advance that I did
or more significantly, i won't have to patch other code to make sure it does
as I see it both autothreading and passed junctions are useful
perhaps this makes sense: 09:01
if there is a branch in control flow due to an external junction, it's a fatal error unless explicitly asked for by the caller
luqui well I'm trying to push autothreading into the list monad for my new proposal
nothingmuch aside from that the user can choose to pass the junction, or to autothread on a lexical pragma basis, with the now safer option (passing) being the default 09:02
luqui: please explain
luqui I'm going to try to steal << >> from qww and turn it into an argument hyper
(the arrows are pointing to the parentheses)
so you can do foo(1, 2, <<3,4>>,5) 09:03
and it will return (foo(1,2,3,5),foo(1,2,4,5))
nothingmuch yeah, i liked that
nothingmuch thinks he brought that up with >><<
but I don't remember now
luqui I think you did
I'm just generalizing it more 09:04
nothingmuch why can't it use foo(1, 2, >>3, 4<<, 5), btw?
luqui well, it *could*
but I think << and >> make more sense, because the arrows always point to the *operator* that is being hypered
and I think qww is not common enough to deserve these very useful brackets
nothingmuch oh, i see 09:05
luqui so that covers autothreading when you want it (plus predictable unpacking, which junctions don't give you)
nothingmuch qww is something I write all the time in perl 5
or wait, qww vs qw? 09:06
luqui qww interpolates and shell splits
nothingmuch ah
right, I agree
i like that a lot
it also has the property, much like hyper, to be so obvious and so noticable, that once a newbie knows what it's doing, they never forget 09:07
luqui hooray
so back to junctions
nothingmuch luqui++; # good proposal
luqui I would like junctions in their current state if we could massage them to obeying logic 09:09
for instance, flipping their junctitude on negation operators and whatnot
or maybe they are just a smartmatch pattern? 09:10
nothingmuch lunch time
luqui happy lunching
luqui gets some food too
nothingmuch how do you make them obey logic?
luqui actually, I don't think you can 09:11
not without implementing Quantum::Entanglement
but we can formalize the notion of "pattern"
against which smartmatch works
and then junctions are just patterns, and there's nothing mathematically evil with them 09:12
nothingmuch hmm 09:13
what's bad about Q::E?
luqui nothing, it's just, um...
not for children :-) 09:14
nothingmuch okay
so again to my "safe" junctions proposal:
boolean logic that involves non uniform control flow is a fatal error
luqui example?
nothingmuch your example for is()
it wouls say 'junctive parameter caused inconsistent program flow at line xxx file x, from line zzz file z' 09:15
with zzz being the first place where the junction was passed
luqui so once you tested it against one thing
er, that sentince won't finish 09:16
nothingmuch heh
luqui so: my $j = 1|2; if $j == 1 { say "foo" } if $j == 2 { say "bar" } # dies
nothingmuch luqui: user def autovivification on p6l
yes
at $j == 1
luqui my $j = 1|2; if $j == 1 { say "foo" } if $j == 1 { say "bar" } # doesn't die
nothingmuch because it's both true, and false, one flow does execute { say "foo" } and the other doesn't
wait, no 09:17
hmm
doesn't make sense
luqui you just make sure that the value the junction represents is always the same
nothingmuch this isn't good enough because sometimes you would like the code you call to "foo\nbar\n" and sometimes you don't
i think you don'
t is the default
so it dies if there is any altered control flow whatsoever
if ($x == 1) { } ...
luqui hmmm
nothingmuch that is death 09:18
unless the code takes that explicitly
luqui then the usual implementation of is() breaks with junctions
nothingmuch right, unless the caller says "that's OK"
but anyway, nevermind, i'm hungry
luqui ahh, the call*er*?
yeah, me too
bbiab
nothingmuch errrr yes, the person who apssed the junction decides under these semantics
unless the calleeeeeeeeee (no Rs) says explicitly that it allows, doesn't allow, and/or autothreads 09:19
*poof*
oh boy! 10:20
nothingmuch is writing a query planner
with Class::Multimethods.
this should be really fun
ofcourse, I don't need it to be efficient yet, but how can I pass on something this fun 10:21
luqui you could use Class::Multimethods::Pure, which is even slower!
nothingmuch yarr 10:22
nothingmuch isn't going to use MMs just yet...
in several minutes
luqui has been using MMs all over the place. They really make programming wonderful 10:35
And they make my programs slow... but I'm not the kinda guy that cares about that
luqui sleep & 10:36
nothingmuch i like the potential of MMD to actually make things fast
wilx MM?
nothingmuch lots of the logic becomes declarative
wilx Is that something like C++ function overloading?
nothingmuch and this can be resolved by a type inferrencing compiler
wilx: yes, multimethods
they're more than overloading though
scook0 and it's based on runtime-type rather than static-type 10:37
nothingmuch scook0: not necessarily
scook0 well, no
nothingmuch it's based also on runtime type
scook0 bah, pedant :) 10:38
nothingmuch =)
static typing is very important to make MMD practical
look at my thread on the runtime linker
scook0 the scary long antihistamine-induced one? 10:39
nothingmuch i don't think antihistamine caused that
it may have degraded it =)
but yes, the long one
it briefly touches on how I see MMDs, the linker, static typing and the optimization pipeline getting along
scook0 but logically speaking, most MMD takes place on dynamic-type, right? 10:41
(even if in practice it might get optimised away)
nothingmuch yes
scook0 claims victory! :) 10:42
nothingmuch but it doesn't seem so close to the truth with this:
my Dog $foo;
my Cat $bar;
fight($foo, $bar);
scook0 huh? (I don't see your point...) 10:43
nothingmuch scook0: the "feeling" of dynamicity is lost here
luqui: how stable is C::M::P? 10:44
wilx What if you have multiple inheritance and more than one instance of one class in the inheritance graph? How does the multimethod pick one? Does that signal compile time/run time error?
nothingmuch wilx: in my opinion yes 10:45
in Damien's opinion manhatten distance wins
we don't3 know yet
scook0 nothingmuch: but if you were to say `$foo = new Poodle`, then it would "feel" dynamic? 10:46
nothingmuch scook0: yes =)
or my $foo = shift;
or my $foo = get_poodle;
then type inferrence can make it really be my Poodle $foo = get_poodle;
scook0 I think we may be violently agreeing -- just a mismatch of personal terminology
nothingmuch oh, i knew we were all along =) 10:47
Can't locate Class/ultimethods/Pure.pm
wow... ultimethods
those must rock
scook0 wants an ultimethod
nothingmuch i wonder what they do
scook0 (or what about penultimethods?) 10:48
nothingmuch i don't even have a guess as to what that would mean
scook0 don't mention them on p6l, or somebody will try to figure it out!
nothingmuch heh
wolverian can we call methods by regex? $obj.*/foo/() # call all methods matching /foo/? 11:30
maybe that's a bad idea...
scook0 wolverian: probably a bad idea to make it that easy 11:38
but with the metamodel, it should at least be /possible/
wolverian ObjectSpace>>.*/.*/; # hm...
scook0 wolverian: the question is, how long until it finds the permutation `0.exit()`? :) 11:39
wolverian heh. that looks totally weird, by the way. 1.kill 11:40
scook0 $?SYSTEM.halt_and_catch_fire 11:41
"~/important_document.txt".unlink 11:42
lots of potential for catastrophe 11:43
?eval 42 11:45
evalbot_6528 42
wolverian I think the integers are more confusing
0.exit, 1.kill, etc.
scook0 ?eval use Test
evalbot_6528 pugs: *** No compatible subroutine found: "&use" at tmp-Prelude.pm line 38, column 30-59
scook0 hmm, does someone test something on a recent build for me? 11:46
(I'm getting weirdness on my local copy, and I want to know if it's my patch)
sorry, thinko -- can someone test something for me?
svnbot6 r6530 | scook0++ | * Merge some more code in Parser.hs 12:00
ycheng I found it's interesting that in perl, access non-exist elements of an array will not stop the program. 13:08
nothingmuch ycheng: search for autovivification 13:15
nothingmuch wants to burn something 13:16
i've had it with character encodings!
wolverian ?eval "foo " ~ ("a" & "c") ~ " bar" 13:17
evalbot_6528 'foo a bar'
nothingmuch i'm so fucking sick and tired of endlessly dealing with this pointless nonsense!
wolverian hm. that doesn't look right, but I don't know what it should be.
nothingmuch ?eval "foo " ~ ("a" & "c")
evalbot_6528 'foo a'
nothingmuch ?eval ("a" & "c") ~ " bar"
evalbot_6528 'a bar' 13:18
nothingmuch hmm
?eval ("a" | "c") ~ " bar"
evalbot_6528 'a bar'
nothingmuch odd
wolverian I was just wondering whether '"<element>" ~ all(open("foo.txt").lines) ~ "</element>" would autothread the concatenation and create <element>..</element> for each line
nothingmuch ?eval ("a" | "c")
evalbot_6528 'a'
nothingmuch wolverian: it should
wolverian that is brilliant.
nothingmuch ?eval ("a" | "c").perl
evalbot_6528 '(\'a\' | \'c\')'
nothingmuch ?eval (("a" | "c") ~ " bar").perl
evalbot_6528 '(\'a bar\' | \'c bar\')'
nothingmuch okay, eval bot is just drunk
it actually works
?eval 1|2 13:19
wolverian yeah, I think he doesn't like outputting junctions
evalbot_6528 1
nothingmuch does anybody know of any cache profiling tools?
(l2 cache, that is)
ycheng nothingmuch: thanks very much for that keyword ! 13:20
nothingmuch you're welcome
stevan hola nothingmuch 13:21
how is live in the construction zone
nothingmuch hi ho 13:23
stevan: yucky 13:24
FUCKING CGI! FUCK YOU FUCK YUOU FUCK YOU!!!!!
elmex nothingmuch: please
nothingmuch sorry
stevan your not using CGI.pm are you?
nothingmuch Kwiki is
=(
elmex nothingmuch: i love fsck-speak :)
nothingmuch kwiki uses a CGI function called query_string 13:25
stevan ah,.. my sympathies
nothingmuch which rejoines the parsed query string
stevan yes
nothingmuch somewhere in the way UTF8 is smashed to bits
stevan has had many text-encoding nightmares lately
nothingmuch it needs to be unicode all the time 13:26
stevan nothingmuch: look... you so offended elmex,.. he quit
nothingmuch that's the only way to deal with it
stevan nothingmuch: I tend to agree
nothingmuch i'm truely sorry, but i'm on the verge of kicking my table
stevan we had to deal with french text,.. which is fine in Latin-1,.. but all the data files were set to us as utf-8
and they were to big to convert
so we ended up with all this mixed up encodings 13:27
nothingmuch why not stick to utf-8?
stevan because the people who were doing the HTML were using Latin-1
nothingmuch then those people should be spanked
stevan and when I said utf-8,.. they said "what is that?"
stevan considered that 13:28
nothingmuch isn't Latin-1 isa utf-8?
stevan I have no idea,.. they did not seem to get along well
nothingmuch =(
wolverian no
nothingmuch ascii isa utf-8?
stevan it is almost over though,.. just one more week
then I can get back to the metamodel :)
wolverian nothingmuch, 0-127 ascii corresponds to utf8, yes 13:29
stevan has been reading papers on sub-typing and generics to help ease the pain
wolverian but latin-1, also called iso 8859-1 has 128-255, which do not correspond to utf-8
nothingmuch to top it off i'm wroking through SSH and vim on the other end is so drunk i can't work
stevan wolverian: I suspect the 128-255 is where all the funny french characters are then
wolverian stevan, yes.
same with finnish.
it is annoying on IRC, since most people use mirc and it doesn't do utf8. 13:30
stevan those were the ones which never seemed to mix properly
wolverian stevan, they don't. you have to recode.
(irssi supports it natively with /recode in 0.8.10 release candidates)
stevan wolverian: yes,.. Encode was my friend :)
wolverian right. :)
nothingmuch why is CGI.pm so big?
stevan that and gaal's DBD::Mysql patch 13:31
wolverian nothingmuch, because it is horrible.
nothingmuch why is it both OO and imperative?
stevan nothingmuch: you should write your own version
then patch Kwiki :)
wolverian if you're on apache, use the apache libs. if not, use CGI::Lite
nothingmuch stevan: there are a ton of CGI things
i don't want to do that
i don't want to patch kwiki too, it takes too long to read ingy code 13:32
stevan was very much kidding :)
wolverian or if you can't help it, use CGI::Simple as a drop-in replacement for CGI.pom
er, s/pom/pm/
nothingmuch too much inheritence
wolverian: does it pollute the CGI namespace?
because kwiki calls to CGI::query_string directly
stevan nothingmuch: I think SocialText is hiring (Ingy's kwiki company)
wolverian nothingmuch, augh.
nothingmuch, no, it doesn't.
it's also strictly OO only, I think.
nothingmuch stevan: i don't want to hack kwiki for a living
stevan nothingmuch: that is easily hacked around 13:33
nothingmuch wolverian: then i'm screwed... I need to rewrite Kwiki:CGI
stevan *CGI::query_string = sub { CGI::Special->querystring(@_) {
nothingmuch: but you would work with ingy ;)
nothingmuch stevan: ugh.... 13:34
stevan :D
nothingmuch (wrt to the hack with the sub)
Juerd CGI as a namespace doesn't make sense for things concerning URIs. 13:36
Query strings are not a CGI thing, they're a URI thing.
wolverian can you reimplement junctions in perl6 trivially yourself? 13:37
Juerd wolverian: Depends mostly on your definition of "trivial".
wolverian Juerd, right. I'm wondering how to easily redispatch everything to the elements
in the object
Juerd I don't understand that sentence 13:38
Or sentance. I never know.
wolverian sentence. :)
I mean: how does the junction object I construct easily run any operation performed on it on each of the elements it has instead? 13:39
method override_all { ... } # :)
Juerd I don't think that'll be easy 13:40
wolverian hmm. new Proxy: FETCH => { for @element { ... } } ? 13:41
nothingmuch oh god.... CGI::init could be fixed 13:43
but it's soooo big and ugly
someone please put a bullet in my head
nothingmuch gives up 13:45
wolverian I just don't know how to do foo(1 & 2) => foo(1), foo(2) 13:47
nothingmuch ? 13:50
wolverian isn't that what the junction does? 13:51
Juerd nothingmuch: What are you doing with CGI stuff? 13:52
nothingmuch Juerd: Kwiki is not supported hebrew usernames
Juerd nothingmuch: So it's not Perl 6 related?
nothingmuch the encodings get screwed
no
Juerd Okay
I was very afraid you were porting CGI.pm literally. 13:53
(Which is okay, but not using that name...)
nothingmuch hell no, aankhen was doing a good job with all the good modules
i think CGI.pm should be left to rot in perl 5 land
nothingmuch should have installed mediawiki instead 13:54
but i was like "hmm... if I have to change something I prefer working with perl"
which sucks, because all of the features I needed so far are supported by mediawiki
nothingmuch goes back to writing fun code3 14:10
sest driven development saved whatever sanity i had left today... 14:23
ciao, i'm going home!
svnbot6 r6531 | iblech++ | * Usual svn props. 14:28
r6531 | iblech++ | * New util/smokeserv/: Pugs Smoke Reports Server
r6531 | iblech++ | * smokeserv-client.pl: Submits smokes to a smokeserv
r6531 | iblech++ | * smokeserv-server.pl: The smokeserv (HTML template included)
r6531 | iblech++ | * README
r6531 | iblech++ | * util/yaml_harness.pl: Include the path to pugs used in the smoke so the
r6531 | iblech++ | smokeserv can easily differentiate between normal runcore/PIL-Run/PIL2JS.
r6531 | iblech++ | * pugs::hack: Add info how to submit a smoke to a smokeserv.
iblech putter: smokeserver is up! :) 14:29
m19s28.vlinux.de/cgi-bin/pugs-smokeserv.pl
Submitting your own smoke is easy, a "$ ./util/smokeserv/smokeserv-client.pl ./smoke.html" should suffice.
(Help on the design is very much appreciated, BTW :)) 14:34
stevan iblech++ 14:48
iblech putter++ # his idea
svnbot6 r6532 | iblech++ | util/smokeserv/smokeserv-server.pl: 15:31
r6532 | iblech++ | * Very minor typo and XHTML validity fixes.
r6532 | iblech++ | * Obsolete smokes are now removed. The algorithm differentiates between
r6532 | iblech++ | pugs_version, osname, and runcore, so it won't blindly delete smokes older
r6532 | iblech++ | than n days.
r6533 | iblech++ | PIL2JS: PIL::Params: Fix invocant parameters wrongly being moved to the end of 15:49
r6533 | iblech++ | the fixed parameter list. reverse.t has a chance of passing now again.
putter hey folks. 16:04
iblech++
putter running smokes
iblech Hi putter :) 16:05
xinming iblech: you there please?
oops. 16:06
iblech: hmm, Did @Larry confirmed about the 'named parameters' passing?
iblech: for my test
wish I can do something before I go to BeiJing. :-P
iblech xinming: Yes: You are not allowed to pass slurpies by name. I.e. sub foo (*@args) {}; foo(args => [...]) is bogus
putter I wonder if it's time to give ext/ an explicit test/smoke inclusion flag. MIME-Base64/t/Base64.t has been running for 15+ minutes, and actually am only interested in the core test suite at the moment. (sometimes I just find ext/|grep '\.t$'|xargs rm ;) 16:07
xinming iblech: and slurpy hash won't slurp the named params, right?
iblech xinming: Right. 16:08
putter: You can use the PUGS_RUNTIME hack
xinming sub foo ( $a, *%h ) { }; foo( a=> 'v', a=> 'u' ); will cause an error; and foo( a=>'v' ); works, but %h get nothing, 16:09
iblech putter: See util/yaml_harness.pl:41
xinming: Right.
xinming bbl
QtPlatypus putter: It shouldn't be taking that long.
putter iblech: which reminds me, it might also be worth distinguishing (avoiding gc) on the summary line. So if someone does an ext/-free smoke, like I am now, it doesnt smash one with. or parrot-working vs not. 16:10
QtPlatypus will have to look at it tommrow morning to see if there is some bug in the code that is screwing it up, and I'll see if I can oplimise it.
putter QtPlatypus: k. it did seem odd. thanks :) 16:11
iblech: looking...
TheOddity Hi, is this the right place to talk about perl threads? 16:12
5.8.x
xinming TheOddity: try #perl instead. :-)
TheOddity: that might be better for you. 16:13
iblech putter: Yep. Will only need an addition to yaml_harness, testgraph.pl or run-smoke -- one of these scripts have to record the setting.
TheOddity sorry to be such a pain, but I could not find #perl ? #perl-es #perl.ch #perl.de #perl6 #perlcafe #perlchina #perlfr but not #perl. do i need to connect to a separate server for that? 16:17
svnbot6 r6534 | iblech++ | * t/rules/: Some more guards to skip the test if rules support is missing and
r6534 | iblech++ | unCOMMENTED a previously hardfailing test.
r6534 | iblech++ | * PIL2JS:
r6534 | iblech++ | * PIL::Params: Invocant parameters shouldn't be allowed to be passed by name.
r6534 | iblech++ | * Prelude::JS::*: Extremenly minor misc. fixes.
iblech TheOddity: Huh? on irc.freenode.net/#perl are 323 people currently
TheOddity i'm using gaim, probably something wrong with the list-rooms function, will try to join room by specifying the room manually. 16:19
it works!!
thanks ..
iblech food & 16:30
xinming hmm, anyone here can tell me if there is way to ban the private message on advertisement? 16:31
putter strawman proposal: create a "runpugs" wrapper script, in p5, which handles command line analysis, and knows where the various backends and repls are. We could add it to pugs (which currently only knows about PIR, which it actually compiles), but perl seems preferable to haskell for this kind of hackery. Thoughts? 16:38
Aankhen`` Hola! 16:40
xinming iblech: hmm, sub func ( $n, *%h ) { }; func( 100, n => 100 ) 16:43
iblech: what the result will be in this example?
iblech: hmm, sub func ( $n, *%h ) { }; func( 100, n => 200 )
an error or 200? 16:44
iblech Aankhen``: Welcome back! :) 16:47
putter: re. Yep, I think Perl is more suited for that job, too 16:48
xinming: Your subbodies were both empty
Aankhen`` Thanks!
How's things?
iblech xinming: But anyway, $n will 200, and the dangling 100 will throw an error ("too any arguments passed to...") 16:49
xinming :-)
iblech Aankhen``: PIL2JS passes 74%, and PIL-Run is actively developed again 16:50
Aankhen`` Nice!
Any progress on the Haskell backend? :-D
autrijus no, because I was away with $work and moving to another house 16:51
Aankhen`` Ah, okay. 16:52
And hi. :-D
autrijus plus a bruising finger jammed by a random door, today
Aankhen`` Ow.
Sounds painful.
autrijus feels rather weird typing with nine fingers
yeah. will heal tomorrow likely
putter iblech: btw, re 74%, nothingmuch's smoke _is_ passing the :perl5 tests (just not the p6 rules tests). so there is perhaps an extra ~10% when you want it. basically just requires ~~ recognize whatever Prelude/JS/Rules.pm creates.
Aankhen`` My condolences. :-(
autrijus thanks. :) 16:53
iblech putter: ooh +10% :) I'll look into your Rules.pm today or tomorrow
autrijus rehi all, btw. I'm delighted to have plenty of things to catch up every couple hours :) 16:54
oh, and good news: my Perl-I18N talk in Euro OSCON is cancelled
replaced by "Introduction to Pugs":)
Aankhen`` applauds! 16:55
putter iblech: one ./util/smokeserv/smokeserv-client.pl ./smoke.html coming up... 16:56
iblech putter, autrijus: ooh :)
autrijus hm, I probably need a small project to acquaint myself with the p5 works 16:58
say, what would take it to plug Inline::Pugs into the pilrun framework? 16:59
or has it already been thought of?
pasteling "Aankhen``" at 203.101.5.144 pasted "Build fails on Win32" (17 lines, 750B) at sial.org/pbot/12780 17:00
putter iblech: is there a delay before an upload smoke shows up? 17:01
iblech putter: No, shouldn't be. Did smokeserv-client.pl say something other than "success"? And was your util/yaml_harness.pl synced up? 17:02
putter I synced up util (though I've been editing yaml_harness). the client dumped html text at me. looks vaguely like the text of your page. 17:03
iblech putter: Checking
putter: I had this problem during debugging, but it magically disappeared
putter autrijus: all the progress on pilrun has been fglock's. I've just tried to keep the rest of it out of his way. smoke coming up (in a while). 17:05
autrijus understood
putter re Inline::Pugs, do you mean p5 user code use Inline::Pugs; and get the metamodel, etc?
autrijus yes.
iblech putter: Problem (hopefully) fixed, try again please? :) (It was discarding your POST because the smoke was too big) 17:06
putter Only two issues come to mind. err, 3. We kind of stomp on the p5 namespace (defining Int, Str, etc). I don't think we have quite gotten to the bottom of what's a p5 method, is it a Method, etc. So users should expect instability, and I'd hate to put any kind of crimp on stevan/flgock/etal making major changes. And there is a GB-eating (deep recursion?) in pilrun or the mm which shows up on some test files. 17:10
iblech: trying...
autrijus okay. I'll do the wrapper thing first 17:12
iblech putter: worked :)
autrijus Fyi: require(P5Runtime/PrimP6.pm) - file not found at lib/Inline/../../perl5/PIL-Run/lib/PIL/Run/PrimP5.pm line 187.
hm, what does this signify? something akin to :precomp needs to be done?
putter hmm, you may be the windows sacrificial guinne pig. 17:13
putter will be intermittantly non-responsive over the coming minutes as memory hungry tests run... 17:14
perl5.t giving PIL2JS indigestion... 17:18
svnbot6 r6535 | autrijus++ | * port Inline::Pugs to use the PIL-Run runtime. So far this works: 17:23
r6535 | autrijus++ | use Inline Pugs => q{
r6535 | autrijus++ | sub intone { say "Hello, $_!" }
r6535 | autrijus++ | };
r6535 | autrijus++ | intone('World');
autrijus seen fglock 17:39
hm. I'll check back when I wake up.
autrijus sleeps &
putter good night autrijus... 17:40
ibech++! :) looks great. more smokes coming...
err, iblech++ ;)
iblech Night autrijus :) 17:42
Thanks, putter++, for the idea :)
putter iblech: does a single js get used for multiple tests? I have a js, now running for 12+ minutes, which has been slowly growing, now almost up to 1/2 GB. 17:53
svnbot6 r6536 | yiyihu++ | Change the test according to www.nntp.perl.org/group/perl.perl6....uage/22883
iblech putter: No 17:54
putter: 12+ min -- if you set PIL2JS_RESOURCE_GUARD=1, then all js's taking longer than 35s are killed by the kernel
putter I may have broke something by ^Z'ing the test to get some swapping-free breathing room. 17:55
iblech no, that shouldn't have changed anything
putter do many tests exceed 35s?
xinming ?eval [+][1,3,5] 17:57
evalbot_6528 (1, 3, 5)
iblech putter: At least on my machine (1.667 GHz, 512 MiB RAM) probably only the rules/ tests 17:58
putter ok. tnx. 17:59
iblech xinming: "dies_ok { position_with_slurpy_hash1 n => 100, test => 40, 1, 3, 5, 7 }," -- why?
xinming iblech: hold on please 18:01
iblech xinming: Finished reading the rest -- looks correct :) 18:02
xinming iblech: :-P forget to add 100 before (n => 100) pair. T_T
iblech: help me fix it please. :-) 18:03
iblech xinming: Ah, ok, then it's correct :)
xinming: err, you mean forgot to add (n => 100) before the pair (n => 100), right? 18:04
xinming dies_ok { position_with_slurpy_hash1 100, n => 100, test => 40, 1, 3, 5, 7 }
iblech: the comment above the test shows what I wisht o do. :-) 18:05
iblech xinming: You've to add another n => "does not matter" to the arglist, or you've to s/dies_ok/lives_ok/ 18:06
xinming iblech: ... 18:07
bbl
## `sub func ( $a, *%h ) { }; func 100, a => 200` should be an error 18:09
dies_ok { position_with_slurpy_hash1 n => 100, test => 40, 1, 3, 5, 7 },
'Testing: passing positional parameter with pair which has the same key name.';
I mean, change this dies_ok { .. }... 18:10
to dies_ok { position_with_slurpy_hash1 100, n => 100, test => 40, 1, 3, 5, 7 }, ' ... ';
because you ever said that might raise an error.
ods15 evalbot_6528: say "a" for (1...); 18:11
err
?eval say "a" for (1...);
evalbot_6528 pugs: out of memory (requested 1048576 bytes)
ods15 ?eval $a = 1+2 for (1...);
evalbot_6528 pugs: out of memory (requested 1048576 bytes)
ods15 ?eval $a = 1+2 while (1);
evalbot_6528 Error: Undeclared variable: "$a"
ods15 ?eval my $a; $a = 1+2 while (1);
evalbot_6528 (no output) 18:12
ods15 did i break it?
ah, damn, smart :P
iblech xinming: Right, "sub func ( $a, *%h ) { }; func 100, a => 200" should be an error, but &position_with_slurpy_hash1 has got an additional *@slurpy_array which accepts the additional "100"
ods15: evalbot is unbreakable :D
ods15 i've gotta find something :P
xinming iblech: hmm,
ods15 ?eval say "hello\nworld" 18:13
evalbot_6528 hello world bool::true
ods15 ?eval say "hello\nworld";
evalbot_6528 hello world bool::true
iblech ods15: It s:g/\n/ / :)
?eval "\n" x 100 # would be many lines 18:14
evalbot_6528 ' '
xinming iblech: Oh, Ok, thanks for reminding me this.
ods15 iblech: ya otherwise i could send irc commands
iblech xinming: Sure. Thanks for fixing the rest of the test :)
ods15 say "bla\nquit"
iblech ods15: And you could flood the channel
ods15 only if i did 'privmsg #perl6 :' after each \n .... 18:15
iblech ods15: But no, that wouldn't be possibly, because Net::IRC is smarty enough and appends PRIVMSG #perl6 to each line
ods15 ay
xinming iblech: by the way. 18:19
putter ok, while that runs... 18:20
svnbot6 r6537 | putter++ | Consolidated ENV{PUGS_RUNTIME} hackery in run-smoke. yaml_harness is now backend-clean.
r6537 | putter++ | yaml_harness.pl: --exclude now takes a regex pattern rather than a string, and it is applied to the path, not just the bare filename.
putter iblech: is there a ~~ in piljs?
xinming sub func ( $a, *%h ) { }; func( 100, x => 1, y => 2, 11, 12, 13 ); will this be an error?
I think so. :-) 18:21
iblech xinming: Right, as there's no place where the 11,12,13 could go. It would work if you added a *@slurpy_array, though.
putter ah well. I tried smokeserv-client'ing a copy of Johnathan's smoke, so we would have something windows up, but got a error: The submitted smoke has an invalid format! 18:26
iblech Right, as it doesn't contain a pugs-path-info 18:27
Only new copies of yaml_harness.pl include that field 18:28
(It is needed to associate a smoke with normal/PIL2JS/PIL-Run)
putter Ok. So in future we'll be able to use them. good.
iblech Right.
xinming ?eval { sub foo { $_.say } }; foo( "haha" ); 18:29
evalbot_6528 haha bool::true
xinming Is this a bug?
sub is defined in a closure, how can it be "package" scoped?
putter perhaps you meant my sub ... 18:30
iblech xinming: Pugs' behaviour is correct: sub foo {...} is really BEGIN { our &foo ::= sub {...} }
xinming: Thus all sub declarations are automatically package-scoped, unless you prefix them with "my"
?eval { my sub foo { $_.say } }; foo( "haha" ); 18:31
evalbot_6528 Error: No compatible subroutine found: "&foo"
svnbot6 r6538 | putter++ | crude_repl.pl: added -w command line argument (ignored). make smoke-perl5 again works. 18:38
putter iblech: can spidermonkey do a system() call? (google is proving unhelpful...) 18:42
iblech putter: Not that I know of...
putter: At least the standard bin/js does not. I tried embedding spidermonkey in a P5 script (to provide printWithoutNewLine, sleep, etc.), but I got many segfaults 18:43
Aankh|Clone goes to sleep. 18:44
See y'all in the morning.
G'night.
iblech Night Aankhen`` :) 18:45
putter yeah, I dont see it in the source. ah well, no parrot pge for p6 rules. 18:46
g'night Aankhen``. welcome back. 18:47
iblech: you've seen the p5 cpan module to use spidermonkey?
iblech putter: Right, I've used that
putter wow. dont think I've ever seen a 3GB perl process before. 18:49
iblech I have seen some of them since PIL2JS was able to make smoke :) 18:52
putter hmm, there seem to be two of them search.cpan.org/~claesjac/JavaScript-0.55/ search.cpan.org/~mschilli/JavaScrip...nkey-0.11/
the first seems under active development. 18:53
iblech I took the first
putter mmm... Inline::Pugs::JS ... ;) 18:54
iblech :)
putter autrijus: btw, re one of your soft typing posts, one option I didnt see mention is the compiler emitting both simple and optimized code, and using observer hooks to invalidate the optimized version. 19:02
JavaScripdt::SpiderMonkey make test works at least. 19:07
iblech JavaScript.pm worked fine for small scripts -- but with the embedded Prelude (at that time ~~5000 lines), it segfaulted all the time... 19:09
putter ;) pugs, its not just perl6, it's a stress test. 19:11
fglock hi all 19:17
iblech Hi fglock 19:19
putter woot. JavaScripdt::SpiderMonkey handles a "commented out a couple of problematic lines" version of PIL2JS.js ! :) 19:27
pasteling "puttern" at 66.30.119.55 pasted "non-functional diff to PIL2JS.js to run under JavaScripdt::SpiderMonkey" (76 lines, 2.1K) at sial.org/pbot/12782 19:31
putter hi fglock. 19:32
fglock hi putter
putter "handles" as in "survives well enough to run trivial p5 callback at end". none/little of the code was actually run. 19:36
svnbot6 r6539 | fglock++ | * PIL-Run - fixed lazy slices - @a[1..1000]=@a[1001..2000] 19:38
putter fglock: have you seen iblech's new smoke server? m19s28.vlinux.de/cgi-bin/pugs-smokeserv.pl 19:39
iblech: I added a link from pugs.kwiki.org/?SmokedPugs 19:40
fglock nice - are you placing PIL-Run there too? 19:41
putter ha. couldn't ask for a better straight-line. smoke done. uploading...
iblech putter: Thanks :) 19:42
:D
putter :( error: The submitted smoke has an invalid format!
Hmm, no pugs-path-info field... 19:43
iblech old yaml_harness.pl?
putter nope, brand new recently edited and commited yaml_harness. maybe I broke something? 19:44
iblech checking... 19:45
hm, your commit seems sane 19:47
manually add "pugs-path: something/PIL-Run/something" so you can upload it? 19:48
but yaml_harness should add pugs-path, hm 19:49
putter I see a $Config{"pugs-path"}, is that what should be doing it?
iblech Right.
putter $Config{"pugs-path"} = $ENV{HARNESS_PERL};
hmm, no other action required...?
iblech I think so. At least, it worked when you and I tested it 19:50
putter I see pugs-path in the config: in the smoke file. where does the -info come in?
iblech ERR! I misread your "Hmm, no pugs-path-info field..." as no pugs-path (without -info) 19:51
-info is not necessary
Could you nopaste the header section from the smoke.html please? 19:52
putter pasting...
xinming ?eval { my foo { 'a'.say } foo;} { my foo { 'b'.say } foo;}]
evalbot_6528 Error: unexpected "{" expecting variable name or "("
xinming ?eval { my foo { 'a'.say } foo;} { my foo { 'b'.say } foo;}
evalbot_6528 Error: unexpected "{" expecting variable name or "("
xinming ?eval { my foo { 'a'.say } foo;}; { my foo { 'b'.say } foo;}
evalbot_6528 Error: unexpected "{" expecting variable name or "("
iblech xinming: You need the "sub" keyword -- my *sub* foo {...
xinming ?eval { my sub foo { 'a'.say } foo;}; { my sub foo { 'b'.say } foo;} 19:53
evalbot_6528 a b bool::true
wolverian er, should that output anything actually? I thought bare { }s are closures now
pasteling "putter" at 66.30.119.55 pasted "requested smoke-perl5 header section" (27 lines, 697B) at sial.org/pbot/12783
wolverian or are they called automatically in void context?
iblech wolverian: Right, they're called automatically 19:54
wolverian I don't know if I like that.
sub foo { { ... } } foo; # calls the inner closure?
iblech wolverian: Right, you need "return {...}" if you want to return the closure
wolverian ew. 19:55
that feels exceedingly ugly.
iblech putter: I changed smokeserv-server.pl to be a bit more verbose -- could you try uploading it again?
putter uploading... 19:56
xinming ?eval sub foo(+$n, *%h, *@a){ @a.sum }; foo 1, x => 20, y => 300, 4000;
evalbot_6528 4000
xinming is it a bug?
iblech wolverian: If you want blocks, there's no way around it -- or do you want to type "{ my $a... }()" (with the ()) all the time?
xinming named param act the same as positional
putter error: The submitted smoke has an invalid format![[summary!runcore!duration!pugs_revision!]]
iblech xinming: It is
wolverian iblech, do { }
iblech putter: Aha! It lacks the osname! 19:57
xinming iblech: oh, I ever thought there might be something wrong with my test.
iblech putter: Should I $osname ||= "(unknown)"?
xinming and even think that the bug is about function scope
wolverian iblech, I don't like that I have to remember that closures are a special case when you return them from a function 19:58
iblech, it feels like perl5
putter hmm. smoke complains each time I run it that I should do something with a configure file, or give it an --anonymous option. neither of which are easy enough to have been pursued. could that be part of the problem?
iblech putter: That should only be relevant if you want your name to be included in the smoke.html 19:59
fglock PIL-Run thinks (1..2,5) is ((1,2),5)
svnbot6 r6540 | fglock++ | * perl5/Array - fixed "@a[4..5]=6"
iblech putter: I've fixed smokeserv, try again please? :)
fglock: Probably because * in subroutine signatures isn't implement correctly (as &infix:<,> has a signature of (*@things)) 20:00
putter iblech: "unknown" isnt a fix... something about smoke is broken...
iblech fglock: * has to flatten arrays
putter its a workaround...
iblech putter: right...
putter uploading anyway...
error: The submitted smoke has an invalid format![[summary!runcore!duration!pugs_revision!]] 20:01
why dont I paste the entire beginning of the file, so you can easily test?
iblech You're definitely right, it's something weird with the smoke system 20:02
pugs_version wasn't set
oooh!
I know why!
build_info: &quot;--version unimplemented\n&quot; 20:03
It's crude_repl
putter ?
iblech the run-smoke calls $pugs --version to get the build info
osname, version number, etc.
putter ahhhhhh.
fixing...
iblech :)
putter++
putter err, but pugs --version doesnt have osname...?? 20:04
does say Version: 6.2.9 (r6531) 20:05
iblech hrm
right.
looking...
ah, it calls -V 20:06
see yaml_harness.pl:121
$self->{_build_info} = join '', qx{$executable -V};
putter ahh, not --version, but -V
ok, I'll fudge that. use pugs's -V. what does js smoke do? 20:07
iblech pugs-smokejs.pl proxies most calls to the real pugs 20:08
putter ok
iblech only pugs-smokejs.pl pugs -w t/foo/bla.t is handed over to PIL2JS
svnbot6 r6541 | fglock++ | * PIL-Run - p6_from_a has a signature of *@a 20:11
r6541 | fglock++ | - this fixes <,> - @a[1..2,5]=@b works now
fglock bbiab 20:12
putter I fixed crude_repl, and hand patched smoke-perl5.html with the output. uploaded.... 20:16
works! :)
Duration a bit arbitrary. Took me a while to notice one of the tests is currently non-terminating. 20:17
svnbot6 r6542 | putter++ | crude_repl.pm: -V passed on to pugs. The version information is used by smoke.
iblech :)
Great! 20:18
putter smoke-pir in progress... ;)
iblech: spiffy site. 20:20
iblech Hm, problem: pugs-path is "./pugs" with make smoke-pir
Fixed, testing 20:21
svnbot6 r6543 | yiyihu++ | Correct some mistake I made. Rewrite Part of test, which makes it clear... (to me at least) 20:23
iblech fix doesn't work... 20:25
putter the smoke is running far too well... I suspect its not -BPIR.
looks like Makefile passes -BPIR to run_smoke.pl. exploring... 20:26
iblech Hm, it seems $optional_args doesn't get used
Probably changing line 17 of run-smoke.pl to include $optional_args suffices 20:27
xinming iblech: which line please? :-)
iblech xinming: Hm? I meant line 17 of run-smoke.pl (not your test)
xinming I am watching for your feedback. ;-)
iblech: oh,Ok, hmm, Need to take a shower. bbl. 20:28
putter yes, fix in.
svnbot6 r6544 | putter++ | run_smoke.pl: was ignoring optional command line arguments. so make smoke-pir wasnt -BPIR. Fixed.
iblech xinming: Ah :)
putter or not fixed... 20:29
ah, Null PMC access. fixed! :) 20:30
iblech :) 20:32
putter So, iblech. (putter briefly dons "interviewer" hat), what are your plans for PIL2JS?
;) 20:33
iblech My goal for the very near feature is to make .pairs work again.
putter lol
iblech Really, PIL2JS is still a fun-only project, I don't have any real plans :) 20:34
putter Quite a remarkable fun-only project. Did you think it would get as far as it has when you started? 20:36
putter notes that if not for rules support, with its unusual number of tests, PIL2JS would be neck and neck with the pugs backend... 20:37
iblech Thanks :) No, absolutely not. The prototype was a 400 line long script, and it worked, then I wanted to pass 01-sanity, then mandel.p6, then Test.pm, and suddely PIL2JS passed 74% :) 20:38
putter And that's doing things well enough to make up for not receiving any oo data from pugs via pil.
:) 20:39
obra how fast is PIL2JS running these days? 20:40
putter So perl6 may someday be a JSAN module? ;)
iblech obra: A complete smoke (without ext/), with a guard so no test may take longer than 35s, takes ~~44min. 20:41
obra: I.e., it's probably half as fast as normal Pugs
obra nod 20:42
iblech obra: It used to be much faster (before the CPS change), but without CPS, PIL2JS couldn't have continuations (and thus proper &return, coroutines, gather/take, &sleep, etc.) 20:43
obra *nod*
iblech obra: If JavaScript had goto, PIL2JS would be at least twice as fast as it is now.
obra does the next rev of js? 20:44
iblech Not that I know of.
obra 16:41 <iblech> obra: I.e., it's probably half as fast as normal Pugs
16:43 <iblech> obra: If JavaScript had goto, PIL2JS would be at least twice as fast as it is now.
Those statements taken together scare me ;)
iblech ;)
clkao mad iblech 20:47
iblech putter: I wonder if PIL-Run passes more tests than PIR :D 20:50
xinming back 20:52
iblech: do you think test is much bettter? :-P 20:53
iblech xinming: Sorry, haven't looked
xinming: (Making .pairs in PIL2JS work again :))
xinming iblech: hmm, well, It's Ok, Just wish to make my test better. 20:54
I need to sleep now, If you think there is any place which can be improved, please let me know. 20:55
:-)
bye all
putter iblech: almost certainly. smoke in progress... ;)
bye xinming 20:56
iblech Night xinming :)
putter: :)
Yay! .pairs fixed! :) (PIL2JS have had this bug for three weeks) 21:03
putter oh my. like wow. hey iblech. working pairs++. how would you like to run them linked with perl5? 21:04
say "Hello, world" just ran in a test-jig... 21:06
!!!
Sorry. Got distracted. Yay! .pairs fixed! :-) 21:07
./runjs.pl -e 'say "hi from linked monkey"' 21:09
# Number of lexical scopes: 1
Skipping &*END.
hi from linked monkey
#IGNORE NEXT LINEFEED#
1..0
#IGNORE NEXT LINEFEED#
iblech ooooh! :)
No segfaults! :)
pasteling "putter" at 66.30.119.55 pasted "run_js_on_jssm" (13 lines, 368B) at sial.org/pbot/12784 21:11
putter oops, left the "write to file" debugging line in.
leo_ [ot] PyPy runs on its own! Hurray!. - PyPy release 0.7.0 21:12
putter JavaScripdt::SpiderMonkey rather than JavaScript.pm. There are two reasons for a module...
leo_ - but a bit related ;-)
putter leo_: !! huzzah! :)
iblech putter: Can you try running a bigger script (say, t/builtins/lists/elems.t)? If it doesn't segfault, check it in please :) 21:14
putter iblech: testing... 21:15
a moment of paranoia... are Test runs now independent? (I'd hate to stomp on the takes-a-long-time PIR smoke...) 21:17
iblech I think they are. And I think that all tests which do IO (= which could interfer) won't run under -BPIR 21:18
putter ah, what the hell... array/elems.t 12 for 12! :)
what's a harder one...? 21:20
;)
iblech Wow! 21:21
arith.t
putter testing...
iblech this one should take relatively much memory
I.e. it's a good stress test 21:22
putter ReferenceError: _26main_3a_3atryeq is not defined ? 21:24
iblech Works fine here... 21:28
svnbot6 r6545 | iblech++ | Pairs should stringify to "$key\t$value", per the
r6545 | iblech++ | pairs-behave-like-one-elem-hashes. Modified
r6545 | iblech++ | t/builtins/arrays_and_hashes/pairs.t and t/data_types/pair.t accordingly.
r6546 | iblech++ | PIL2JS: Fix the long-standing bug of .pairs not working.
r6546 | iblech++ | * PIL2JS.js: Repair %hash = @array
r6546 | iblech++ | * PIL2JS.js: Croak on unknown container combinations
r6546 | iblech++ | * PIL2JS.js: Make $foo := $bar call $foo's .STORE **with an invocant**.
putter I get the same error, modulo the extra #IGNORE NEXT LINEFEED#s, with normal run_js().
js gets over 1GB big doing arith.t! 21:29
putter svn up's
iblech Hm, weird. Do you get the error at the very beginning or later on? 21:30
putter very later on.
has actually printed the 1..183. error is the next line show. 21:31
iblech Very weird. &tryeq is used at the very beginning, so it should not die later on 21:32
("_26main_3a_3atryeq is not defined" == "&tryeq not declared") 21:33
d'oh, misread you. 21:34
pasteling "putter" at 66.30.119.55 pasted "runjs arith.t error message" (15 lines, 1.2K) at sial.org/pbot/12786
iblech It has to have do to something with your modifications, as arith.t runs fine here (I ^Ced after ok 117) 21:35
Hm, I don't have any local modifications ATM either... 21:36
BTW, the huge memory usage of pil2js.pl may be because of a bug in perl5 -- I've found at least one bug (lib/PIL/PApp.pm:146 -- a lone "$self;" changes things), so it's possibly that some internal state is b0rked 21:39
svnbot6 r6547 | putter++ | PIL2JS - JavaScripdt::SpiderMonkey support 21:40
iblech oooh /me goes to look
putter I svn up -r6546 . Same error. Could it be a PATH problem? Ie, a file not loaded? Do you normally invoke runjs elsewhere? Puzzled... 21:42
well, smoke-pir is quite boring. :/ 21:44
wilx How do you run the JS tests? 21:46
Just with Firefox?
iblech putter: I normally run runjs.pl from perl5/PIL2JS/, but that shouldn't be a problem.
putter: Want to send me the .js runjs would execute?
putter uploaded. those %ok are very misleading. PIR choked on almost every test. but of those few it didnt, it got some oks. just not the right thing...
iblech wilx: make smoke-js
putter does make test-js work now? 21:47
iblech: all 1.6MB of it? 21:48
ok, make test-js isnt working yet. t/01-sanity/01-tap.................................pugs: user error (Cannot generate code for Js). so just make smoke-js. or running the tests individually. 21:49
iblech: there is a 21:51
var _26Test_3a_3atryeq but no var _26main_3a_3atryeq. 21:52
iblech 23:48 < iblech> putter: WRT PIR -- the two tests which work are both run at compile-time,
the runtime doesn't have anything to do...
23:48 < iblech> putter: Haven't tested
23:48 < iblech> putter: No, it doesn't know about -BJS 21:53
putter tries to decode iblech's post... 21:54
iblech I mean, the only test PIR passes are exit_in_check.t and exit_in_begin.t
putter got it.
;) 21:55
so, _26Test_3a_3atryeq is defined, and used once or twice. _26main_3a_3atryeq isnt defined, but is used a lot.
iblech That's weird. But I do have an assumption: 21:56
Try ./pugs-smokejs.pl pugs -w ../../t/operators/arith.t
This is the same as runjs, but uses the precompiled Test.pm
(And I tested arith.t using pugs-smokejs.pl) 21:57
It's possible that somehow, PIL gives PIL2JS wrong sub names -- it prefixes the "tryeq" with "Test::", probably because it doesn't get that it has left Test.pm's namespace 21:58
I.e. it thinks the declaration of tryeq is still in the "Test" namespace 21:59
putter ok, ./pugs-smokejs.pl pugs -w ../../t/operators/arith.t worked.
??
iblech ok, assumption confirmed :) 22:00
putter :)
iblech To summarize: Pugs thinks "module Test {...}; sub tryeq {...}" really means "module Test { ...; sub tryeq {...} }"
(but Pugs is wrong.)
putter And in answer to your long-ago question... JS::SM does arith.t! :) 22:02
iblech :))
putter++
putter use Inline::Pugs is definitely going to need some options now... ;) 22:03
iblech :D 22:04
putter I suppose the right thing is runjs or something taking a --compile-to-js -o - and then doing the rest in some p5 module? 22:05
oh, no. there isnt any reason to have a separate process. a PIL2JS module? 22:06
iblech There's jsbin_hack(compile_perl6_to_standalone_js("-e", $code)) 22:07
putter use PIL2JS; PIL2JS::compile_p6_to_js() ?
ah.
perl -Ilib -we 'use PIL2JS; PIL2JS::eval_p6(q/say "hi"/);' 22:14
iblech Hm, but I don't see a possibility of "PIL2JS::eval_p6(q/sub foo {...}/); foo(...)" working 22:16
Unless we use IPC, similar to how Pugs communicates with an external Parrot
putter is there a mechanism for js to wait on/watch an object? the JSSM api lets one modify objects in the (I assume) running runtime. 22:20
iblech Err. Ignore what I said. I forgot JS can run in the same process as the main P5 program now. 22:21
In this case, I see the possibililty of it working :)
putter ;)
iblech ok, clearly I need to sleep now ;) 22:22
Night all! :)
putter though perhaps PIL2JS should just provide code, rather than eval and such. else you have to worry about multiple $js instances, throwaway vs persistant, etc.
ok,
good night iblech!
Fun day. Thanks! 22:23
fglock putter: ping 23:04
putter pong 23:08
fglock: pong
fglock hi putter 23:09
how are the tests going?
putter hi
tests?
There is now a smoke-all on iblech's server...? 23:10
fglock you were trying to make Test work on PIR-Run?
putter briefly. ;) PIL-Run currently uses its own Test.pm in its lib6. 23:12
basically, 23:13
fglock maybe I'll write some tests with lazy arrays - but I'll need some way to time-out the failing tests
putter ok, that should be easy to add... working.. 23:15
fglock I'll take a look on making hashes work 23:19
putter fglock: 6548 has a --timeout argument. it may even work. hmm, perhaps an ENV variable would be easier to use with make smoke, etc? 23:28
fglock thanks - I'll check it 23:29
putter 6549 has a PUGS_HACK_TIMEOUT env var. 23:30
svnbot6 r6548 | putter++ | crude_repl.pl: new --timeout Nsec argument.
r6549 | putter++ | crude_repl.pl: can also set timeout with an environment variable.
r6550 | fglock++ | * perl5/Value - Pair stringifies
fglock should I use the t/ directory? 23:35
PIL-Run/t
putter err... 23:40
whatever is easiest. the perl5/ slushpile needs to go away, but for now, I guess, whereever. 23:41
fglock maybe pugs/t ? the problem is that pugs doesn't implement lazy lists, and will loop forever 23:42
putter ahh.
hmm, maybe crude_repl should define a "you are running on X backend" variable?
then the test can be if X'ed. 23:43
fglock $?BACKEND ? 23:44
putter eh, howabout pilrun t/ for now, and then we'll deal with the integration as part of perl5/ going away...?
(the "eh" was for the concept - you subsequent suggestion, $?BACKEND, is plausible)
does JS do lazy, or only P5? 23:45
fglock don't know 23:47
I'm trying to implement postfix:<...> - Inf works, but 1..Inf gives an empty list 23:49
putter err, whats the name of some $?FOO variable? (putter has to leave rsn and is drawing a blank)
fglock ? 23:50
putter got it. Run.hs. tnx
fglock bbiab & 23:57
putter fglock: 6552. 23:58
good night & 23:59

Powered by Cro and the Raku® Programming Language.

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

Thank you!