»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'p6: say 3;' or rakudo:, std:, or /msg camelia p6: ... | irclog: irc.perl6.org | UTF-8 is our friend!
Set by moritz on 25 December 2014.
00:02 ggherdov left
raydiak so I defined a submethod DESTROY for my class, but it doesn't ever get called, that I can tell...say, die, etc should do something in there, right? 00:05
00:05 ggherdov joined
raydiak m: class Foo { submethod DESTROY { say "oh noes" } }; Foo.new # idk how DESTROY should interact with the evalbot, but this doesn't work locally either 00:07
camelia ( no output )
ab5tract .tell Kristien reading the backlog i notice it sounded like i was saying i prefer a syntax choice over an optimal solution -- that's only in golfing, if then :) this was just an example from the article i linked to in the comment. i thought it was cool that it's been stable syntax since 2009
yoleaux ab5tract: I'll pass your message to Kristien.
ab5tract (if not well before 2009) 00:08
japhb: one heck of a golf indeed! i have a feeling that the ::Sci namespace in Perl 6 is going to be one very interesting realm of compunits 00:11
00:15 madcapJake joined 00:20 kurahaupo1 joined 00:22 pmurias left
masak pro tip: if you max= on a bunch of values and they all come out as '9', it's probably because they're strings from a SQL database and not numbers as you assumed. 00:34
00:36 muraiki_ joined
vendethiel *g* 00:38
00:40 yeahnoob joined 00:41 adu left 00:42 kurahaupo2 joined, kurahaupo1 left
raydiak idk if anyone actually tries using this stuff yet, but App::Pray and Image::PNG::Portable no longer segfault 00:43
00:44 BenGoldberg joined 00:54 raiph joined, BenGoldberg left, BenGoldberg joined 00:56 prime left
masak "You might actually not have that much time to try Perl6 [...] before it is declared production ready and becomes mainstream." -- programming.tudorconstantin.com/201...dable.html 01:04
BenGoldberg Love the pic on that page :) 01:28
01:40 yeahnoob left
japhb
.oO( Where's *my* jet-powered camel? )
 01:53
masak 'night, #perl6 01:55
01:55 kurahaupo2 is now known as kurahaupo
japhb o/ masak 01:58
02:00 espadrine left
skids I like the method in those slides. Show ugly, then "fix" it to what most people find acceptible, then surprise then by fixing it even more. 02:00
02:11 docdurdee joined
docdurdee p6: say 3 02:12
camelia rakudo-{parrot,moar} ac7ee6: OUTPUT«3␤»
docdurdee p6: say $_ for (0 .. 100)
camelia rakudo-{parrot,moar} ac7ee6: OUTPUT«0␤1␤2␤3␤4␤5␤6␤7␤8␤9␤10␤11␤12␤13␤14␤15␤16␤17␤18␤19␤20␤21␤22␤23␤24␤25␤26␤27␤28␤29␤30␤31␤32␤33␤34␤35␤36␤37␤38␤39␤40␤41␤42␤43␤44␤45␤46␤47␤48␤49␤50␤51…»
skids o/ 02:13
02:17 docdurdee left 02:31 chenryn joined
muraiki_ meow 02:34
02:34 muraiki_ left 02:37 IllvilJa left 02:38 fwilson left 02:44 fwilson joined 03:10 kurahaupo left 03:11 telex left 03:12 telex joined 03:17 dj_goku joined
colomon raydiak: Lastfm, eh? 03:23
raydiak colomon: more of a digitally imported/radiotunes guy myself 03:24
but maybe you meant github.com/avuserow/perl6-webservice-lastfm/ :) 03:25
colomon THere’s a www.last.fm/user/colomon
raydiak oh I see
colomon yes, that’s what I meant
raydiak do I have an account on there?
colomon wait, you’re totally right. how did I get confused and think it was by you?
raydiak lotta new modules lately :) 03:26
colomon oh, it was right next to Image::PNG::Portable in the smoke output.
simple off-by-one reading error. :)
raydiak maybe not as engaging as music, but I am excited to be able to output PNGs from Pray soon 03:27
03:29 raiph left
colomon heck, that’s more exciting than last.fm! 03:31
FROGGS_ so, that's happening here at night :o)
morning
raydiak \o FROGGS_
FROGGS_ yawns into his coffee 03:32
raydiak it's pretty thin in terms of features atm, but it works, and I'm testing a patch to moar right now so CArrays are zero-filled by default and I can get all the malloc/memset/free garbage out of my module too :) 03:33
03:34 BenGoldberg left
raydiak though I have to see if there's some way I can package a precompiled zlib.dll with my module and point compress::zlib at it for the windows folks, I guess 03:35
haven't actually tried windows yet but it's one of the main goals of course given the name
colomon o/ night 03:37
FROGGS_ gnight colomon
raydiak \o colomon
FROGGS_ raydiak: I also thought about shipping dll's to windows boxes...
(my libxml2 bindings) 03:38
03:38 kurahaupo1 joined
raydiak it'd be awesome if it became a common practice for nativecall modules, imo 03:39
03:39 vendethiel left
FROGGS_ aye 03:40
gtg 03:42
03:42 FROGGS_ left
raydiak \o/ my first moar patch ever (or even C, for that matter) seems to work on the first try! 03:43
03:43 IllvilJa joined 03:49 noganex_ joined
skids star: use NativeCall; my @f := CArray[int].new; @f[0] = 1; @f[1] = 2; @f.elems.say; @f[0].say; @f[1].say; @f[].say; @f[0,1].say 03:50
camelia star-{m,p} 2014.12: OUTPUT«1␤1␤2␤␤0␤»
03:52 noganex left
skids Those used to work I am pretty sure. 03:58
04:11 oetiker left 04:22 chenryn left 04:27 araujo left 04:28 araujo joined 04:38 adu joined 04:50 araujo left 04:51 FROGGS[mobile] joined
FROGGS[mobile] does somebody need more air? I'm at an air port... 04:52
does somebody know if it is as cheap and easy in Belgium to bye mobile broadband cards like it was in Austria? 04:58
05:00 mr-foobar left 05:08 chenryn joined, vendethiel joined 05:23 fil_ joined
fil_ hey 05:24
quick question
any1 here?
raydiak nope :)
moritz that *was* a quick question
fil_ yay!
TimToady ain't here neither
moritz another satisfied customer. NEXT!
fil_ so I've got a grammar going
I can't find good docs for how to work with Match objects 05:25
I figured out how to get what I want
but it looks messy
my $parsed = QFXXML.parse($xml); for @($parsed<itm>) -> $itm { my $pair = $itm.list().[0].caps().[0];
moritz fil_: doc.perl6.org/type/Match
fil_ Is there a shorter way to get at the $pair I'm interested in?
yes I found that doc
but it wasn't that helpful 05:26
moritz fil_: show the grammar please
fil_ grammar QFXXML { token TOP { ^ <itm>+ $ } token itm { ( <doctype> | <closetag> | <opentag> | <cdata> ) } token doctype { \< \? <-[>]>+ \> } token closetag { \< \/ <-[>]>+ \> } token opentag { \< <-[>]>+ \> } token cdata { <-[<>]>+ } }
I'm iterating over items
it all works
it's just that this part feels messy: my $pair = $itm.list().[0].caps().[0]; 05:27
FROGGS[mobile] an action class might be more elegant
fil_ is there a clearer / shorter way to get at what I'm after?
the code after it looks pretty 05:28
given $pair.key { when <opentag> {
...
I just don't like my $pair = $itm.list().[0].caps().[0];
it feels like a common scenario that there should be helper functions for this sort of thing 05:29
FROGGS[mobile] hmmm, that really sounds like you want an action class
fil_ can you point me to a doc?
they work with grammars?
FROGGS[mobile] not from my phone
moritz fil_: github.com/rakudo/star/blob/master...f?raw=true the chapter on grammars also talks about action methods 05:30
/ action classes 05:31
raydiak if you didn't want to write a whole actions class, could you put "{ make .list().[0].caps().[0] }" in token TOP or something along those lines?
s/in/at the end of/
FROGGS[mobile] m: grammar G { token foo { a +% b} }; class A { method foo($/) { say $/} }; G.parse( 'ababa', :actions(A))
camelia rakudo-moar ac7ee6: OUTPUT«No such method 'TOP' for invocant of type 'G'␤ in method parse at src/gen/m-CORE.setting:15013␤ in block <unit> at /tmp/PHFDBpQc8o:1␤␤»
fil_ ok 05:32
FROGGS[mobile] my..parse method wants a :rule('foo')
fil_ I will study actions
thanks for the pointer
I agree it sounds like Actions is what I'm looking for 05:33
FROGGS[mobile] and I cant copy+paste here
fil_ I knew I must be doing something wrong when I had to type my $pair = $itm.list().[0].caps().[0];
05:33 vendethiel left
fil_ thanks for the help! 05:33
moritz fil_: which token from the grammra do you want to extract?
FROGGS[mobile] and then you reinspect it, just after parsing its tokens :o) 05:34
05:34 fil_ left
raydiak yes if you don't use an actions class, you do get a real mess: github.com/raydiak/Math-Symbolic/b...c.pm6#L918 05:35
FROGGS[mobile] messy like my P5 code often looks 05:36
raydiak wow yer right...it looks a lot like my P5 code did; never noticed before
FROGGS[mobile] :o) 05:41
05:52 adu left
raydiak working on C code feels like reaching my hand into muddy opaque crocodile-inhabited waters compared to P6 05:57
05:59 FROGGS[mobile] left 06:00 baest left 06:19 xfix joined 06:42 Rounin joined 06:46 tinyblak joined 07:07 jack_rabbit joined 07:15 noganex joined 07:18 espadrine joined, noganex_ left
tadzik to me it always feels like chopping wood with my bare hands 07:20
raydiak I like that; seems more thoroughly accurate 07:24
(mainly because sticking your hand in muddy water isn't very rewarding most of the time) 07:25
07:27 kurahaupo1 left 07:31 rurban_ joined 07:34 kaleem joined, sqirrel joined 07:48 jack_rabbit left 07:55 mvuets joined 08:00 yeahnoob joined
jnthn Not to mention if you bring your axe down off the end of the log, you chop your foot off... :) 08:08
yoleaux 29 Jan 2015 23:49Z <raydiak> jnthn: #123700
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=123700
jnthn raydiak++
08:10 bjz joined 08:11 virtualsue joined 08:12 darutoko joined
jnthn wanders off to the airport, for le FOSDEM trip & 08:14
08:15 bjz left, chenryn left 08:18 bjz joined 08:19 chenryn joined, IllvilJa left 08:25 Kristien joined 08:26 petercommand left, BigBear joined 08:27 petercommand joined 08:28 IllvilJa joined 08:31 bjz left
pmichaud bonjour, #perl6 08:38
lizmat pmichaud: good morning 08:41
.tell raydiak re DESTROY not working: that is exactly why I think it is a bad name 08:42
yoleaux lizmat: I'll pass your message to raydiak.
lizmat .tell raydiak because atm DESTROY only gets called on GC cleanups
yoleaux lizmat: I'll pass your message to raydiak.
lizmat .tell raydiak and if there isn't one, then it just won't get called 08:43
yoleaux lizmat: I'll pass your message to raydiak.
pmichaud maybe "GOODBYE" instead of "DESTROY"? ;-)
raydiak ah, so there is *no* reliable way to do that? 08:44
yoleaux 08:42Z <lizmat> raydiak: re DESTROY not working: that is exactly why I think it is a bad name
08:42Z <lizmat> raydiak: because atm DESTROY only gets called on GC cleanups
08:43Z <lizmat> raydiak: and if there isn't one, then it just won't get called
lizmat raydiak: there is, with phasers
but not in the class itself
now, I think we need a way to export a phaser to a caller
raydiak hrm 08:45
lizmat but on the other hand, that might be too much action at a distance
you could for instance do a:
raydiak why not just always call DESTROYALL?
lizmat LEAVE $dbh.?disconnect
you mean at program exit ? 08:46
raydiak aye 08:47
pmichaud Wait, I know! The DESTROY method should be called STAY-PUFT
lizmat probably because in most of the case you don't need it
there was a discussion recently on p5p where it was shown that the teardown of a Perl process in a forked environment
pmichaud that's the form of the destructor as named in Ghostbusters :) 08:48
lizmat can be really slow because the teardown dirties COW pages in the OS
08:48 fhelmberger joined
raydiak so performance reasons...hm 08:48
lizmat so, again, why I think DESTROY is a very bad name for the functionality being offered in Perl 6
raydiak well it does exactly what we need except for not being reliable, and much more cleanly than putting the onus of cleaning up my mess on the caller 08:49
lizmat raydiak: but what is it that you need to clean up?
raydiak so maybe we need a copy of it called something else which *is* reliable
lizmat memory?
database connections?
raydiak memory 08:50
lizmat but why would you care?
the system should be able to do that for you
the GC of the system :-)
raydiak b/c I nativecall malloc'd it
and so need to free it
lizmat so we would need to somehow inform the GC that it needs to free it when it need it
*needs to 08:51
08:51 bjz joined
raydiak maybe by giving it something to hook into...like a DESTROY method but called something else? 08:51
lizmat too bad jnthn is travelling atm, he could chime in on this
08:51 jack_rabbit joined
lizmat well, in this case, I would more think along a special malloc that hooks into the GC somehow 08:52
pmichaud that sounds special-case-y to me
raydiak but I think saying "you don't ever need to reliably call something at exit" feels very wrongish
08:53 kjs_ joined
pmichaud I think there ought to be a generalized "you're about to be destroyed" method that lets an object take care of any cleanup it needs done 08:53
lizmat pmichaud: there Is: it's called DESTROY 08:54
but the timing of it, and whether it really runs, are unclear
08:55 Otterpocket joined
pmichaud well, for most objects there probably doesn't need to be a DESTROY 08:55
I could see it being a role... something like Destroyable
raydiak in my specific case, I'm hacking around a series of limitations which are all temporary, but it has revealed a more serious limitation, imo...we really need to be able to do this in a composable way, which phasers don't seem to be, on this axis
pmichaud so, objects that don't have the Destroyable role don't bother with a DESTROY call on destruction
08:56 abraxxa joined
raydiak I like it 08:56
pmichaud that way the check for whether or not to invoke DESTROY could be made quicker
arnsholt I've some thoughts on how to handle the whole "should this memory be freed on GC" problem in NativeCall, but nothing very implementable just yet
lizmat m: my $d; class A { method DESTROY { $d++ } }; for ^10000 { A.new }; say "seen $d DESTROYs"
camelia rakudo-moar ac7ee6: OUTPUT«seen 6566 DESTROYs␤»
lizmat pmichaud: afaik, the existence of a DESTROY method at compose time, already sets such a flag 08:57
m: my $d; class A { method DESTROY { $d++ } }; for ^10000 { A.new }; say "seen $d DESTROYs"
camelia rakudo-moar ac7ee6: OUTPUT«seen 6566 DESTROYs␤»
pmichaud okay, that can work also, but somehow I think it might be better exposed as a role 08:58
timely destruction is something we abandoned/postponed long ago
lizmat fwiw, I think the existence of the method, is flag enough
JimmyZ 200~m: my $d; class A { method FINALIZE { $d++ } }; for ^10000 { A.new }; say "seen $d DESTROYs"
mathw morning
JimmyZ m: my $d; class A { method FINALIZE { $d++ } }; for ^10000 { A.new }; say "seen $d DESTROYs"
camelia rakudo-moar ac7ee6: OUTPUT«use of uninitialized value $d of type Any in string context␤seen DESTROYs␤»
moritz pmichaud: I've even tried to document it: doc.perl6.org/type/Metamodel%3A%3AFinalization 08:59
JimmyZ m: my $d; class A { method FINALIZE { state $++ } }; for ^10000 { A.new }; say "seen $d DESTROYs"
camelia rakudo-moar ac7ee6: OUTPUT«===SORRY!=== Error while compiling /tmp/q29k1_jXpb␤Preceding context expects a term, but found infix + instead␤at /tmp/q29k1_jXpb:1␤------> ; class A { method FINALIZE { state $++ ⏏} }; for ^10000 { A.new }; say "se…»
raydiak good morning mathw
arnsholt pmichaud: Yeah, timely destruction (and guaranteed order) restricts the options for GC strategy quite a bit, no?
moritz well, for timely we have LEAVE block 09:00
tadzik yeah, I was wondering if guaranteed order is still something to aim for
lizmat arnsholt: I think the order follows out of the reachability constraint that is there for getting finalized anyway
pmichaud somewhere very long ago... almost back to pugs days I think ... it was decided that Perl 6 wouldn't bother with timely destruction (for now?)
mathw Whenever destruction comes up in the context of a language with GC I suddenly find myself wanting to be a C++ programmer again.
09:00 jack_rabbit left
lizmat moritz: or the specced (and half implemented) "will leave" clause on variables 09:01
09:01 sqirrel left
moritz lizmat: right 09:01
lizmat my $dbh will leave { .disconnect };
pmichaud I'm afk for a bit
mathw lizmat: that is very cool. I like that.
moritz python has with dbh = connect(): do stuff
raydiak often you don't care about any of that, just that you know for sure it will get called exactly once unless something *really* catastrophic happens 09:02
mathw C# has using() blocks that call Dispose on things when they go out of scope
Which alleviates most of the pain
lizmat mathw: how is that different from LEAVE ? 09:03
Kristien mathw: F# has use variables, which I like better since they don't require a new scope
yoleaux 00:07Z <ab5tract> Kristien: reading the backlog i notice it sounded like i was saying i prefer a syntax choice over an optimal solution -- that's only in golfing, if then :) this was just an example from the article i linked to in the comment. i thought it was cool that it's been stable syntax since 2009
moritz lizmat: it's the same semantic, just different presentation
Kristien for example:
use $foo = Bar.new; 09:04
raydiak I guess I'm wondering what DESTROY *is* meant to be useful for
Kristien it would also be useful if this were possible for attributes
09:04 mvuets left
moritz raydiak: avoiding resource leaks in the long run 09:04
mathw lizmat: it's nowhere near as flexible and it has an annoying syntax.
Kristien: yes, F#'s version is preferable to me as well, but I have yet to convince my colleagues of the value of the language as a whole. 09:05
Many of them think C# is the cutting edge
raydiak moritz: how is it useful for that if it's not reliable?
except as a "mitigate the damage enough to make it bearable" kinda measure :/ 09:06
moritz raydiak: in your exmaple, the program terminated before all of them were DESTROYed
09:06 zakharyas joined
mathw But I do like the idea that 'will leave' gives you of providing the cleanup code right there where you do the thing that needs cleaning up later, so it's all together. Phasers are a concept I'm growing to like more the more Perl 6 I've been writing lately. 09:06
moritz raydiak: and when a program terminates quickly, it can't leak resources 09:07
raydiak: if it ran longer, all of them would be collected.
lizmat to give you a datapoint:
raydiak even if it doesn't free memory, it won't leak?
moritz raydiak: which care are you talking about?
lizmat in mod_perl, there is a special setting that will tear down a child process *without* running any of the normal Perl (5) teardown
moritz raydiak: program exit frees all memory 09:08
lizmat at former $work, I've seen this make the difference between being able to run the site and not
raydiak yes perhaps I'm getting things tangled...tis 1 AM here
lizmat because the normal teardown of a process was in the ~20 seconds range (yes, twenty whole seconds)
Kristien mathw, I thought it was about Perl 6
lizmat taking 100% of CPU during the teardown, as it was all CPU bound 09:09
Kristien I should give Boehm GC a try. 09:10
09:10 jack_rabbit joined 09:12 kjs_ left
lizmat In a way, the timely / unknown when discussion about finalization, reminds me of the discussion of driving a stick shift or an automatic 09:12
driving a car with a stick shift, may make you feel more in control 09:13
but whichever way you look at it, it is a hack around the fact that a petrol engine cannot give enough power at some rotational speeds
moritz and these days, even Formula 1 pilots drive with automatic 09:14
(iirc)
lizmat let's say that Perl 6 is an automatic, and Perl 5 has a stick shift :-)
raydiak so perhaps make it harder or at least more weird-looking to do, but are you really trying to tell me that it's not important to let people do that reliably, at all, for any reason ever, in a module instead of the code using the module?
lizmat indeed, yes, generally you shouldn't care 09:15
although Perl 6 is not a CVT, it's still an automatic
tadzik hehe, the way I always look at it is similar to a "should I write my web crawler in C or Perl". Of course the one in C will be faster, but it's just a web crawler, and Perl tools will do better what I'd do wrong in C
so why would I drive to the supermarket with manual? :)
I'm not a racecar drivre
lizmat In about an hour, we'll be picking up a van for moving a lot of stuff to FOSDEM 09:16
the van has a nice "feature"
it shows a light on the dashboard when you're supposed to shift up or down a gear
to me, that it just stupid: if the car knows it should have a higher/lower gear, why doesn't it switch itself? 09:17
moritz and my reaction to that would be "why don't you f$#$#ing do it yourself, smarta$$?" :-)
lizmat indeed
Kristien tadzik, the first rule of C is that you should never use C for anything 09:18
raydiak given that my problem only exists because I've already gone a bit off-road with nativecall mallocing stuff I only half understand to get around problems which won't exist, and I don't have a good counter example off the top of my head...I'll concede the point for now and meditate on this wisdom :) 09:19
lizmat :-)
09:21 kjs_ joined
raydiak but so even in C, if you don't free stuff, the OS reliably reclaims your memory when the program exits? I am fuzzy on these sorts of things 09:21
lizmat yes, the OS is expected to
Kristien that's language-agnostic
lizmat the only exception to that, are unclaimed zombies, afaik
but it has been a while since I've seen any of those 09:22
Kristien no
zombie processes have their memory freed
mathw Memory leaks are only an issue if your process stays alive. 09:23
Kristien their only overhead is an entry in the process table that contains their PID, exit status and a few flags
raydiak hm so I should just use DESTROY then and not worry if it doesn't get called, like how I was going to :P
09:24 kjs_ left
Kristien and wait() deletes the entry by violating CQS 09:25
TuxCM www.xs4all.nl/~hmbrand/t.pl <= what is the perl6 syntax to do self!$m($x); (and will self!can($m) work as I want) 09:26
raydiak this all makes sense I would have noticed "magical diasappearing memory" as a user...I guess jumping in to C just reminds me of that "don't assume ANYTHING" feeling 09:27
09:27 kjs_ joined
Kristien it means you jump into the realm of unnecessary bugs like heartbleed 09:28
TuxCM ! => . → Cannot find method 'postcircumfix:<( )>' 09:29
moritz TuxCM: you want to call a private method with a variable name? 09:30
TuxCM: that would be self!"$name"(arguments), I think
TuxCM I want to call a public/private method *named* $m with variable $v
moritz m: class A { method call() { my $name = "priv"; self!"$name"(42) }; method !priv($x) { say 42 } }; A.new.call 09:31
camelia rakudo-moar ac7ee6: OUTPUT«42␤»
Kristien raydiak: the only real reason to use C for anything is because you have to maintain something written in C. for anything else you should always use higher-level tools that give stronger guarantees and require less effort to get stuff to work
TuxCM if (self.can ($m)) {
self."$m"(%init{$attr});
*that* works 09:32
message to FROGGS: «self."$m" (%init{$attr});» does not work under Tuxic
moritz TuxCM: fwiw without the quotes, it would try to use $m as a code object 09:33
09:33 molaf__ joined
moritz TuxCM: no whitespace allowed before the opening paren 09:33
m: 42.&say()
camelia rakudo-moar ac7ee6: OUTPUT«42␤»
TuxCM whitespace is allowed under Slang::Tuxic
moritz oh
09:33 Kristien left
TuxCM That is why I am active in perl6 again 09:33
09:34 espadrine left
moritz didn't know 09:35
09:36 alpha- joined, molaf_ left
raydiak good night, #perl6 09:37
09:38 sqirrel joined
moritz good night, raydiak. Dream of timely destruction! 09:42
09:45 chenryn left 09:48 Kristien joined, pnu left, pnu joined
tadzik hehe, out of context that sounds pretty grim :P 09:50
moritz uhm, right, I wasn't aware of that :-) 09:51
jnthn The rules on DESTROY are that it's called at most once, upon the GC deciding the object is unreachable (with the exception of it being on the queue of things that need a DESTROY call) 09:57
If you never reach the point of GCing the object, the call will never happen. 09:58
We don't attempt to go and GC the world at shutdown, because it's a huge waste of time. The OS can clean up stuff way faster than we can.
And if your DESTROY makes the object reachable again, then you won't get your DESTROY called a second time when it becomes unreachable. 09:59
"You only live twice"
09:59 rindolf joined
moritz jnthn++ 09:59
nwc10 jnthn++ 10:01
TuxCM how does p5 «my $v = "x"; my $w = "x${v}x"» translate to p6? ${v} is obviously not the correct syntax
I know I can use "x"~$v~"x" 10:02
but I wonder about interpolation syntax
moritz "x{$v}y" 10:03
code blocks in general interpolate
so you could also write "x{ $v }y" and get the same result
lizmat class A { method foo { say "foo" } }; my $a = A.new; $a.can("foo")[0]($a) # use the result of .can directly, like in P5
TuxCM Top, I'll try to remember. It does have a lot of logic
lizmat m: class A { method foo { say "foo" } }; my $a = A.new; $a.can("foo")[0]($a) # use the result of .can directly, like in P5
camelia rakudo-moar ac7ee6: OUTPUT«foo␤»
jnthn Looks like it's nearly boarding time :) 10:04
See some of you in Brussels :)
lizmat have a safe flight!
TuxCM safe flight, see you
moritz good journey!
10:05 espadrine joined
lizmat m: class A { method foo { say "foo" } }; my $a = A.new; .($a) for $a.can("foo") # more correct use of .can 10:06
camelia rakudo-moar ac7ee6: OUTPUT«foo␤»
TuxCM lizmat, Type check failed in binding <anon>; expected 'Text::CSV' but got 'Bool' 10:07
10:07 labster left, labster_ joined
TuxCM my @can = self.can ($m); 10:07
if (@can.elems > 0) {
@can[0](%init{$attr}); # space before ( not (yet) allowed under Tuxic
next;
lizmat that code seems wrong to me 10:08
why check for more than 1 elem, and then only do the first ?
TuxCM my @can = self.can ($m); 10:09
if (@can.elems > 0) {
.(%init{$attr}) for @can; # space before ( not (yet) allowed under Tuxic
next;
even longer error messgae starts with
Type check failed in binding <anon>; expected 'Text::CSV' but got 'Bool'
wait, ($a) passes self, not a variable 10:10
lizmat ah, yes
that's the problem
you need to pass the object as the first parameter 10:11
in that sense, P6 is just the same as P5 :-)
TuxCM if (@can.elems > 0) {
.(self, %init{$attr}) for @can; # space before ( not (yet) allowed
lizmat .(self,%init{$attr})
TuxCM ok, nice enough. tests pass
masak good antenoon, #perl6 10:12
lizmat m: class A { method foo { say "foo" } }; my $a = A.new; $a.$_() for $a.can("foo") # aaah, you *can* use the code object as a method :-) 10:13
camelia rakudo-moar ac7ee6: OUTPUT«foo␤»
TuxCM both «$_.($v) for @can» and «$_!($v) for @can» fail though 10:16
10:16 yeahnoob left
TuxCM «.(self, %init{$attr}) for @can;» is good nuf for me 10:17
10:17 FROGGS[mobile] joined
FROGGS[mobile] hotel /o/ 10:17
nwc10: is that you in that little room? 10:18
10:18 thilp left
nwc10 no, I'm in Vienna 10:20
FROGGS[mobile] nwc10: must be your lost twin then O.o 10:21
TimToady there are no little rooms in Vienna?
10:21 pecastro joined
nwc10 I've seen at least one, but I was guessing that it wasn't "that" little room 10:22
amd right now I seem to be the only person in a room with 8 desks
FROGGS[mobile] that room here has one table but eight ppl 10:23
10:24 espadrine_ joined 10:25 espadrine left 10:30 kjs_ left, jack_rabbit left, espadrine_ is now known as espadrine 10:33 Kristien left 10:34 mvuets joined 10:35 FROGGS joined 10:36 FROGGS[mobile] left 10:40 vendethiel joined
timotimo t/spec/S02-types/multi_dimensional_array.rakudo.moar .......... Dubious, test returned 255 (wstat 65280, 0xff00) 10:44
am i perhpas on an outdated rakudo?
moritz timotimo: I think I saw that problem yesterday too 10:45
timotimo: I was working on a (fresh) branch, so I wasn't sure if maybe the branch was to blame
timotimo 'k 10:46
lizmat I think it's Mouq's recent LoL patches that are to blame
timotimo has some things to look at here and there to pass him some time
lizmat either 5275eb90c15cd35 or 166cd3c5ed3cf9d8 10:47
10:49 abraxxa left 10:50 andreoss joined 10:55 spider-mario left
lizmat afk fior a bit& 10:56
dalek kudo/nom: e4883a5 | timotimo++ | src/core/Cursor.pm:
eqat is nicer than substr + eq.
 10:57
TuxCM can I check the context of the caller in a method like perl5's caller? I'd likke to check if I am Void/Int|Num/Str/List|Array/Hash 10:58
nwc10 timotimo++ 11:00
timotimo: git grep 'substr.*\beq\b'
two more to get in src/Perl6/Actions.nqp ?
FROGGS m: class Foo { method Numeric { say "I'm called in Num context!"; 42 } }; say 1 + Foo.new
camelia rakudo-moar ac7ee6: OUTPUT«I'm called in Num context!␤43␤»
11:01 tinyblak left
FROGGS TuxCM: that's how context works in Perl 6 11:01
11:01 tinyblak joined
FROGGS timotimo / nwc10: I intentionally left some substr's untouched... I think eqat does not queck bounds, so that can cause trouble 11:02
11:04 tinyblak left, tinyblak joined
FROGGS hmmm, seems I am wrong 11:05
TuxCM FROGGS, does code explain better? www.xs4all.nl/~hmbrand/t.pl 11:10
timotimo nqp: my $hs = "hello world"; my $ndl = "world"; say(nqp::eqat($hs, $ndl, 6))
camelia nqp-jvm: OUTPUT«(signal ABRT)*** Error in `java': double free or corruption (!prev): 0x00007f230401a8d0 ***␤*** Error in `java': double free or corruption (!prev): 0x00007f230401a8d0 ***␤#␤»
..nqp-parrot: OUTPUT«Assignment ("=") not supported in NQP, use ":=" instead at line 2, near " \"hello wo"␤current instr.: 'panic' pc 15755 (gen/parrot/stage2/NQPHLL.pir:5751) (gen/parrot/stage2/NQPHLL.nqp:425)␤»
..nqp-moarvm: OUTPUT«Assignment ("=") not supported in NQP, use ":=" instead at line 2, near " \"hello wo"␤ at gen/moar/stage2/NQPHLL.nqp:485 (/home/camelia/rakudo-inst-1/languages/nqp/lib/NQPHLL.moarvm:panic:105)␤ from <unknown>:1 (/home/camelia/rakudo-inst-1/languages/nqp/li…»
timotimo nqp: my $hs := "hello world"; my $ndl := "world"; say(nqp::eqat($hs, $ndl, 6))
camelia nqp-jvm: OUTPUT«(signal ABRT)*** Error in `java': free(): invalid pointer: 0x00007f1cc801d750 ***␤#␤# There is insufficient memory for the Java Runtime Environment to continue.␤»
..nqp-{moarvm,parrot}: OUTPUT«1␤»
timotimo nqp: my $hs := "hello worl"; my $ndl := "world"; say(nqp::eqat($hs, $ndl, 6))
camelia nqp-jvm: OUTPUT«(signal ABRT)#␤*** Error in `java': double free or corruption (fasttop): 0x00007f7d30013450 ***␤# There is insufficient memory for the Java Runtime Environment to continue.␤*** Error in `java': double free or corruption (fasttop): 0x00007f7d3001d160 ***␤# ***…»
..nqp-{moarvm,parrot}: OUTPUT«0␤»
moritz TuxCM: try 'list' instead of 'Array', 'sink' instead of 'Void', 'hash' instead of 'Hash' 11:11
though .keys doesn't trigger hash context
m: say <a b c>.keys
camelia rakudo-moar ac7ee6: OUTPUT«0 1 2␤»
FROGGS moritz / TuxCM: though, it wont flatten... you cannot iterate over it just like that
11:12 virtualsue left
FROGGS m: class Foo { method list { say "I'm called in Array context!"; 42, 43 } }; .say for Foo.new 11:12
camelia rakudo-moar ac7ee6: OUTPUT«Foo.new()␤»
TuxCM FROGGS, that is list for the *class* in context 11:13
what about the context of a *method*
moritz TuxCM: methods aren't called in contexts
TuxCM method Numeric foo () {} is invalid
timotimo only return values are evaluated in contexts
moritz TuxCM: context doens't flow inwards like in perl 5 11:14
timotimo so you'd return an object that'd have methods for the contexts you're interested in handling differently
TuxCM Tue is PM, I'll ask Liz to do some explaining
FROGGS what we call contexts in Perl 6 is what you might have done with 'use overload' in Perl 5 11:15
TuxCM www.xs4all.nl/~hmbrand/t.pl <= like this? Why do String and list do not what I expect? 11:21
(the rest does)
timotimo you want .Str or .Stringy
but not .String
TuxCM top! 11:22
and list?
timotimo may want to have "is Iterable" or "does Positional" or something similar?
i'm not experienced in making listy objects, but there was an advent calendar post on that exact topic
11:25 sqirrel left 11:40 abraxxa joined
colomon is completely baffled by TuxCM’s code 11:42
TuxCM :) 11:43
can a class inside a class access the private variables of the surrounding class?
m: class Foo { has Int $!x = 0; class Bar { method Numeric { return $.x; }}; method bar { return Bar.new; }; say 0 + Foo.new.bar; 11:45
camelia rakudo-moar e4883a: OUTPUT«===SORRY!=== Error while compiling /tmp/_ekcmJ33gg␤Unable to parse expression in block; couldn't find final '}' ␤at /tmp/_ekcmJ33gg:1␤------> return Bar.new; }; say 0 + Foo.new.bar;⏏<EOL>␤»
tadzik what was this opcode to force a GC run?
colomon TuxCM: pretty sure that has to be return $!x
TuxCM no, as that would be the $!x of Bar 11:46
masak TuxCM: no, you can't reach the `self` of an outer class.
TuxCM that much I understand of OO :)
damn
masak TuxCM: my workaround tends to be to pass (an instance of) outer class as a parameter (or drop it in a contextual) 11:47
tadzik ah, nqp::force_gc that is
TuxCM my current thought are to pass all that is needed to fullfill my context wish to the subclass
masak but there's no analogue of Java's OuterClass::self, for example.
heh, if that were indeed the way Java wrote it ;) 11:48
they write OuterClass.this
timotimo yo japhb what's your status? 11:49
i suspect "sleeping" or "at work" or something ,so i'll .tell
colomon TuxCM: Stringy certainly works where you had String before 11:50
timotimo .tell japhb i'd like to have an integrated tool for perl6-bench that can build a version of a compiler and the version before it, or alternatively you could pass a little patchset and it'd build the things with and without the patchset, time all things and immediately spit out a comparison graph file that might even contain the diff
yoleaux timotimo: I'll pass your message to japhb.
11:51 BigBear left 11:53 kaleem left
TuxCM got it working as I want. /me collects the hoops that were not caught in this circus act 11:54
11:55 abraxxa left, abraxxa joined
TuxCM colomon, in what way baffled? 11:55
colomon TuxCM: I was able to get the list case working with class Error::Diag is Iterable and method iterator { return [1, 2, 3].iterator; }
TuxCM: didn’t understand what you were trying to do 11:56
have it working on my machine now.
colomon just woke up...
TuxCM one moment, I'll push
timotimo do you actually need to return ....iterator?
11:58 rindolf left
colomon timotimo: I didn’t try it the other way, but my naive assumption is that the iterator method should return an iterator and not just an Iterable 11:58
TuxCM doesn't work here with or without .iterator 11:59
colomon, github.com/Tux/CSV test-t.pl + t/10_base.t should show what I want 12:01
make test
colomon ah, yes, that makes more sense now 12:04
though I’m not sure I see the logic of having list and hash options there. wouldn’t it be easiest / more idiomatic to skip the list interface altogether, and make the hash names correspond to the actual attribute names and then not have the hash interface either? 12:06
12:09 kjs_ joined
TuxCM the iterator doe not work on $csv.error_diag[2] 12:10
12:10 andreoss left 12:11 BigBear joined
TuxCM nor on ($csv.error_diag)[2] 12:11
12:11 chenryn joined
TuxCM it does however on my @ed = $csv.error_diag; is (@ed[2], ... 12:12
colomon TuxCM: right, you’d need does Positional and at_key (or is it at-key now?) for ($csv.error_diag)[2] 12:13
moritz (but please tell me you want to return an object from .error_diag...) 12:14
colomon but again, why would you ever want to do that? wouldn’t $csv.error_diag.pos be 100% clearer always
?
moritz aye, the positional interface looks funny to me
like "objects are weird and expensive" in p5 :-) 12:15
TuxCM I want that TOO
moritz listens to Queen: "I want it TOO"
TuxCM but I first want to create an interface that would make the transition from Text::CSV_XS as transparent as possible
12:16 chenryn left, kjs_ left
colomon ah, compatiability with crappy old interfaces is a valid design concern. :) 12:19
moritz fwiw, in p5 I might have done the interface similarly 12:21
12:21 kjs_ joined 12:28 molaf__ left 12:35 telex left 12:36 telex joined 12:38 rmgk_ joined, rmgk left, rmgk_ is now known as rmgk 12:39 Kristien joined
Kristien I want to make a Perl 6 backend for IPython. 12:40
12:41 kjs_ left
ab5tract Kristien: i've been thinking along those lines as well... have you made any progress hacking on the REPL? i seem to remember you saying that you had a few ideas along those lines 12:45
Kristien No. 12:48
But REPLs that can show graphics and offer syntax highlighting are very useful.
and with re-evaluation and such 12:49
arnsholt IIRC timotimo++ experimented with a Perl 6 IPython kernel a while back 12:50
Kristien I often make errors in REPL sessions, and it's annoying when I cannot undo them
12:50 skids left
Kristien especially when I plan to archive the REPL session later 12:50
timotimo aye 12:53
i had trouble finishing it up because we didn't have multithreading at that point
and i was a tiny bit reluctant to implement the ipython protocol, as it has some crypto to it
github.com/timo/iperl6kernel 12:54
the iperl6kernel.nqp approach to implementing it isn't super nice either 12:55
Kristien cool 12:56
12:58 IllvilJa left 12:59 [Sno] left
vendethiel m: nqp::::get_interp__P() 12:59
camelia rakudo-moar e4883a: OUTPUT«===SORRY!=== Error while compiling /tmp/ABtDeFkuoe␤Name component may not be null␤at /tmp/ABtDeFkuoe:1␤------> nqp::⏏::get_interp__P()␤»
vendethiel misread
moritz quad colon is a bit exaggerated, IMHO :-) 13:01
timotimo get_interp? do you perhaps want getcomp?
moritz iirc in PIR it's getinterp or something like that 13:02
timotimo ah 13:03
13:03 rurban_ left
timotimo but why would you want a pir op on moarvm? 13:03
vendethiel was just trying to get a parse :) 13:04
timotimo t.h8.lv/p6bench/2015-01-30-six_mont...akudo.html - random benchmark results 13:05
t.h8.lv/p6bench/2015-01-30-six_months_nqp.html - for nqp as well
moritz while_int2str_concat is funny 13:06
on moar it's faster to do 100k iterations than 50k? 13:07
13:08 abraxxa left
timotimo not guaranteeing my desktop at home is 100% noise-free ... 13:09
moritz it's just.. a logarithmic scale, right? 13:10
and it's a factor 2 faster two do twice as many iterations
13:10 IllvilJa joined
timotimo i hadn't realized rakudo-moar was this far ahead on many of our microbenchmarks 13:10
moritz and it shows a trend that's also confirmed in the _native
that doesn't look like random noise to me 13:11
does it start a new process for for each number of iteration?
timotimo aye 13:12
and it runs each benchmark 4 times
moritz then it's not the JIT warming up
timotimo i suppose some part of the benchmarks begins to dominate the algorithmic complexity at some tipping point? 13:13
moritz oh wait, number of seconds per iteration
that means it divides by the number of iterations that's on the x axis
that means if the second 50k iterations are basically free (because the JIT kicks in), that's what we'd see
all those 2d plots basically measure the startup time of a benchmark vs. the actual run time 13:14
if the startup time is low, you get a flat line
if it's high, you get a straight line with slope 1 13:15
timotimo we subtract startup time, though 13:16
moritz well, I don't mean the compiler startup time
rather the one-time overhead of the benchmark code 13:17
if you have sub bench-for($x) { 1 for ^$x } you have a one-time cost for calling, returning, and creating the range (for example)
if that dominates the actual benchmark, you get a line with slope close to 1
like, if the call costs 5ms and the iteration is 1ms, you get 6ms for 1 iteration 13:18
erm
lizmat commute to Brussels&
moritz and then 7ms for 2 iterations
13:19 lizmat left
moritz which whould be 3.5ms for 1 iteration 13:19
... and slope close to -1
moritz-- # can't do math anymore
colomon Is it just me, or did rakudo-moar’s progress in the last six months force the system to add a line or two to the top of most of these graphs?
13:20 kaleem joined 13:21 BigBear left
timotimo moritz: actually, bench calculates the start-up time for each benchmark separately based on what happens when you pass a workload of 0 13:22
moritz clever 13:25
timotimo that's japhb for you :) 13:26
moritz japhb++
13:31 kjs_ joined 13:33 rurban_ joined 13:34 BigBear joined 13:43 brrt joined 13:44 Kristien left, kaleem left 14:03 brrt left 14:04 woolfy1 left 14:09 Kristien joined 14:20 Kristien left 14:26 rindolf joined 14:35 tinyblak left, skids joined 14:36 tinyblak joined
TuxCM $ perl6 --profile test-x.pl </tmp/hello.csv 14:38
50000
===SORRY!===
Don't know how to dump a Str
that 50000 is the last line of my script «$sum.say»
is profiling broken? 14:39
masak makes a wild guess that some string or something is too large
14:44 raiph joined 14:45 rurban_ left 14:46 rurban_ joined
masak m: repeat -> $n { say $n } while (^5).roll 14:47
camelia rakudo-moar e4883a: OUTPUT«(Mu)␤1␤1␤3␤4␤»
masak weird :)
masak is reading S04
m: repeat -> $n { say $n }␤while (^5).roll 14:48
camelia rakudo-moar e4883a: OUTPUT«(Mu)␤1␤3␤3␤3␤2␤3␤4␤1␤2␤1␤1␤3␤2␤4␤1␤»
masak I was curious whether that worked. it does. because the while/until is required, so the '}\n' can't end the statement just yet.
14:52 rurban joined 14:58 fwilson left
pmichaud bon après-midi, #perl6 14:59
mst waves from brussels
vendethiel bonjour, pmichaud!
jnthn is also in Brussels by now :) 15:00
JimmyZ \o/ 15:16
15:19 fhelmberger left 15:22 tinyblak left 15:23 tinyblak joined
TuxCM /pro/3gl/CPAN/rakudobrew/bin/perl6: line 2: 27389 Segmentation fault 15:26
is that something you want to analyse?
15:27 Mouq joined
colomon TuxCM: how did you get the segmentation fault? 15:27
TuxCM perl6 -I. t/55_combi.t 15:28
15:28 tinyblak left
TuxCM many tests to FAIL, that is what I am trying to fix 15:28
Mouq TuxCM: Re class Foo { method list { }} not working: That's by design. Iterating with `for` or `map` shouldn't be calling .list on each of its arguments. For this reason, if you want to make sure it's treated like a list, the way to get that is @$foo or $foo.list 15:29
Same with .hash
15:30 fhelmberger joined
TuxCM at_pos and at_key are my frieds, right? 15:30
15:31 alinbsp joined
Mouq The same thing happens with, e.g., $0, $<foo>, etc. Trying `for $<foo> { … }` should just run the loop once. Doing `for @<foo> { … }` will do what's intended 15:31
15:31 brrt joined
TuxCM lists are still surprising me 15:31
Mouq TuxCM: I believe those work regardless, yeah
PerlJam TuxCM: you may have to wait for the GLR before lists are less surprising 15:32
TuxCM «@foo = (1, "foo", 3, Str); for @foo -> $foo {}» ok, but «for (1, "foo", 2, Str) -> $foo {}» is not
Mouq Even after the GLR, $ => item, @ => list, % => hash :P 15:33
15:33 Rounin left
Mouq m: for (1, "foo", 2, Str) -> $foo { say $foo.perl } 15:33
camelia rakudo-moar e4883a: OUTPUT«1␤"foo"␤2␤Str␤»
Mouq TuxCM: What's the problem? 15:34
TuxCM m: for (1, "foo", 3.14, Str) -> $fo { .say; }
camelia rakudo-moar e4883a: OUTPUT«Nil␤Nil␤Nil␤Nil␤»
Mouq Or are you expecting $foo to be (1, "foo", 2, Str) ?
TuxCM m: for (1, "foo", 3.14, Str) -> $fo { $fo.say; }
camelia rakudo-moar e4883a: OUTPUT«1␤foo␤3.14␤(Str)␤»
TuxCM tries again
Mouq m: for $(1, "foo", 2, Str) -> $foo { say $foo.perl }
camelia rakudo-moar e4883a: OUTPUT«$(1, "foo", 2, Str)␤»
Mouq m: for [1, "foo", 2, Str] -> $foo { say $foo.perl } 15:35
camelia rakudo-moar e4883a: OUTPUT«[1, "foo", 2, Str]␤»
TuxCM must have been another error (too) that blurred my view 15:36
15:38 tinyblak joined 15:39 tinyblak_ joined 15:41 Alina-malina left
TuxCM ok (my @ret = $csv.fields, "fields"); <= that surprised me. "fields" was pushed to @ret too 15:42
15:42 tinyblak left
pmichaud TuxCM: what would you have expected there? 15:43
15:43 Alina-malina joined, fhelmberger left
TuxCM ok ((my @ret = $csv.fields), "fields"); 15:43
pmichaud so, list assignment shouldn't take a list on the rhs? ;-) 15:44
El_Che if someone is still looking for a room for @fosdem: Olivier Mengué @omengue · 14m 14 minutes ago
.@FOSDEM I still have a bed available for saturday in my twin-beds room at Hotel Beau Site. Your price: 34 €, including breakfast.
TuxCM it should :) but it still surprised me. .fields is now a list of objects, and "fields" caused the next line to barf on an attempted object call 15:45
PerlJam TuxCM: that's less of a listy problem and more of a P6 v. P5 problem (i.e. you're still thinking in P5)
TuxCM is *fighting* to separate those :)
gtodd nine: did you do your talk about Inline::Perl5 yet? 15:46
I meant did you finish the slides ;-) 15:48
15:50 darthg0d joined 15:51 kaleem joined
Woodi_ hallo #perl6 :) 15:52
I think discussions about finalizations needs some decompozition... or maybe I just misunderstand something 15:53
15:54 andreoss joined
gtodd nine: can't recall if it was you or PLICEASE (author of numerous perl5 FFI modules) that mentioned there was an ongoing effort to standardize/document the description of functions available in foreign libs in an an "easy to digest" "X-Platform and/or X-Language" way (in JSON or YAML errm maybe not XML!) so that someone writing an FFI back module to work with errm libopencv.so would create a JSON file that could be used 15:54
nine: oh ... errm it *was* G. Ollis metacpan.org/pod/FFI::Platypus#DESCRIPTION 15:56
Woodi_ first programmer want to declare .disconnect be called when $dbConnection is last seen. and he realy do not care about memory at the moment. but in discussions this nice to have "domain" feature is forgotten and slips into gc don't wanting to make a run...
gtodd nine: "a goal of [ FFI::Platypus ] is to reduce common interface specifications to a common format like JSON that could be shared between different languages." 15:58
15:58 jluis joined, brrt left
PerlJam Woodi_: what do you mean by "when $dbConnection is last seen"? When it goes out of scope? 15:59
Woodi_ but actually finalization is perfect hint for gc - that object do not need to be dragged by all generations... just quick call to $gc->flag($dbconnection, READY_TO_FREE) and $dbc do not need to be looked for while traversing... 16:00
16:00 araujo joined, araujo left, araujo joined
Woodi_ PerlJam: depends on finalization implementation... scope is nice 16:00
16:00 darthg0d left
Woodi_ or gc prefer to personally find all objects to be freed ? :) 16:01
16:01 dayangkun joined 16:02 gfldex joined
jnthn Woodi_: You're conflating memory management and resource management I think, when in Perl 6 - any *many* other languages - those two have been taken apart. 16:03
Woodi_ jnthn: I get impression that all finalization discusions I ever seen do that... 16:05
moritz that doesn't quite make it better 16:06
Woodi_ and $gc-> flag is just "marker", not "GC do free NOW!"
domain objects like handles want actions to be done, for consistency. 16:07
16:09 lizmat joined
lizmat has arrived in Brussels 16:09
16:10 dagurval_ left
jnthn lizmat: Welcome :) 16:10
Woodi_ and that flag (memory menagement) is for free, just happens to be consequence of freeing resources. 16:12
as for cars with auto gear changing :) : new cars with auto feature tends to introduce some manual settings like sportmode or ecomode :) and computer in car do not see that corner is comming (yet)... some nice description of stick using feature :) www.motivemag.com/pub/feature/drive...lick.shtml 16:16
16:19 nige joined
FROGGS jnthn / pmichaud / lizmat: breakfast will start at 7 if you want (not at 8am as described) 16:20
nige can't make FOSDEM - but just wanted to wish everyone well in spreading the "Joy of Six" go for it! ;-)
yoleaux 20 Dec 2014 08:56Z <moritz> nige: I've published your advent post (scheduled for tomorrow, 21st) today already, because today's post isn't written yet
nige thanks moritz++ 16:21
FROGGS nige: you have to pop in more often :o) 16:22
lizmat FROGGS jnthn pmichaud : already arranged for breakfast at 7:15
jnthn 7:15???!! 16:23
16:23 woolfy joined
FROGGS jnthn: only if you want to be at the fosdem at 8am, to prepare the booth and stuff 16:23
so, since I wanna help, I will break fast and early 16:24
16:24 mr-foobar joined 16:25 rindolf left
nige yes - Froggs - kids, work and building a new start up - make it tricky 16:25
am a keen follower of the weblogs however
- speaking of start ups I've got a cool idea for a Perl6 app to aid adoption 16:26
FROGGS do tell
nige :-)
can't go into too much detail ;-) ... but it will need a Mojo style engine 16:27
or some web app engine
FROGGS we've got Template::Mojo and Bailador
nige yes - will look again at these 16:28
have been enjoying using Mojolicious in Perl5 - and would defn use a Perl6 version 16:29
16:29 andreoss left
FROGGS both need better docs (or docs at all), but they work 16:29
nige cool - that's good to know
FROGGS could be faster of course, but I think it is usable 16:30
El_Che that's great news 16:32
16:33 mvuets left
FROGGS the testers.perl6.org uses it (both Template::Mojo and Bailador), and it is only slow for the reason that it fetches thousands of rows out of its database every time 16:34
dalek rl6-roast-data: b37fb34 | coke++ | / (5 files):
today (automated commit)
 16:36
16:37 kaleem left
woolfy El_Che: lizmat and I arrived in our hotel, good parking place for the truck, checked my mail (now 59 people for dinner on Saturday, 31 for Sunday), and on my way to Delirium for some drinks. lizmat is not joining me, she will hack with Perl6-ers (ah well, Perl 6, lovely) 16:38
El_Che woolfy++ 16:39
16:39 alinbsp left
El_Che woolfy: is 70 a max, or does the resto accept an overflow? 16:39
hackers overflow
woolfy 70 is the max for Saturday, 50 for Sunday 16:40
El_Che so, not a lot space left
woolfy El_Che: No overflow allowed or possible, 70 is just the max
El_Che maybe keep an small % open for people that say they come but bail?
woolfy El_Che: already 2 people bailing out for Saturday, and their seats were taken the same 15 minutes 16:41
El_Che those are the *nice* people telling you they bail out 16:42
anyway, 59 is pretty greay
great
woolfy indeed
Now I am going to Delirium
Byebye
El_Che 100% chance of filling the 70 places 16:43
ok bye
16:43 woolfy left
jnthn Wow, so dinner! 16:43
16:43 dayangkun left, sjn left
El_Che jnthn: are you better? 16:43
nwc10 so dinner, many conversation, nom 16:46
jnthn El_Che: Largely. Still not at 100% strength, but nothing to worry about wrt my talk (except I've some of the slides still to write...doing that now ;))
El_Che jnthn++ 16:47
16:47 pecastro left
vendethiel xfix: doing some coffeescript :)? 16:51
16:54 Mouq left 16:55 adu joined
FROGGS and again I wonder how well my talk fits in 50 minutes :/ 16:56
pmichaud I figure if my talk gets to 50 minutes, I'll just stop. :) 16:57
FROGGS :P
vendethiel "and now, for the interesting part... Oh wait, it's been 50 minutes. have a nice day!"
pmichaud vendethiel: Exactly.
vendethiel *g*
FROGGS my problem is that when I am nervous I start to talk faster... I mean, I did the v5 talk in 11min instead of 20 :o(
pmichaud followed by: "Be sure to come see my next talk ."
jnthn FROGGS: You could do a practice run... 16:59
FROGGS so, if one wants to support me and I finish early, then please ask simple but enlighting questions :o)
jnthn: will do that
jnthn FROGGS: And put a horror movie on while you practice so you're nervous, or something, to make it realistic... :P 17:00
FROGGS jnthn: so if you hear mumbling at night, it is your own fault :P
pmichaud I watched r0ml (Robert Lefkowitz, fantastic speaker) do something similar with some of his talks. He would start out his talk noting that before he could talk about what he had written for the schedule, he needed to first cover the material that he ran out of time for in his previous talk. So he would do that, then upon getting to the end of his alloted time he'd say "Oh, sorry, I didn't get to what I was supposed to talk about today. You'll just have
FROGGS hehe
jnthn FROGGS: Hmm, I know enough about nativecall to ask some questions... :)
FROGGS jnthn++
pmichaud: limits all the way down :o) 17:01
El_Che well, if someone wants me to give a subtle sign to slow down, please tell me so 17:02
anyway, after the time is up, I'll remove you from the stage manu militari :)
pmichaud How about a not so subtle sign? ;-)
FROGGS weird images are now popping up in my head 17:04
17:04 xinming left
pmichaud For once I don't expect to be using my entire time allotment. That rarely happens for me. :) 17:04
jnthn pmichaud: I'll believe that when I see it :D 17:05
jnthn has never been to a Pm talk that ran short of stuff :) 17:06
pmichaud I actually had several presentations last year that went exactly on time, or 10 minutes under.
which, considering they were 3-hour presentations, I thought was pretty good. :)
also, I'm the talk that is right at the lunch hour, so I figure I should err on the side of running short rather than long :) 17:07
17:08 zakharyas left 17:15 kjs_ left 17:16 rurban left, vendethiel left 17:23 rurban joined 17:24 sqirrel joined 17:25 rurban left 17:26 konsolebox joined
FROGGS .tell fil_ soon meant 7 hours later :o) (I had no idea when I would get around to it, and what it would involve... happily it is just about setting a flag to true) 17:27
yoleaux FROGGS: I'll pass your message to fil_.
17:30 Kristien joined
Kristien hi 17:31
17:32 Otterpocket left
FROGGS hi Kristien 17:33
17:36 silug left
japhb .botsnack 17:40
yoleaux 11:50Z <timotimo> japhb: i'd like to have an integrated tool for perl6-bench that can build a version of a compiler and the version before it, or alternatively you could pass a little patchset and it'd build the things with and without the patchset, time all things and immediately spit out a comparison graph file that might even contain the diff
:D
japhb .tell timotimo That sounds eminently doable and a very good idea. GitHub Issue it and I'll try to get to it next time I'm not slammed by $day-job? 17:42
yoleaux japhb: I'll pass your message to timotimo.
pyrimidine Re: DESTROY, was the main issue re: conflating it vs. use of DESTROY in p5? 17:45
dalek c: 3b4a118 | paultcochrane++ | lib/Type/Blob.pod:
Use consistent capitalisation in template letter table
c: 760cd53 | paultcochrane++ | lib/Type/Blob.pod:
Minor sentence flow improvements in update() docs
pyrimidine e.g. needing to rename it?
17:45 xinming joined
moritz \o/ more doc contributions 17:46
jnthn \o/ 17:47
.oO( Ship more patches to the doc dock! )
[ptc] moritz: is htmlify working for you with current rakudo (moarvm) HEAD? 17:48
moritz [ptc]: dunno, I'll give it a try 17:49
17:49 nige left, slavik left
[ptc] moritz: thanks. It doesn't run for me atm, and I'm just wondering if I have a case of PEBCAK 17:50
moritz [ptc]: the cron job that rebuilds doc.perl6.org runs moarvm 2014.01.1
... but it's good to catch any errors in the tool chain early
[ptc] nods 17:51
pyrimidine nm Re: DESTROY, just caught this in backlog # moritz++ for IRC logs
moritz fwiw I only use htmlify.p6 with rakudo-m; I'm fine with it using r-m specific features 17:53
[ptc] ah ok. Nice to know :-) 17:54
17:54 rurban_ left
[ptc] I've written something to document pack(), however have just realised that Blob.pod could be the wrong place 17:54
maybe in Buf.pod?
since the sub is defined in Buf.pm? 17:55
moritz [ptc]: no, Blob.pod is correct
[ptc]: that's a historical accident
[ptc] ok.
17:55 silug joined
moritz (we used to have only Buf, not Blob) 17:55
[ptc] ah, was just about to ask why
dalek c: e4a91e7 | moritz++ | README.md:
README: adjust vision statement

explicitly include language features, not just built-in types and routines		 17:56
moritz now running htmlify... 17:57
[ptc]: does it hang for you?
dalek c/add_pack_doc: a208c53 | paultcochrane++ | lib/Type/Blob.pod:
Document the pack() sub
pyrimidine So, if the main need for DESTROY is to run garbage collection, why not use a name like TRASH?
[ptc] moritz: I had it hanging (at Regexp)
muraiki I prefer ANNIHILATE
moritz [ptc]: yes, I have that too
[ptc] then I thought I had a problem with my rakudo, so I rebuilt it (make realclean; perl Configure....; make install) 17:58
moritz [ptc]: and I recently patched regexes.pod, that might be why
17:59 Kristien left
[ptc] and then I get this: Missing or wrong version of dependency 'src/gen/m-CORE.setting' 17:59
jnthn pyrimidine: The point isn't the run GC; the point is it's a callback if GC detects that all references to the object are now gone and, were it not in the list of things with a DESTROY method, it'd have been freed already
*to run
[ptc] moritz: now, I'm running the full rakudo test suite (just in case), and it runs again (gah!)
moritz: however, still hangs on regexp.pod 18:00
FROGGS [ptc]: that means you have to rebootstrap panda: "Missing or wrong version of dependency 'src/gen/m-CORE.setting'"
[ptc] FROGGS: ah, that'll be it :-) Thanks!
18:01 Mouq joined
[Coke] m: my $foo = qw‘bar’; 18:01
camelia ( no output )
FROGGS .oO( caches, you can't live with and without them )
[ptc] I'd tried to run perl6-debug-m, however Debugger::UI::CommandLine was missing, hence I rebootstrapped panda...
he
Mouq [ptc]: Just note, if you haven't run it before, that htmlify currently takes a really long time to run. O(minutes) 18:02
[ptc] Mouq: yeah, I know. I guess that's why people want to parallelise it 18:03
18:03 slavik joined
[ptc] however atm it's hanging on regexes.pod and just burning cpu 18:04
moritz but that won't help with a genuine hang :-)
[ptc]: Debugger::UI::CommandLine is a user-space module and must be installed with panda
ok, rakudo itself hangs when presented with regexes.pod
so, rakudobug
18:04 madcapJake left 18:05 pyrimidine left, madcapJake joined, pyrimidine joined
pyrimidine jnthn: ok, makes sense. I'm so used to having an explicit cleanup step with some classes. 18:06
[ptc] moritz: just a quick heads up: I added the pack docs in the "add_pack_doc" branch. Could you let me know if it looks up to standard? TIA!
18:06 espadrine left 18:07 ivan`` left
moritz [ptc]: looks good 18:07
[Coke] m: my $foo = ‘bar’;
camelia rakudo-moar e4883a: OUTPUT«===SORRY!=== Error while compiling /tmp/rigR999ZRG␤Malformed initializer␤at /tmp/rigR999ZRG:1␤------> my $foo = ⏏‘bar’;␤ expecting any of:␤ prefix or term␤ prefix or meta-prefix␤»
[ptc] moritz: thanks :-) 18:08
moritz [ptc]: though I don't think the 1:1 correspondence to unpack is quite correct
but better to have it in that form that not at all
dalek c: a208c53 | paultcochrane++ | lib/Type/Blob.pod:
Document the pack() sub
c: 400bdf2 | paultcochrane++ | README.md:
Merge branch 'master' of github.com:perl6/doc
18:08 Maddingue left, djanatyn left, epochbel1 left, Maddingue joined 18:09 btyler left, retupmoc1 left, ingy left, Bucciarati left, revdiablo left, mathw left, masak left, dylanwh left, cxreg left, rjbs left, awwaiid left, epochbell joined
moritz ok, 09867b7eb092d6eabfd305df931a35b548fa1b30 in perl6/doc causes rakudo to hang on regexes.pod 18:09
18:09 retupmoca joined, mathw joined, pmichaud left, revdiablo joined, btyler joined, masak joined 18:10 awwaiid joined, cxreg joined, rjbs joined, masak is now known as Guest27290, pmichaud joined
dalek c: 08b2628 | moritz++ | lib/Language/regexes.pod:
Revert "Regexes: document :overlap and :exhaustive"

This reverts commit 09867b7eb092d6eabfd305df931a35b548fa1b30.
It causes rakudo to hang on regexes.pod. TODO: reduce to minimal example, file rakudo bug, restore the commit once fixed (or work around it)		 18:10
[ptc] moritz: thanks for the help, I appreciate it. In time I should be able to do more stuff with less supervision (I hope!) 18:12
Mouq moritz: It still shouldn't hang, but what you reverted has an "=begin code" with no "=end code"
I'm pretty sure that normally errors out..
m: =begin code
camelia rakudo-moar e4883a: OUTPUT«===SORRY!=== Error while compiling /tmp/8s40WVOB7i␤Preceding context expects a term, but found infix = instead␤at /tmp/8s40WVOB7i:1␤------> =begin ⏏code␤»
18:12 konsolebox left
Mouq m: ␤=begin code␤ 18:12
camelia rakudo-moar e4883a: OUTPUT«===SORRY!=== Error while compiling /tmp/n1bOuTfuYy␤'=begin' not terminated by matching '=end code'␤at /tmp/n1bOuTfuYy:2␤------> <BOL>⏏<EOL>␤» 18:13
moritz [ptc]: fwiw I'm fine with you (or anybody else) pushing to master directly and asking for review later
18:13 BigBear left
FROGGS jnthn: btw, would be nice if we could have 'has MyFooEnum $.foo' in CStructs... should be fairly doable I think 18:13
moritz [ptc]: in general, bad docs are better than no docs. And forgiveness > permission
pyrimidine jnthn: re: DESTROY, I recall someone (Peter_R maybe) who brought this up in the past (some explicit way of defining cleanup upon GC). May have to backlog to see if anything came about from that 18:14
moritz Mouq: sounds like a very probable hypthesis
18:14 Bucciarati joined
pyrimidine moritz: re: pod and code documentation, would you suggest keeping it separate from code for now? e.g. separate .pod file? 18:16
moritz pyrimidine: I don't have a recommendation for you 18:17
the reasons for keeping them separate for core stuff don't apply to user-space modules
PerlJam pyrimidine: I'd keep them together were I you. 18:19
pyrimidine moritz: np. just checking whether it makes much of a difference; would rather keep ...
18:19 ivan`` joined
pyrimidine what PerlJam++ said 18:19
18:20 djanatyn joined, sqirrel left 18:21 dylanwh joined, ingy joined
Mouq pyrimidine: And you can also attach documentation to the things you're documenting in your code ( hoelzro++ ). We don't have much functionality to expose this into generated documentation, but I'm sure that'll be a thing eventually 18:22
moritz Mouq: it might be not a "genuine" hang, but a very, very slow backtracking to death 18:23
Mouq m: #| my sub rocks␤sub foo ( $a, #= This parameter does stuff␤ $b, #= So does this one␤){ 42 }␤say &foo.WHY
camelia rakudo-moar e4883a: OUTPUT«my sub rocks␤This parameter does stuff␤»
pyrimidine Mouq: will have to check into that, thx! 18:26
Mouq moritz: Oh, hm.. is it allowed to backtrack like that?
Oh
|| '=end' \h+ 'code' [ <pod_newline> | $ ] || <pod_string>**0..1 <pod_newline> <delimited_code_content($spaces)>
18:27 Kristien joined
Mouq The formatting's weird, but that means that the thing that parses the code content builds up a giant callstack 18:27
Which then has to get unwound
Or, not unwound, but failed over
moritz my current plan is to eat dinner, and then submit a bug report 18:28
PerlJam food++ 18:29
happy noms moritz! :)
Mouq moritz: Sounds good
Mouq tests a fix 18:31
18:33 Guest27290 is now known as masak_grr 18:35 telex left 18:36 telex joined 18:39 abraxxa joined 18:40 sqirrel joined
FROGGS goes down to the lobby to meet other lobsters 18:48
PerlJam Was there a desire to move from RT to github issues? I seem to recall a general lean in that direction. Anyway, I just saw www.dagolden.com/index.php/2397/mov...-improved/ which has some nice helpful software if that's desired.
18:49 FROGGS[mobile] joined
pmichaud I was pushing for a move from RT to github in 2013. 18:50
Since I've been less active, I've been less pushy, though. :)
I know that when I went to work on bugs in 2013, I felt like the RT interface was an impediment 18:51
masak++ was the chief proponent of keeping the existing rt bugset, if only because he had tools and mental reference points built up over a period of time that enabled him to be an effective historian 18:52
at yapc::na 2013 we left it as an agreement that I'd investigate the difficulty of migration and we'd revisit again
nwc10 pmichaud: is there a plan for handling security tickets? ie CVE-level pain? 18:53
pmichaud plan?!
PerlJam pmichaud; Is your attendance at fosdem symbolic of your return to activity? :) 18:54
hoelzro (GitHub tickets)++
nwc10 ie stuff that should be private until fixed
avuserow colomon, raydiak: I'm gradually working on converting an old Perl5 project to Perl6, which was a network jukebox, hence many of my libraries being audio-related.
pmichaud nwc10: I'm not aware of a plan as yet, no. We can develop one.
nwc10 pmichaud: "plan", as in, "I love it when a ... comes together" :-)
avuserow Also it's a good way for me to try things out, see what's missing or broken
pmichaud PerlJam: my attendance at FOSDEM is best interpreted as a strong intent to return to activity. :)
I'm still robotic-competition-bound through Feb 14, at least. 18:55
pyrimidine pmichaud: I've look at RT -> Github migration for our project, but we have a handful of RT tickets. I could see migrating perl6 tickets being a bit more... painful
nwc10 pmichaud: in that, I think to make a private issue queue on github, @you would need to pay them.
pmichaud (two robotics teams at their respective regional championships)
nwc10: I have no problem with paying github.
moritz nwc10: if this is their consumer prices (and not their enterprise offers), I'm sure we'll find a solution. 18:56
nwc10 pmichaud: I also have no problem with github getting money. But it's an open-ended requirement to keep paying them (a small amount of money)
pyrimidine nwc10: GitHub is pretty nice about requests from open-source projects. Doesn't hurt to ask
nwc10 whereas on perl.org RT, private queues are just as free (or costly) as public queues
PerlJam what pyrimidine said
pmichaud at any rate, I'd rather choose based on capability than (likely low) expense. 18:57
pyrimidine: I agree that migrating perl6 tickets may end up being painful. But I'm guessing it's same O(pain) as it was for me to use RT to try to categorize and resolve them. 18:58
nwc10 I think that's a sane rational.
pyrimidine pmichaud, nwc10: note they are *very* helpful about open science: github.com/blog/1840-improving-git...or-science
nwc10 having dealt with (I think) two CVEs, I'm just mindful that the project needs to be ready for them.
18:59 espadrine joined
nwc10 sorry, having done the work and release @pain for at least one. 18:59
pyrimidine pmichaud: the link that PerlJam posted from dagolden: seems like it would tremendously for this
pmichaud time for dinner for me -- bbl
nwc10 I tend to block it out. It's not fun
hence why I can't remember if was one, or two
pyrimidine * would be tremendously useful
jnthn eek, yes, dinner!
&
dalek kudo/nom: 99021cf | Mouq++ | src/Perl6/ (2 files):
[Pod] Eliminate recursion in delimited_code_content
 19:05
kudo/nom: 7fd2f64 | Mouq++ | src/core/Cursor.pm:
Merge branch 'nom' of github.com/rakudo/rakudo into nom
moritz Mouq: part of the problem seems to be that the code doesn't stop looking when reaching EOF 19:08
Mouq: and instead backtracks
19:08 FROGGS[mobile] left
Mouq moritz: The above patch fixes it for me 19:08
moritz Mouq: could you please run gist.github.com/moritz/8ceacf454fb1f5af9461 ? 19:09
it nicely illustrates the timing :-)
Mouq moritz: Awesome! And even awesomer: gist.github.com/Mouq/e1a1aba877724259d5ce 19:10
moritz Mouq++ indeed 19:11
moritz doesn't even have to write up his bug report
raydiak g'mornin #perl6 19:13
masak_grr good antipode, raydiak
19:13 masak_grr is now known as masak 19:14 sqirrel left
raydiak \o masak...thanks for enduring the darkness for a while every day so that I can have a turn with the sun :) 19:14
masak you're welcome. 19:15
hoelzro on a side note, I've been learning Octave for a machine learning course on Coursera, and I'm pleased to see that many powerful features from Octave are also present in Perl 6 =)
o/ raydiak
raydiak heya hoelzro
moritz where octave implements mostly the same language as matlab
hoelzro mhmm 19:16
interestingly enough, I remember *hating* Matlab in college
I think I had to stop worrying and learn to love the matrix 19:17
dalek c: c2be389 | moritz++ | lib/Language/regexes.pod:
Revert "Revert "Regexes: document :overlap and :exhaustive""

This reverts commit 08b2628e7023c02710f1ac9daef556f6e7b36f1f, and thus re-instates commit 09867b7eb092d6eabfd305df931a35b548fa1b30, while also fixing broken pod. Mouq++ and [ptc]++ for reporting and diagnosing/fixing
moritz what I really don't like about matlab/octave is that the compiler don't catch undeclared names at compile time
that's really disappoing when you're used to Perl 6 :-)
masak yes, that's the thing. when you know it *can* be done, anything short of it just feels a bit... cheap. 19:18
hoelzro heh
moritz aye 19:25
19:27 kjs_ joined
gtodd maybe something like Inline::Matlab could help ;-) 19:27
moritz to be fair, if you use matlab with the IDE, you get some nearly-real time linting 19:28
gtodd perl5 is to duct tape as perl6 is to .... 19:29
moritz duct concrete? 19:31
nwc10 several rolls of duct tape at the same time, with O(1) emojii.
oh, wait, no, that makes no sense :-)
[Coke] Having lived through a migration process: I would vote against migrating unless the benefits of the new system are INCREDIBLY compeling. 19:32
gtodd flame resistant waterproof double-sided nanotech enhanced super glue coated duct tape with teflon coating and kevlar fibre reinforcement
masak the O(1) emoji depicts a cyclops devouring a ruler? 19:33
gtodd lol 19:35
while lying down
19:35 abraxxa left
gtodd can't wait to see/hear nine's Inline::P5 talk 19:36
19:36 colomon left
nwc10 yes, exactly - they're adding it in Unicode א.0 :-) 19:37
(and I discover that as I typed that it, I'll be confusing Left to Right or Right to Left)
jdv79 i guess they'll announce the streaming urls "later"? 19:41
19:41 Mouq left
gtodd some talks will end up on youtube ... liz will probably post links in p6weekly ;-) 19:42
aka lizmat 19:43
jdv79 i thought they were going to try to stream 19:44
gtodd oh probably that too
masak tell them that the stream is only good if they tell us in advance :P 19:45
gtodd there's ... time travel 19:46
jdv79 i guess i'll only know after i wake up at 3am EST...
19:46 colomon joined 19:47 espadrine left
colomon avuserow++ 19:50
20:00 raiph left, raiph joined
raydiak do we have weak references? 20:09
masak not to my knowledge.
I've heard the need for them called a "design smell" on this channel. 20:10
20:10 spider-mario joined
TimToady well, something like that can be useful for caching, I suppose 20:10
but it would take the connivance of the GC to manage it 20:11
raydiak blast....and END phaser would solve my DESTROY problem, but if I can't keep a list of the objects without interefing with GC, that seems counterproductive 20:12
masak right. it's not pure user space.
20:13 kjs_ left
raydiak fwiw I don't really have a problem any more, I'm just kinda wondering about the corner I stumbled in to...what does a person do if they want something like DESTROY but reliably always called? 20:15
20:16 darutoko left
raydiak but yeah I guess mostly what I've heard is design smell and YAGNI 20:17
20:21 retupmoca left
moritz raydiak: for me, the question is really: what kind of cleanup would DESTROY do that can't be done by the OS cleaning up after the process? 20:26
20:29 raiph left
geekosaur I can imagine e.g. a database client implementation (think berkeleydb, not postgres) that write-behind buffers some things, and needs to flush the buffer 20:29
moritz if there are good examples of such cases, it would be much easier to make a good point for global destruction
20:29 raiph joined
moritz geekosaur: well, but that has to be able to deal with the not-flushed case anyway, in case of power loss or a kill -9 20:29
geekosaur also, releasing resources when the application requires some ordering on it
PerlJam geekosaur: that sounds more like the timing of the destruction is what's really important
geekosaur moritz: that does not mean (a) that the normal exit should be just like a panic abort, (b) that the user should have to explicitly manage state (what is this, C?) 20:30
bartolin moritz: a few days ago you made ObjAt.new(str) work (and thereby fixed RT #122094). I wonder where to place a test for that -- do you think it would make sense to add a new test file S02-types/built-in.t? 20:31
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=122094
20:33 retupmoca joined
raydiak I find it odd that I am the only one who finds it odd that we give people a plethora of ways to say most things, even the stuff you really *shouldn't* be doing, but I want to say something which seems like a fairly natural way to think, but most answers are just "we don't let you say that" 20:37
I guess I think a better question is why *shouldn't* we have a few decent options for doing this? 20:38
PerlJam If we had a phaser that fired just before an object was about to go out of scope (let's call it CLEANUP), would that satisfy people's need for "destruction'?
masak PerlJam: go out of lexical scope? 20:39
PerlJam aye
masak PerlJam: because that's LEAVE.
raydiak but I can't, from my own module, install a LEAVE in whatever is *using* my module, afaik
masak and I think moritz raised a good point earlier: given that the GC is what it is, and given whe have LEAVE and `will leave`... are there any good uses left for DESTROY? 20:40
raydiak lizmat++ had the passing thought of exporting phasers, which is kinnda interesting
but really something that works exactly like DESTROY but if there are any left at exit they get called, would work too 20:41
there are a few subtly (but fiendishly) different things people might want in this area
wrt to timing and ordering guarantees, for one 20:42
20:42 adu left
PerlJam masak: not quite. LEAVE approaches the problem from the wrong end (the routine rather than the object) 20:43
20:43 raiph left
raydiak so I understand that using this stuff if it existed would make things slower on exit, but it shouldn't be very concequential unless you're doing heavy stuff in there, just like it would be if you do heavy stuff in a LEAVE 20:43
PerlJam raydiak: "exactly like DESTROY" isn't well-defined. Different people have different ideas of what DESTROY is for (IMHO)
20:43 Mouq joined, raiph joined
raydiak PerlJam: I mean working precisely as it does now in rakudo, except with the reliability bit 20:44
masak sometimes I feel in the discussions about DESTROY that people desperately want it to work like it does in Perl 5... but that world is gone. timely destruction simply doesn't exist, and can't coexist with the other tradeoffs we've made, mostly GC-related, but also related to threading.
raydiak okay so throw out the timely part
PerlJam raydiak: if you do that, then I don't see a problem. 20:45
masak if you throw out the timely part, *it may not fire at all*
because the process may end.
PerlJam or (statistically unlikely) the GC may never fire. :)
masak or may fire, but not for your object. 20:46
(even though it could)
20:46 adu joined
raydiak if it gets killed or something, that's one thing, but I don't see why it's so hard to just make a copy of DESTROY which also gets called on any controlled exit if it hasn't already 20:46
masak jnthn already answered that.
because it's expensive, and nothing the OS can't do.
moritz raydiak: then propose a patch 20:47
masak +1
moritz and if it doesn't slow down those programs that don't need it, and isn't overly complicated, I'm sure it'll be accepted
masak on the face of it, it might incur an O(n) slowdown for n allocated objects. 20:48
raydiak someone said we already have a flag that gets set when you declare a DESTROY
so if we copy that, and people use it, they get the slowdown cuz they asked for it...otherwise shouldn't be hardly any 20:49
PerlJam still thinks that DESTROY should be renamed to CLEANUP or something with less conceptual baggage
moritz raydiak: yes, the meta object stores it
raydiak: doc.perl6.org/type/Metamodel%3A%3AFinalization
masak PerlJam: there's a very strong symmetry between BUILD and DESTROY
ab5tract PerlJam: i think today's discussions bear out the wisdom in that position
UNBUILD>
raydiak yes note I'm not necessarily talking about changing DESTROY itself and don't have a proposed name because there are too many for my brain to consider atm
ab5tract ?
PerlJam masak: that only exacerbates the problem as far as I'm concerned. 20:50
raydiak DISMANTLE?
masak PerlJam: my point is, the resource things you would allocate or initialize in BUILD, you would deallocate or tear down in DESTROY. 20:51
raydiak (it's even uncomfortable to type, which would be good imo)
masak watches in mild despair as the discussion once again morphs into a bike shedding
20:52 raiph left 20:53 raiph joined
raydiak I think it's a natural way to think about objects, so letting people write code that way...is it a bad idea? we allow it with phasers, they just don't handle this case I'm look at very well and wish there was a way to do so, for the same reasons that the related phasers are cool to have 20:53
raydiak wishes TimToady were around 20:56
PerlJam notices that DESTROY is under-specced 20:57
raydiak yeah I don't think the commit that meant to put it back, put everything back that I thought I saw about it before 20:58
ab5tract masak: is balancing your BUILD with a DESTROY a suggestion or a requirement? what's the practical consequence of not writing a DESTROY to parallel your BUILD? i don't have any horses in this race, i only ask out of curiousity.
in discussions about perl 6 with people smarter than me, the existence of BUILD itself has come up as a question mark 20:59
PerlJam ab5tract: Just like BUILD is only need *if* you need to do something special when you create your objects, DESTROY is only needed *if* you need to do somethign special when they go away. 21:00
masak ab5tract: what PerlJam said. I'm likely to write a hundred BUILDs for every DESTROY I'll need to write. 21:01
ab5tract: while they could go together and do opposite tasks, one can exist without the other.
ab5tract these are comforting sentences to read
21:02 Bucciarati left
masak what I was addressing was changing a name that's already pretty optimal in its connotations to something else just because people feel that the Perl 5 meaning interferes too much with the Perl 6 meaning. 21:02
moritz
.oO( either can live while the ohters survive ... )
masak moritz: IN THE SAME VULD
moritz oh, right, I forgot about multi world theories 21:03
[Coke] perljam: "LEAVE approaches the problem from the wrong end" - I disagree. can you show me an example where dealing with this in a LEAVE block doesn't work?
raydiak people want me to give examples of what it'd be useful for, but I wonder what examples there are of things DESTROY is useful for as it exists now 21:04
masak raydiak: me too.
geekosaur LEAVE is lexical. I may well intend that my object survive my lexical scope, so that the caller can use it to do things and then it cleans up when it goes out of dynamic scope 21:05
raydiak [Coke]: but I can't, from my own module, install a LEAVE in whatever is *using* my module, afaik
masak it'd have to be something that "would be nice" if it ran at object GC, but not absolutely essential.
moritz closing a TCP connection is a classical example
if you don't close it, it'll time out
but you're a nicer team player if you close it sooner 21:06
raydiak yeah just something that is guaranteed to be called exactly once, at or after the last ref goes away, and at or before the program is terminated
[Coke] and if you know you're done with it, you can close it yourself.
masak moritz: was gonna say. close it explicitly in code.
moritz: can probably be done with LEAVE in most cases.
geekosaur my thinking is again that caching/buffering thing. where you need to update two different parts of the file when you flush, so the standard buffering isn't sufficient
PerlJam Coke: IT does work, it's just potentially cumbersome. To do something to a particular type of object every time it goes out of scope requires that I have a LEAVE in each of those scopes. I'd rather have defined a CLEANUP method that gets called for me automatically.
moritz masak: right. It's more of a fallback when somebody forgets it 21:07
masak raydiak: it's not guaranteed to be called exactly once. it's guaranteed to be called at most once.
raydiak masak: as it is now, yes...that's my problem
21:07 Bucciarati joined
masak raydiak: not even Perl 5 guarantees "exactly once". 21:07
geekosaur also, think about this one, how is buffered file I/O currently implemented? does LEAVE work to do the final buffer flush if you're passing a file handle around? 21:08
moritz and in p5, when you're in global destruction, all bets are off
[Coke] perljam: so, LEAVE { $myObj.CLEANUP }, but where we automatically inspect all myobj's for a CLEANUP method on block close?
masak raydiak: if that's your problem, then you have a pretty problem with a pretty fundamental part of memory/threading design in Perl 6.
geekosaur (p5 gets away with this by handling it internally)
moritz some of your attributes can be undfined
raydiak masak: I am not thinking anywhere in the P5 region of my brain when I am talking about what I think we should have
masak s/pretty//
[Coke] that seems expensive.
timotimo when do i start watching fosdem streams so that i don't miss anything perl6y?
yoleaux 17:42Z <japhb> timotimo: That sounds eminently doable and a very good idea. GitHub Issue it and I'll try to get to it next time I'm not slammed by $day-job?
[Coke] s/all myobjs/vars/
PerlJam Coke: aye and aye (I'm just spit-balling hoping to find something better)
[Coke] what if you've got a bound object that you then cleanup, making it useless for wherever else it's bound? 21:09
PerlJam Coke: also, I think if there were something like that, it would satisfy what people think they want DESTROY to really do.
[Coke] Is there any prior art in Java we can look at? 21:10
ab5tract timotimo: there's perl 6 sprinkled all the way through the perl devroom tomorrow :)
[Coke] at least there, it's been GC all the time, so there's no "java5" expectations. :)
PerlJam Coke: that seems like a DIHWIDT situation
masak ab5tract: I think you may have misunderstood timotimo's question... 21:11
timotimo tomorrow, yeah?
masak ab5tract: oh, nope. I misunderstood the question. :)
ab5tract masak: :)
masak ('when', not 'where')
do we know where the streams are yet?
masak .oO( all I know so far is not to cross 'em ) 21:12
retupmoca The current DESTROY is very useful in NativeCall code, as a way to handle the C-side memory management (when this object is cleaned up, free this C struct / call this library cleanup function)
I haven't had any other use for it yet
21:13 denis_boyun joined
retupmoca only caught the last half or so of the conversation 21:13
japhb WRT to the timely destruction thing -- do we still have an NQP op to force a GC run? If you use that, does DESTROY become reliable? 21:14
Basically, so that you can say "As of now, I want all pending DESTROY to happen. I'm ready and willing to wait."
ab5tract if that
21:14 denis_boyun left
masak japhb: arrgh 21:14
[Coke] what type is a nativecall object?
ab5tract if that GC run could be scoped in any way, it could be useful
PerlJam japhb: does a GC run guarantee that all objects waiting to be freed actually are?
21:15 denis_boyun joined
japhb masak: Why are you arrgh'ing me? 21:15
masak japhb: (more)
japhb PerlJam: Well, in our current implementation, I'm not sure if 1 would, but I believe 2 will. :-)
21:16 raiph left
japhb is mentally playing the Jeopardy final question music 21:16
ab5tract raydiak: do i recall correctly that the impetus of your DESTROY desires also come from NativeCall?
raydiak ab5tract: yeah that's where the thoughts came from originally 21:17
masak japhb: because I think forcing GC is definitely a design smell.
21:17 raiph joined
PerlJam masak: I agree. 21:17
masak japhb: and it's just another layer of failing to acknowledge the GC model we have, incurring overhead for wishful thinking 21:18
japhb masak: I agree with you in almost all cases. The exception is soft-real-time, e.g. games.
I disagree with that. I believe it is intentionally breaking an abstraction to deal with a 1% case, when the GC is tuned for the 99% cases.
That's why I was asking about an NQP op, not saying Perl 6 feature. 21:19
PerlJam japhb: and I agree with that :)
japhb And it was a *question*, not a suggestion, to boot.
raydiak I'm not even necessarily talking about a change that has to be tied to GC at all, the only implementation hurdle I can see is knowing if anything is left that uses a metaclass with that DESTROY flag set at END time
masak ok :) sorry for overreacting
japhb hugme: hug masak 21:20
hugme hugs masak
moritz hugs masak and japhb without bot involvement
japhb heh
masak :)
El_Che live.fosdem.org/
21:20 colomon left
masak \o/ 21:20
El_Che live stream
PerlJam El_Che++
El_Che live.fosdem.org/watch.php?room=K.3.401 21:21
(if Perl is your thing :) )
and on sunday, this is Larry's room: live.fosdem.org/watch.php?room=Janson
japhb El_Che: I missed where you pointed to how (and when) we'll see the videos if we're on the other side of the planet. YouTube, I'm guessing? 21:22
21:22 raiph left
raydiak perhaps I'm just being thick-headed...will re-read the backlog and reflect later, maybe look at what that patch would look like...afk for now 21:22
El_Che japhb: no, you can livestream from the url I posted
later the videos will also be available at the fosdem site
starts tomorrow at 10:30 CET 21:23
japhb El_Che: There's enough buffering to start it the night before and rewind it the next day?
21:23 FROGGS left, colomon joined, denis_boyun left
El_Che well; depending on where you are you can just watch the raw video 21:23
that will be online soon
21:23 diana_ joined
japhb is in America/Los_Angeles timezone (AKA PST) 21:23
21:23 raiph joined
japhb -08:00 21:24
El_Che I'll post more info on the channel if I know more
japhb ok
21:24 pecastro joined
El_Che going to bed now (had to get up today at 4 am :() 21:24
and tomorrow will be early as well :) 21:25
bye
japhb o/
21:25 diana_ left, FROGGS joined, colomon left, diana_olhovik joined 21:26 colomon joined
moritz \o 21:26
retupmoca CET is +0100 ?
21:26 krunen left
japhb .ask andreoss How is the perl6-bench pull request coming? 21:26
yoleaux japhb: I'll pass your message to andreoss.
japhb .tell timotimo Don't forget to GH-issue your perl6-bench feature request 21:27
yoleaux japhb: I'll pass your message to timotimo.
ab5tract i wasn't able to follow the whole conversation today, but if the main (only?) trouble spot with DESTROY is related to smoothing over the memory management of *another language*, then a) perl 6 is doing great, and b) NativeCall is probably the layer we should be focusing on
21:28 colomon left, colomon joined
timotimo aye 21:29
yoleaux 21:27Z <japhb> timotimo: Don't forget to GH-issue your perl6-bench feature request
japhb wonders if we can bodge together an only-one-reference-ever object, the way some languages can ensure stray references can't be created, so whoever holds the one reference is guaranteed to be the one responsible for destroying it 21:30
masak japhb: that would be interesting! 21:31
(but hard)
japhb Seems like once in a while, that guarantee is exactly what you need to retain sanity -- and be sure that if you have the reference, it is safe for you to create a LEAVE handler to destroy it, because no one else could possibly care about it.
moritz isn't that exactly what rust is offering? 21:32
masak it is
ab5tract .tell El_Che streams!! much excite, many popcorn. thanks for all the hard work! 21:33
yoleaux ab5tract: I'll pass your message to El_Che.
moritz but they only achieve it by anchoring it very deep in their type system
japhb doesn't know rust, but has seen it in C++, and knew of some Microsoft research that supported it for other languages (though they had gone a bit farther with the idea than just that)
masak the theory is called affine algebra, IIRC 21:34
japhb ISTR they had the ability to define "islands" in an object graph that have only one inbound reference
(But possibly many objects within that island that reference each other)
masak japhb: I remember that paper
very aggregates
21:37 raiph left 21:39 yoleaux left, kjs_ joined 21:40 FROGGS[mobile] joined 21:41 colomon left
moritz such DDD :-) 21:43
21:43 colomon joined 21:44 Mouq left
ab5tract solving it at that layer would be even better :D 21:45
i'm signing off for the night. i'm excited for this weekend! i feel a bit like a bum for being near-but-not-at fosdem, but there's streams! and i will hopefully be releasing my first p6 module before the bell of monday tolls. 21:47
good night and many thanks for all the insightful discussion :) 21:48
hugme: #perl6
21:48 FROGGS[mobile] left
FROGGS gnight ab5tract 21:48
21:48 pecastro left
ab5tract (i guess hugme prefers legitimate nicks) 21:49
moritz sleeping is a good idea
FROGGS hugme: hug #perl6
hugme hugs #perl6
21:49 pecastro joined
FROGGS sleep well moritz :o) 21:49
ab5tract o O ( hugme knows how to do other things? ) 21:50
21:51 diana_olhovik left
PerlJam ab5tract: hugme has the important but all-too-overlooked task of hugging others. It doesn't need to do other things :) 21:51
ab5tract :) 21:52
21:52 ab5tract left
japhb It used to also hand out commitbits, IIRC 21:54
hugme: help 21:55
hugme japhb: (add $who to $project | list projects | show $project | hug $nickname | tweet $twittername $message )
japhb Functionality appears to still be there, though I dunno if it has the right credentials still
hugme: list projects
PerlJam hugme: list projects
hugme PerlJam: I know about Math-Model, Math-RungeKutta, MiniDBI, bench-scripts, book, gge, hugme, ilbot, java2perl6, json, modules.perl6.org, november, nqp-rx, nqpbook, perl6, perl6-examples, perl6-wtop, proto, rakudo, star, svg-matchdumper, svg-plot, tardis, try.rakudo.org, tufte, ufo, web, yapsi
japhb: I know about Math-Model, Math-RungeKutta, MiniDBI, bench-scripts, book, gge, hugme, ilbot, java2perl6, json, modules.perl6.org, november, nqp-rx, nqpbook, perl6, perl6-examples, perl6-wtop, proto, rakudo, star, svg-matchdumper, svg-plot, tardis, try.rakudo.org, tufte, ufo, web, yapsi
PerlJam heh
japhb Great minds. ;-) 21:56
hugme: show hugme
hugme japhb: the following people have power over 'hugme': P⁣erlJam, T⁣imToady, [⁣particle], c⁣olomon, j⁣nthn, m⁣asak, m⁣berends, m⁣oritz_, p⁣michaud. URL: github.com/moritz/hugme/
japhb nice
21:59 xfix left
PerlJam Everybody at FOSDEM have an excellent time! 22:00
PerlJam is headed out
22:03 kjs_ left
raydiak heh and if yer really lonely... 22:04
hugme: cuddle me
hugme cuddles raydiak
japhb 'hugme: help' patch clearly needed. :-)
tadzik oh, I didn't know it can cuddle :) 22:05
raydiak looks like it'll blush and hand out karma once in a while too 22:06
FROGGS hugme++ 22:08
japhb .karma hugme
oh well
raydiak I thought maybe the help omission was intentional...some people might object to being cuddled without consent :) 22:09
japhb Does it allow cuddling anyone, or just 'cuddle me'?
tadzik hugme: cuddle japhb
hugme cuddles japhb
tadzik ha
a proxy :D
22:11 raiph joined, alinbsp joined, rurban joined 22:12 colomon left
japhb imagines hugme looks like Big Hero 6 22:12
22:13 rurban left
masak :D 22:13
22:14 raiph left
raydiak hugme: hug strings with spaces 22:14
hugme hugs strings
22:14 raiph joined
masak hugme: hug strings with non-breaking spaces 22:15
hugme hugs strings with non-breaking spaces
22:16 pecastro left 22:18 colomon joined
raydiak retupmoca: would you consider packaging a zlib dll with compress::zlib::raw to make it more windows-friendly? 22:23
skids
.oO(sheesh selinux is one big festering DRY violation. Yay. Now I can go home.)
22:24 raiph left 22:25 raiph joined 22:28 skids left 22:33 vendethiel joined 22:34 raiph left, raiph joined 22:35 raiph left, raiph joined 22:38 espadrine joined 22:43 raiph left 22:44 raiph joined 22:47 raiph left 22:48 raiph joined 22:52 raiph left, raiph joined 22:57 [Sno] joined
Kristien I want to learn Fortran. 22:57
FROGGS why? 22:58
masak Kristien: lesson one: it's spelled 'FORTRAN'
:P
Kristien No, not that old crap. :( 22:59
masak oh, you're right. they seem to have ameliorated that. en.wikipedia.org/wiki/Fortran 23:00
since Fortran 90.
vendethiel Kristien: we do have a lot in common, :P 23:01
23:01 IllvilJa left 23:04 jimmy joined 23:06 jimmy left 23:12 raiph left 23:13 raiph joined
timotimo .tell japhb i think github.com/japhb/perl6-bench/issues/13 should be fixed by the automatic timeout detection stuff we have 23:13
23:18 raiph left, raiph joined
sergot m: say 1 andthen say 2; 23:20
camelia rakudo-moar 7fd2f6: OUTPUT«1␤2␤»
23:22 raiph left, raiph joined 23:23 adu left 23:28 raiph left 23:29 raiph joined 23:34 pecastro joined
Kristien m: 42.numerator.say 23:34
camelia rakudo-moar 7fd2f6: OUTPUT«No such method 'numerator' for invocant of type 'Int'␤ in block <unit> at /tmp/nKus3bDaHC:1␤␤»
Kristien :(
Why does Int not do Rational?
masak because Perl 6 doesn't have a numerical tower based on subtyping, as you seem to assume. 23:37
japhb .tell timotimo Thank you! I've commented and closed it.
masak both Int and Rat are Real and Numeric, though.
japhb chuckles: All remaining open issues for perl6-bench were opened by timotimo++ 23:39
Wait, my .tell didn't work. What happened to yoleaux?
Kristien yoleaux is dead
masak long live yoleaux 23:40
japhb Someone hit it with the paddles ... or a resurrection spell, your choice of genre solutions
MHF (Medium Hanging Fruit) for the interested: Get Proc::Async working on r-j 23:41
vendethiel
.oO( not long live the yoleaux ! )
 23:45
23:47 woolfy joined 23:50 kjs_ joined 23:54 kjs_ left 23:59 adu joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!