»ö« | perl6.org/ | nopaste: paste.lisp.org/new/perl6 | evalbot usage: 'perl6: say 3;' or rakudo:, alpha:, pugs:, std:, or /msg p6eval perl6: ... | irclog: irc.pugscode.org/ | UTF-8 is our friend!
Set by lichtkind on 5 March 2010.
jnthn Don't make me break out the ASCII YMCA dance. 00:00
diakopter refrains from making you 00:01
00:01 c1sung left, c1sung joined 00:06 crythias1 joined 00:08 crythias left 00:22 crythias joined 00:25 crythias1 left
mberends colomon: monitor updated with sampling and display options: gist.github.com/369869 00:25
colomon mberends: would change, but in the middle of a run with it right now. ;)
mberends o/ # going to sleep 00:26
colomon mberends: actually, for my taste the nicest change would be to switch it to using stderr instead of stdout.
mberends++
mberends colomon: thanks for the feedback, can make that another option 00:27
jnthn o/, mberends++ 00:28
00:35 dual left
colomon jnthn: hmmm. mberends++ tool is not showing an unexpected increase in memory consumption in mandelbrot. (yet, still at smallish numbers.) 00:37
oh, except I'm not factoring it correctly. 00:38
hmmm....
oh, there we go. 00:39
looked at properly, it suggests doubling the edge size a bit more than doubles the memory usage.
but it's nothing crazy yet. 00:40
jnthn OK.
colomon or maybe it is. 00:41
okay, here's what's crazy.
jnthn At the moment, a Complex is heavy. It's IIRC 2 PMCs at a Parrot level per Perl 6 object. A Complex is the Complex object itself + a Num for real part + a Num for imaginary part = 6 PMCs.
colomon it looks to me like it's doubling like it should be, more or less.
BUT... it's doubling from a HUGE base. 00:42
00:42 dual joined
jnthn I may try and patch Parrot pre-R* to get it down to 3. After R* when I do the native types work I should be able to get it down to 1. 00:43
colomon actually, Complex might be two Rats in this case.
jnthn Oh
Then wow. :-)
Rat = Rat object + the two parts = 6 PMCs. 00:44
So your total Complex number of two Rats is 14 PMCs.
.oO( no damm wonder it's slow )
colomon hold a sec 00:45
jnthn I suspect I may need to try writing that Parrot patch. :-)
colomon looks to me like it's taking about 4 MB per unit of edge size. 00:46
ack, it's hard to figure out what units we should be talking about.
If you think switching Rat to Num will make a noticeable, that's an easy enough test to do. 00:47
jnthn I'd be interested to know.
colomon anyway, 4 MB is insanely high for this. 00:48
jnthn What does a "unit of edge size" actually mean? 00:49
colomon well, you give it the width / height (ie edge size) on the command line. 00:51
doubling that quadruples the number of pixels calculated (because the image is square)
that affects the overall timing, and what I'm seeing is right in line with that.
double that also doubles the memory usage. 00:52
that sort of makes sense, because since for does not lazily handle its arguments, the entire list of complex values for each row of pixels must be in memory at once. 00:53
but at the same time, that implies that 161 Complex numbers requires 1.3 GB to store (according to the numbers I just got), which is insane. 00:54
If I'm figuring correctly (based on earlier numbers, too, not those) that's 4MB for each Complex number. 00:56
jnthn erm. 00:58
That...can't be right.
colomon I hope not.
jnthn Or at least, can't be the whole story.
colomon sure. 01:01
if you look at the code, it's a nested loop. 01:02
diakopter maybe parrot is allocating more memory unnecessarily (before it actually needs it)? 01:03
colomon all results are dumped each time through the outer loop, so that shouldn't be accumulating memory.
and the results suggest it isn't -- if it was, we'd be seeing the memory go up by 4 rather than by 2.
jnthn diakopter: Not sure. Allocation stratergy tweaks have in the past made notable differences though. 01:04
colomon my @line = subdivide($re + ($upper-right.im)i, $re + 0i, ($width + 1) / 2).map({ mandel($_) });
my $middle = @line.pop;
(@line, $middle, @line.reverse).map({ @color_map[$_] }).join(' ').say;
diakopter maybe more tweak possibilities exist
jnthn Undoubtably. 01:05
colomon is the inner loop, and increasing $width by 1 increases the memory usage by about 4MB.
diakopter in the meantime, strategery abound
01:07 rv2733 left
colomon going to run load the dishwasher, will check back in a few minutes to see how the Numified version is doing. 01:08
01:16 Chillance joined 01:19 wknight8111 left 01:20 ob1 joined 01:22 ob1 left
colomon jnthn: if mberends++' script is accurate, switching to Num shaves 17 seconds off the execution time, and 01:23
rakudo: say 1520201728 - 1516560384
p6eval rakudo 78faa0: OUTPUT«3641344␤»
colomon 3.6 MB off the memory usage. 01:24
jnthn :/ 01:25
Which is a tiny fraction of the whole. 01:26
colomon yup
jnthn Which would seem to suggest we're blowing our memory on something other than the storage.
colomon other than storage of Complex components anyway, yes. 01:27
jnthn Yes, that's what I meant.
diakopter guesses scopes & frames
jnthn The thing is, we seem to be back to while 1 { 1 } leaking again. 01:28
Which is really shit.
colomon could somehow be storage of the Arrays, I suppose...
jnthn Well, what diakopter said makes some sense too I fear. 01:29
I guess we're leaking those somehow.
diakopter maybe they're not being GCd
colomon even when memory usage gets up into the GBs? 01:30
jnthn The bigger problem is why.
diakopter (even if all references to them are being nulled)
jnthn Is it a GC mis-configuration.
Or is it lost pointers
Or something else.
diakopter GC mis-configuration or GC mis-implementation 01:31
01:31 sorear left
jnthn nod 01:31
01:31 sorear joined
diakopter try this 01:32
just do a nested loop of the same structure, but with no storage, just Int counters
and empty arrays though 01:33
so map is still called
jnthn nod
That'd be revealing.
01:34 drbean_ is now known as drbean
colomon I've just noticed that the inner loop is actually map, so in theory.... arrrgh. 01:34
diakopter that would tell you whether the problem is in the routine passed to map or in all teh corntol fowl around it
the control flow, even 01:35
colomon likes the idea of corntol fowl.
diakopter fights the downwards funargs problem for yet another hour 01:36
I should yearn to write in anagrams all the mite 01:37
at lates, where I see meth
colomon I really don't know how to coherently modify the algorithm to not use storage.
oh hey, though, one thing we could easily do is change max_iterations on the inner mandelbrot calculation. 01:38
increasing that (say) would cause additional inner loops / calculations without any difference to the longer term storage in the script. 01:39
so if just looping is using memory, it would show up that way. 01:40
diakopter apparently ye olde volcano will put Europe back into the Dark Sage
if cable news has anything to say about it. 01:41
colomon Dark Sage! oh no! I prefer rosemary!
my wife just discovered our insurance covers cleanup from volcanic ash! 01:44
01:44 meppl left
diakopter but not damage from it? 01:44
diakopter hides the garage door openers
colomon she's in the debris removal section, so we don't know. 01:45
our shrubs are covered against aircraft, too. 01:46
jnthn Better than having your shrubs covered in aircraft, I guess.
diakopter looks around for the punkings 01:47
colomon web search suggests it's been 2.5 billion years since there were active volcanos in Michigan, so they must be worried about the supervolcanos out west....
diakopter yeah but big enough eruption can have global effect
spaceship-wide, I should say. 01:48
jnthn Cable news is probably full of crap.
Some airlines are even doing test flights to see what happens to their aircraft if they fly them in the closed airspace.
Apparently, nothing much yet. 01:49
colomon diakopter: sure, but is there likely to be a worldwide volcano eruption which: 1) causes measurable amounts of ask damage to my house and 2) doesn't end civilization in the process?
*ash
diakopter I guess not. :'(
someone should use bombs to convert the volcanic cone to something that just oozes magma 01:54
01:57 Psyche^ joined
jnthn What could possibly go wrong? :-) 01:58
01:58 nihiliad left 02:01 Patterner left, Psyche^ is now known as Patterner
colomon just started a size 1001 run for overnight. by my reckoning it should take about 5 hours to run... 02:01
02:03 rgrau left 02:05 envi^home joined
jnthn -> sleep now 02:07
o/
diakopter o/ 02:08
colomon o/
sorear colomon: Why? Stress testing? 02:15
02:18 PZt joined
diakopter .. start with some tomahawks, if that doesn't work, maybe a trident. 02:19
02:21 meppl joined 02:28 crythias left 02:34 _jaldhar joined 02:35 eternaleye joined
diakopter hm. 02:38
sorear also, you do not need volcanos to have fun with ash
diakopter sorear: you might be interested: 02:41
I'm trying to solve the downward funargs problem for perlesque
sorear funargs problem? ick
diakopter er
yes
downwards funarg, I mean. :|
en.wikipedia.org/wiki/Funarg_proble...rg_problem 02:42
The second paragraph under "Practical Limitations" is exactly what perlesque does. 02:43
diakopter keeps at it 02:44
03:06 cledu left 03:13 jaldhar_ joined 03:14 _jaldhar left 03:35 fridim joined 03:36 jaldhar_ left, jaldhar_ joined 03:43 meppel joined 03:46 meppl left, meppel is now known as meppl
sorear diakopter: the Wikipedia Funarg Problem is ... not at all the same as the funarg problem I know 03:47
diakopter oh 03:52
sorear: which one do you know 03:53
the one described in terms of free variables and their binding and activation environments?
03:58 Chillance left
diakopter figured it out. 04:01
diakopter can't decide which is more fun: having cake or eating cake. 04:05
sorear diakopter: yes, the free variable binding problem in Lisp-2 04:07
the one which was solved by the invention of lexical scope
diakopter oh :) 04:09
you're saying "lexical scope" makes the binding environment synonymous with the declaration environment. 04:11
there could have been other ways to do it
since there are a couple levels of "bind" there
sorear the original funarg problem was a semantic problem, not an implementation one 04:13
so it could be fixed by conceptual leaps
diakopter I can imagine a "bind_free_vars" function that is a compile-time instruction to "curry" free variables to particular scopes...
(treating a function declaration like a function template)
(so it doesn't actually curry the values, just the slots) 04:14
now _that_ would be metaprogramming.
04:16 jaldhar_ left
diakopter kindof a generalization of 'compose' and 'curry' 04:16
04:17 jaldhar_ joined 04:19 XaeroOne joined
sorear sooooo.... the CLR uses dynamic scoping? 04:36
I had no idea Microsoft was /that/ crazy
diakopter nope
sorear so why are you having a funarg problem?
diakopter CIL (its assembly language) doesn't have scoping
at all
b/c perlesque has lexical and dynamic scoping (and closures, and continuations for that matter), and I'm compiling perlesque to CIL 04:37
so all closures must be strongly typed on each's full signature
so I need variables of particular closure-types to be able to store closures with identical signatures. 04:38
(but different function bodies, obviously)
.. and I'm ignoring co- and contra-variance on signatures entirely. 04:39
stackframes are reified as CIL objects on the heap, and codegen to resolve all lexical slots is done at compiletime 04:40
sorear sounds like perlesque is a real exercise in Turing completeness 04:47
TimToady oddly, Perl 5 gets around the whole issue without putting the activation stack on the heap 04:48
at the expense of having to clone closures, basically
diakopter sorear: Guy Steele explains the definition drift/specification of the funarg problem: people.csail.mit.edu/gregs/ll1-disc...03278.html
TimToady: but that's the same thing.. 04:49
aren't the cloned closures activation records themselves?
TimToady yes, well, you might say it's using indirection to solve the problem :) 04:50
sorear TimToady: essentially, Perl 5 does keep activation records on the heap; the pad is only an index, lexicals are 'struct sv' in the arena
diakopter ok. perlesque does the same thing in an OO way; a closure class is created for each subroutine, and instances of that closure class hold links to their proper parent scopes 04:51
TimToady yeah, I suppose, depending on whether you consider the lexpads to be part of the activation record :)
diakopter each perlesque subroutine also has a 'frame' class, from which its activation records are spawned; those are the lexpads themselves, and also hold links to each's Caller frame 04:52
04:53 jaldhar_ left, jaldhar_ joined
diakopter sigh. there's also a 'bind' interface created for each unique subroutine interface, and a 'return' interface created for each unique return type. *sigh* so that it's all still strongly-typed. 04:54
Yes, OOdles of casting.
04:55 jaldhar_ left
diakopter s/subroutine interface/subroutine signature/ 04:55
04:55 jaldhar_ joined
sorear TimToady: That reminds me of another question I've had for a long time. Why does Perl 5 have one stack per CV, instead of the more traditional approach of intermixing all activation records on a common stack? 04:55
diakopter someone help the n00b; what's a CV
sorear the perl 5 data structure which represents a sub or closure instance (or format, but let's not go there) 04:57
diakopter ok
sorear the target of CODE references
04:57 gfx joined
sorear s/Perl/perl/ 04:58
04:59 ReiniUrban joined
diakopter sorear: that's an interesting question. actually I considered a model something like that 04:59
diakopter waits for TimToady comment 05:00
sorear: I guess that's one way of tracking recursion depth
05:00 justatheory left 05:01 rurban left, ReiniUrban is now known as rurban
diakopter oh, I see. the caller link on the last frame must point to the CV itself and not to the top record on that CV's stack. 05:01
(I'm not saying I know *why* it was done that way) 05:02
oh 05:03
sorear um, why would anyone want to track recursion depth
diakopter sorear: another benefit is that it separates the code from the lexpads, so that code can be changed
after a frame has been entered
I dunno; warnings tracks it, doesn't it? 05:04
sorear activations and code are separated in every model
TimToady well, last I looked, the CVs themselves were all on their own stack separate from the lexpads, but that might have changed since then
sorear I'm not asking about CVs 05:05
I'm asking why there are so many padlists
Why isn't there a single global PL_padstack?
05:06 c1sung left
TimToady actually, the CX statck 05:06
*stack
one reason for tracking recursion level (besides warning on deep recursion) is that we cache a lexpad at each level
sorear or, for that matter, the lexicals /could/ have been stored on PL_stack
TimToady so it's only allocated from the heap the first time we want one at that level
Perl 5 keeps a lot of stacks that most other implementations combine 05:08
you might think of the various stacks as having a rudimentary form of typing 05:09
diakopter hee 05:10
YA typing layer
TimToady I don't recommend you do it that way
it was mostly driven by trying to fake out C macros to do the right thing
diakopter me?
diakopter wonders if there are "compile Perl successfully" test cases in C compilers' regression suites 05:11
TimToady that was a major test in gcc, at least at one time
diakopter not surprised
seems it would be difficult to break down the tests... 05:12
TimToady at the time, perl put more stress on the compiler than many other open source projects could
this was before the big projects we know and love today 05:13
diakopter I guess that's one way to out-lock-in the vendors
05:14 Xaero joined 05:16 molaf joined
sorear reminds me of the way we used to smoke new DIMMs using gcc before memtest86+ came out 05:16
typically on Linux or Xfree86
05:17 c1sung joined, XaeroOne left
sorear "run make World; make realclean in a loop for 24 hours. Check for cores." 05:17
05:25 justatheory joined 05:33 phenny left 05:34 nsh left, sbp left 05:35 Sarten-X joined 05:37 Maxt0r joined, nsh joined
Maxt0r hi 05:37
05:37 sbp joined
diakopter Maxt0r: hi 05:38
Maxt0r i have a problem with a split function am using can anyone hepl plz
i have a file i open it n read form it as follows :open (FH, $filename) || die ("Could not open $filename!");
then i pass info to an array : my @xxx =<FH>; 05:39
diakopter TimToady: you still around?
Maxt0r: this is a Perl 6 channel
Maxt0r foreach $line (<FH>)
perl 6?^
whats the dif
can u help me out?
diakopter a lot of differences. :) see perl6.org
Maxt0r can u help me 05:40
with this quiestion is simple
my split function is not spliting lols
diakopter probably #perl can help you out more
Maxt0r ok
nvm ty
diakopter yw
sorear Where did all the vowels go? 05:44
diakopter gn wth th wnd
sorear: check this out :) 05:45
perlesque: my $last = 4; my $sub_a = sub int foo (int $arg0) { say(my $b = $arg0 + 5); $last = $b; return $b; }; my $sub_b = $sub_a; $sub_b(4); $sub_b = sub int bar (int $arg0) { say(my $b = $arg0 + 4); if ($b>19) { $sub_b = $sub_a; }; $last = $b; return $b; }; while ($last < 100) { $sub_b($last); };
p6eval perlesque: OUTPUT«9␤13␤17␤21␤26␤31␤36␤41␤46␤51␤56␤61␤66␤71␤76␤81␤86␤91␤96␤101␤»
05:45 fridim left
diakopter full type inference for expressions 05:45
(as long as you like either ints or closures) :D (but yes, it'll work for any CLR type) 05:46
and those $sub_a and $sub_b variables can store each other's routines, since they have identical signatures.
(as you can see, the output series increments by 4 until it's > 19, when it starts incrementing by 5) 05:47
sorear: cool or what? :)
sorear I've implemented Damas-Milner
it's going to take more than /that/ to impress me 05:48
05:48 fridim joined
diakopter it's the closures-as-variables I'm pointing at 05:48
as soon as I can get type-prototyping working, closures can be arguments and return values, too 05:50
TimToady: actually that presents a problem for me.
I mean, I have an idea to work around it in perlesque, but it'll make it no longer a Perl 6 subset 05:51
how to represent the types of closures in signatures.
in C#, it would be Func<param0_type, param1_type, param2_type, ... , return_type> and Action<param0_type, param1_type, param2_type, ...> for routines that return Void 05:53
sorear I think GRIN might be relevant to you
diakopter GRIN? 05:54
05:54 XaeroOne joined
sorear the Graph Reduction Intermediate Notation 05:54
a very interesting body of work concerning strongly-typed implementation formulations for higher-order languages 05:55
not /directly/ relevant, you aren't implementing a global superoptimizer for Haskell, but ... has parallels
05:56 Xaero left
diakopter heh 05:56
in this case, I just need a notation
jql hmm... I guess you can't just steal parentheses as function type description notation 05:57
diakopter TimToady will reply sometime... 05:59
my idea was to use other existing closure variables as prototypes
so in the example above, I worked around it by assigning $sub_a a dummy empty routine 06:00
jql hmm, there is a syntax for closure parameter declarations already. spiffy 06:01
diakopter but instead of a typename, if there was some sig_of($other_strongly_typed_closure_with_signature) operator, I could put that in place of the typename in the function signature
orly?
do tell
jql my sub testit (&testcode:(Int)) {testcode(12)}
diakopter oh hah
where'd you find that
jql: thanks! :) 06:02
jql multi sub t1 (&code:(Str --> Bool)) { 'Str --> Bool' };
return types, too
diakopter ahh, I see
ok, now what if &code and t1 there have types that are mutually recursive 06:03
jql whips out the handy book of combinators
I believe.... $math
diakopter waits 06:05
diakopter goes to get a drink 06:07
jql hmm... perhaps multi sub t1 (::T, &code:(::T --> ::U)) or some such nonsense
with a --> U at the end 06:08
multi sub t1(::T, &code:(T, ::U)) --> U, if that was a valid thing to do. *shrug* 06:09
multi sub t1(::T, &code:(T --> ::U)) --> U # comma is *so* tempting 06:10
06:17 Maxt0r left
diakopter ok 06:31
cognominal nqp: say(<a b c>) 06:32
p6eval nqp: OUTPUT«3␤»
cognominal nqp: my @a := <a b c>; say(@a) 06:33
p6eval nqp: OUTPUT«3␤»
diakopter nqp: (<a b c>).say 06:34
p6eval nqp: OUTPUT«Method 'say' not found for invocant of class 'ResizablePMCArray'␤current instr.: '_block11' pc 0 (EVAL_1:6)␤»
diakopter /o/ 06:35
cognominal nqp: my @a := ('a', 'b', 'c'); say(@a) 06:38
p6eval nqp: OUTPUT«3␤»
cognominal nqp: my @a := ('a', 'b', 'c'); say(@a[0]) 06:39
p6eval nqp: OUTPUT«a␤»
06:41 rv2733 joined, Exodist_ is now known as Exodist 06:52 rv2733 left, phenny joined 07:07 gfx left 07:10 finanalyst joined 07:14 Kyril left
finanalyst perl6: role A { has $.att}; my $x .=new(:att<$y>); say $x.att 07:19
p6eval elf 30402: OUTPUT«Can't call method "att" on an undefined value at (eval 123) line 9.␤ at ./elf_h line 5881␤»
..pugs: OUTPUT«*** No such method in class Item: "&att"␤ at /tmp/Gy8CVyuj6W line 1, column 49 - line 2, column 1␤»
..rakudo 78faa0: OUTPUT«Method 'att' not found for invocant of class 'Any'␤current instr.: '_block14' pc 29 (EVAL_1:0)␤»
sorear is perl6: just an alias for elf: pugs: rakudo: ? 07:20
finanalyst rakudo: role A {has $.col}; my A $x .= new(:col<$y>); $x.col.say 07:21
p6eval rakudo 78faa0: OUTPUT«$y␤»
07:21 XaeroOne left
sorear and if so, why no alpha: ? 07:21
07:21 XaeroOne joined
finanalyst it seems to me that $y should be flagged as not having been declared 07:22
07:22 cotto_w0rk joined
finanalyst any one able to comment? 07:24
07:32 nsh left, nsh joined
mberends finanalyst: the <$y> is equivalent to ('$y'), therefore $y are two literal characters 07:34
finanalyst mberends: thanx, i just realised that too. just as i was writing a rakudobug! 07:36
but didnt send it
mberends re-boots to Windows to try process memory monitoring there too 07:39
07:39 mberends left 07:45 hicx174 left 07:52 mberends joined 07:54 yahooooo7 joined 07:55 gabiruh_ joined, cookys_ joined 07:56 diakopte1 joined, phenny left, yahooooo left, cotto left, cookys left, Helios left, diakopter left, pmurias joined, frettled_ joined 07:59 Helios joined 08:00 iblechbot joined 08:11 justatheory left 08:15 justatheory joined 08:17 ejs joined 08:22 ejs left, ejs joined 08:38 pmurias left 08:48 hicx174 joined 08:49 _jaldhar_ joined 08:50 jaldhar_ left, ejs left 08:52 M_o_C joined 08:55 ejs joined, ejs left
moritz_ good morning 08:55
mberends good morning moritz_ 08:56
09:06 justatheory left 09:17 XaeroOne left
finanalyst rakudo: role A { has $.x = 2; method doit($y) { return $.x * $y } }; my A $z; say $z.doit(1.5) 09:19
p6eval rakudo 78faa0: OUTPUT«Type objects are abstract and have no attributes, but you tried to access ␤current instr.: 'perl6;Attribute;accessor_helper_ro' pc 3554 (src/gen/Attribute.pir:400)␤»
finanalyst can anyone tell me what I am doing wrong here? 09:20
moritz_ finanalyst: you are not creating a new instance of A
and still you access $.x, which in turn relies on $!x
which is an attributte
finanalyst isnt the 'my A $z' doing that 09:21
moritz_ which you can't access outside of an instance
finanalyst: no, that's merely a type constraint
finanalyst so how does a role become a class?
moritz_ automatically, if you want it
finanalyst rakudo: role A { has $.x = 2; method doit($y) { return $.x * $y } }; my A $z = .new; say $z.doit(1.5) 09:22
p6eval rakudo 78faa0: OUTPUT«Type check failed for assignment␤current instr.: '&die' pc 17293 (src/builtins/Junction.pir:404)␤»
moritz_ rakudo: role A { has $.x = 2; method doit($y) { return $.x * $y } }; my $z = A.new; say $z.doit(1.5)
p6eval rakudo 78faa0: OUTPUT«3␤»
finanalyst yep. beat me to it
moritz_ I suspect that's a bug in rakudo 09:23
rakudo: role A { }; my A $x .= new()
finanalyst rakudo: role A { has $.x = 2; method doit($y) { return $.x * $y } }; my A $z .= new; say $z.doit(1.5)
p6eval rakudo 78faa0: ( no output )
rakudo 78faa0: OUTPUT«3␤»
finanalyst why a bug? 09:24
moritz_ I don't see where the type check failure is coming from
but I'm not quite awake yet
09:32 jaldhar joined 09:33 _jaldhar_ left 09:53 frettled_ is now known as frettled
moritz_ sorear: (backloggin) yes, perl6: just runs different Perl 6 compilers 09:57
perl6: say "foo"; # one without differences, I hope 09:58
p6eval elf 30402, pugs, rakudo 78faa0: OUTPUT«foo␤»
10:06 masak joined
masak oh hai, #perl6 10:07
mberends good masak, morning
masak is having a good morning
GGE now runs all tests under master. \o/ 10:08
it passes almost half of them, too :)
(as opposed to 100% on alpha)
(well, except for TODOs, of course)
rakudo: role A { has $.x = 2; method doit($y) { return $.x * $y } }; my A $z = .new; say $z.doit(1.5) 10:09
p6eval rakudo 78faa0: OUTPUT«Type check failed for assignment␤current instr.: '&die' pc 17293 (src/builtins/Junction.pir:404)␤»
masak rakudo: class A { has $.x = 2; method doit($y) { return $.x * $y } }; my A $z = .new; say $z.doit(1.5) 10:10
p6eval rakudo 78faa0: OUTPUT«Type check failed for assignment␤current instr.: '&die' pc 17293 (src/builtins/Junction.pir:404)␤»
masak rakudo: class A {}; my A $x .= new
p6eval rakudo 78faa0: ( no output )
masak interesting.
rakudo: class A { has $.x }; my A $x .= new
p6eval rakudo 78faa0: ( no output )
masak rakudo: class A { has $.x = 2 }; my A $x .= new
p6eval rakudo 78faa0: ( no output )
masak oh! 10:11
.new
never mind :)
sorear GGE?
masak it's a port of PGE to Perl 6. 10:12
PGE is pmichaud++'s Perl 6 grammar engine, written in PIT.
*PIR 10:13
10:13 snarkyboojum joined
masak snarkyboojum! \o/ 10:13
snarkyboojum masak: hello! o/
masak snarkyboojum: I might start that 'immediate blocks' branch today in Yapsi.
snarkyboojum masak: just having fun wasting time getting my new macbook pro up and running - rakudo builds quickly on an i7 :) 10:14
masak ooh, nice.
snarkyboojum masak: sounds awesome - I'll have to follow along 10:15
masak I'm hoping to keep that branch fairly short-lived.
snarkyboojum sure 10:16
masak basically just add tests like 'my $a; { ++$a }; say $a', and implement the bare minumum to get it working.
snarkyboojum would be interesting to watch it in progress however short lived :)
masak I see it as the big step Yapsi will take this month. 10:18
snarkyboojum masak: feel free to let me know if there is anything I can add or think about
masak it'll enable a lot of other nice things, including ifs, loops and subs.
snarkyboojum: how's it going with Tardis? 10:19
snarkyboojum masak: ah - looked at it, I'll probably have to bounce stuff of you, as I haven't gotten my hands dirty yet
masak ah, ok.
snarkyboojum looked like a rewrite of Tardis.pm 10:20
etc
masak aye, seems likely.
and it'll need yet another one when we have nested scopes...
snarkyboojum excellent - perhaps I can forge ahead and see what I can accomplish/break :)
masak \o/ 10:22
10:31 Alias joined 10:36 _jaldhar joined, jaldhar left 10:37 IllvilJa left
masak continuing the 'our'/'my' discussion from yesterday: what default scoping do 'regex' et al. have, such that one is now required to add 'our' or 'my' to it to make it stick in an appropriate lexpad? 10:38
moritz_ masak: 'has' 10:39
sorear I thought 'regex' could only appear in classes?
masak moritz_: thought so.
moritz_ sorear: that restricted was lifted for convenience 10:40
masak moritz_: and I suppose that is because they're really methods in sheep's clothing.
moritz_ masak: aye 10:41
sorear moritz_: wasn't you or masak complaining about language bigotry last week on not allowing regex outside grammar
masak that was I.
it works now, if you do 'my'/'our'. 10:42
moritz_ in rakudo?
10:42 iblechbot left
masak aye. 10:42
moritz_ rakudo: our regex foo { '.' }; say '.' ~~ /<foo>/;
p6eval rakudo 78faa0: OUTPUT«Method 'foo' not found for invocant of class 'Regex;Cursor'␤current instr.: '_block61' pc 613 (EVAL_1:239)␤» 10:43
moritz_ rakudo: my regex foo { '.' }; say '.' ~~ /<foo>/;
p6eval rakudo 78faa0: OUTPUT«Method 'foo' not found for invocant of class 'Regex;Cursor'␤current instr.: '_block65' pc 601 (EVAL_1:229)␤»
masak hm.
or not.
moritz_ what am I doing wrong? :-)
masak just too early, I guess. :/
I guess what I'm asking -- and this can be seen as the last throe of my 'bigotry' claim -- is why 'regex' et al. don't default to 'my' scope, and things would work just like they did in alpha. 10:44
or, if that helps, they could default to 'my' scope outside of classes and roles, and have 'has' scope inside. 10:45
moritz_ because then they couldn't be called as methods, and thus benefit from inheritance 10:46
masak I won't press it. maybe the consistency is important somehow. 'my' is not that hard to remember to add outside of a class, grammar or role. 10:48
moritz_ well
there are two things where context sensitive defaults would be benefitial
1) multi foo could default to 'method' inside classes, roles and grammars
and default to subs elsewhere 10:49
masak I've had that same thought.
moritz_ 2) regex foo could default to 'has' inside classes, roles and grammars, and to 'my' everywhere else
masak it seems we are in violent agreement.
moritz_ that goes against the "list of arbitrary things to remember" rule
masak I guess so. 10:50
one could put those context-sensitive defaults in a module, I guess.
moritz_ use magic::default; 10:51
on a totally unrelated note I've taken care that Date objects are immutable 10:53
I guess DateTime objects are not, right?
masak no, and I thought quite a bit about that. 10:54
CPAN's DateTime objects are mutable, and make quite good use of it.
moritz_ aye, I think it makes sense
masak if you look at the IRC logs, you'll see that my goal was to produce a fairly unchanged subset of CPAN's DateTime :)
I didn't want to break that rule by introducing immutability. 10:55
moritz_ Perl 6 has this tendency to keep simple objects immutable, and more complex objects mutable
so it kind of fits
masak also, it was fun to make use of intelligent rw accessors (not yet impl in Rakudo)
moritz_ Mark J. Reed on p6l: 10:57
my $bday = new Date.new(1968, 5, 5);
my $specifically = DateTime.new(:date($bday), :hour(20), :minute(47));
or at least, Date should have a method that returns it's value as
pairs suitable for passing to DateTime.new.
another syntax I thought about was 10:58
$bday.DateTime(:hour(20), :minute(47))
masak I think I prefer the DateTime.new version. 10:59
moritz_ currently I lean towards DateTime.new(:date($bday), :hour(20), :minute(47)) - but since DateTime is your domain, I want your opinions... right
mberends this discussion interests me, but I'll be out for a few hours. Happy bikeshedding!
11:00 mberends left
masak seems the bikeshedding is done already :) 11:00
moritz_ mberends: so far we kept the bikeshedding low by having just masak and me discussing it :-)
masak: do we still need DateTime.today() then? 11:01
masak oh!
no, we don't. we'll just force people to use .Date :) 11:02
or, I suppose Date.new(...)
I wouldn't mind seeing both in that case.
moritz_ what's the ...?
masak Date.new( $dt ), perhaps. 11:04
or some aptly named named :)
moritz_ ok
masak rakudo: say "foo".subst( /(o)/, { $0.uc } ) 11:07
p6eval rakudo 78faa0: OUTPUT«Method 'postcircumfix:<[ ]>' not found for invocant of class ''␤current instr.: '!postcircumfix:<[ ]>' pc 10733 (src/builtins/Role.pir:90)␤» 11:08
masak I'd really like for the above to work again by Rakudo Star.
moritz_ again? never did
afaict
masak alpha: say "foo".subst( /(o)/, { $0.uc } )
p6eval alpha 30e0ed: OUTPUT«fOo␤»
moritz_ huh?
masak I use it all over the place in my code. 11:09
moritz_ rakudo: say "foo".subst( /(o)/, -> $/ { $0.uc } )
p6eval rakudo 78faa0: OUTPUT«Method 'uc' not found for invocant of class 'Regex;Match'␤current instr.: '_block58' pc 757 (EVAL_1:257)␤»
masak not cool...
moritz_ aye
masak but that's not what I meant.
I meant the closure having the correct $/
moritz_ I know 11:10
11:10 rgrau joined
lisppaste3 moritz_ pasted "DateTime <-> Date integration" at paste.lisp.org/display/97950 11:14
masak looks
s/aboved/above/ 11:15
moritz_ is glad for git rebase -i 11:16
masak it rocks.
moritz_ masak: for a moment I was thinking you meant by patch :-)
masak :) 11:17
moritz_: looks fine. I'm glad Date and DateTime start talking to each other.
Trashlord hello, friends
masak moritz_: do you think that it makes sense to keep DateTime mutable but Date immutable?
Trashlord: hi! 11:18
moritz_ masak: yes
pugssvn r30403 | moritz++ | [S32::Temporal] Date <-> DateTime integration
r30403 |
r30403 | * DateTime constructor which takes a Date object, as suggested by Mark J. Reed
r30403 | * DateTime.Date coercion method
r30403 | * Remove DateTime.today, since date-only calculations are meant be done by the
Trashlord what's up?
pugssvn r30403 | Date class
r30403 | * Date.new() constructor taking a DateTime object
moritz_ the sun is up
masak Trashlord: GGE is back up.
(but still fails tests)
Trashlord I like the sun
masak I hear the sun is coming back to this hemisphere this summer. 11:19
heh, github.com/masak/gge/commits/ng-compat 11:21
(you can see where my focus has been this month) 11:22
pugssvn r30404 | moritz++ | [t/spec] initial tests for Date, taken from my Date module
11:26 fridim left, baisong joined
Trashlord hey, is there a significant difference between rakudo and pugs? 11:27
sorear yes 11:28
masak both are Perl 6 implementations.
one stopped being in active development three years ago.
sorear pugs has been undeveloped for several years
masak even at that point, it was due for a major refactor. 11:29
I don't think it ever landed :/
Trashlord oh
sorear that's the most common time for software to die
right when an overhaul is needed
masak Pugs had a bus number of 1, and though the main dev wasn't hit by a bus, the effect was pretty much that :/ 11:30
Trashlord haha 11:31
sorear as compared with Parrot, which has a bus number of 0 11:32
masak Trashlord: no, not much haha at all :/
sorear the developers are all still here but we don't have a clue what we're working on
replacing the entire Parrot development team today would have no significant effect 11:33
Trashlord heh 11:39
dalek kudo: 83650e0 | moritz++ | (2 files):
first shot at Date
 11:48
masak I just had an idea, tried to implement it in Rakudo master, and it worked! \o/ 11:54
1. implement a custom data structure, e.g. Tree
2. implement &map on that data structure
3. ???
4. profit!
gist.github.com/370183
11:54 wknight8111 joined
moritz_ you know that ??? fail()s? :-) 11:55
masak it didn't in this case :)
I'm thinking whether an example such as this belongs in the book.
it's very self-contained, and it shows something which you can definitely not do easily in Perl 5.
colomon I think the spec expects you to implement an iterator for the tree, BTW. :) 11:56
masak colomon: could you produce code that shows how to do that?
I'd be much obliged.
11:57 clintongormley joined
colomon masak: I don't think I have time at the moment, but if no one has by afternoon I'll give it a try. 11:57
masak nodnod, no rush.
moritz_ rakudo: say 0 cmp 1
jnthn o/
p6eval rakudo 78faa0: OUTPUT«-1␤»
masak anytime before 2038 would be good :)
jnthn masak: Yes, regexes are has scoped
colomon though in that case, at least with current code, you wouldn't get a tree back.
masak jnthn: \o 11:58
jnthn masak: The reason <foo> doesn't work yet is that it always goes looking for a method, rather than checking out the lexical scopes first.
colomon masak++ for picking a good simple project to be exploring at this time.
moritz_ jnthn: so NYI, right?
jnthn Well, the compiler can go look I guess...
moritz_: Correct.
masak colomon: but wouldn't the resulting &map give you back a Tree?
jnthn moritz_: I did the first half of the work (making our and my work on regexes). 11:59
colomon masak: no, if it takes a normal iterator, it will give you back a MapIterator, not a Tree.
jnthn moritz_: Still need to do the next bit.
masak colomon: I have simple ideas all the time. it's just that some of them turn out to be very difficult in practice :P
colomon automatically getting back a Tree is something we discussed WRT the hypers, but none of that is implemented yet afaik.
11:59 Trashlord left
masak colomon: I confess I don't grok MapIterators. do they have to do with laziness? 12:00
colomon yes.
gotta run now. more later
masak anyway, my point was that with a few lines of code I can do map over a Tree, and that's pretty useful. 12:02
jnthn heh, once for === map, that means you'd be able to do a for loop over the tree too. ;-) 12:03
masak o.O
jnthn Maybe.
:-)
masak didn't even think of that 12:04
moritz_ ohh 12:05
I forgot to actually add Date.pm to the repo
bad moritz_--, no cookie
done that now 12:06
dalek kudo: 5ab1c03 | moritz++ | src/core/Date.pm:
actually add Date.pm, moritz--
 12:11
moritz_ 82 new tests due to Date.pm 12:12
12:15 rv2733 joined
masak uses the slowness of Rakudo to his advantage by starting a script, modifying it, and starting a second, modified version of it 12:15
12:15 Trashlord joined 12:16 Trashlord is now known as Guest61693
arnsholt masak: I'm not entirely sure that counts as an advantage =p 12:17
masak it's one of those conditional advantages, you know :P 12:18
at the very least, it helps me find bugs faster than if I weren't doing it.
arnsholt Heh
masak hm. it seems I've finally arrived at a test failure which requires me to think about the semantics of the program, as opposed to consmetics-level stuff. 12:20
jnthn oh noes! ;-) 12:24
masak for some reason, the *: quantifier is backtracking. bad, bad quantifier. 12:25
moritz_ in GGE, or in nqp? 12:27
masak in GGE.
wknight8111 How do I run the test t/spec/S06-signature/introspection.t? 12:37
masak perl6 t/spec/S06-signature/introspection.t
moritz_ make t/spec/S06-signature/introspection.
masak but that's a weird use of make 12:38
moritz_ masak: not at all
masak: it runs fudge if necessary
masak make specifies the target, not they source.
s/they/the/
moritz_ well, the target is to run that test file 12:39
s/target/goal/ perhaps
masak yes, that's the weird bit.
make makes things.
and side effects don't count, usually :)
moritz_ it's only an implementation detail that sometimes there's a .rakudo file involved
wknight8111 build...takes....too.....long.....
moritz_ wknight8111: then speed up parrot :-)
12:42 jaldhar_ joined
wknight8111 Speedups to Parrot aren't going to cut down this build time by too much 12:43
12:43 _jaldhar left
jnthn Why not? The compilers are Parrot programs. 12:43
That's like saying speedups to Parrot aren't going to cut down any times by too much. 12:44
Which may be true...but... :-/
wknight8111 how do I jnthn: Speeding up parrot by 50% will speed up the build by at most 50% 12:45
to really cut time in that build process, we need to reduce the algorithmic complexity of it
moritz_ well
one step is to split the compilation of these .pm files
which is non-trivial
jnthn wknight8111: I think we'd all very happily take a 50% speedup in the build! 12:46
masak moritz_: 'split' in what sense?
moritz_ masak: compile each of them separately
masak ah, ok.
moritz_ I don't know if it's really going to be faster in sum, but it's much easier to parallelize
12:46 finanalyst left
masak moritz_: I was thinking along those lines the other day. 12:47
12:47 finanalyst joined
masak moritz_: one things that one could already parallelize quite easily would be compilation of 'leaves', or more exactly, clades in the tree which don't depend on other things. 12:47
moritz_ masak: well, they would be all independent, more or less 12:48
jnthn wknight8111: We can surely get the parse time's complexity down, but the grammar action methods I'd be less sure on. I mean, most of them just build ASTs. Stages after that walk said AST to make POST, which is walked to make PIR. Beyond that it's IMCC, which I guess could be sped up some by algorithmic improvements.
moritz_ one thing we could improve is memory footprint by not keeping those $/ around for every node of the parse tree 12:49
I mean it's enough to store the position inside the string, and the string once
jnthn Yes, that is certainly true.
(I guess you mean "every node of the PAST tree") 12:51
moritz_ yes 12:54
otoh the parse tree still keeps references to those same Match objects
so not much gain at all
unless we implement parse tree pruning... do we? :-) 12:55
jnthn The trouble is knowing when could you prune.
I mean, what if an action method wants to dig a few levels down, etc. 12:56
moritz_ kinda hoped that EXPR would be such a boundary
jnthn That could be a good one. 12:57
jnthn afk for a bit 12:58
12:58 meppl left 12:59 ReiniUrban joined
masak sushi & 13:00
13:01 rurban left, ReiniUrban is now known as rurban
wknight8111 mmmm...sushi 13:02
13:03 wknight8111 left, fridim joined 13:04 M_o_C left 13:06 xinming_ joined 13:08 fridim left 13:19 dual left 13:24 iblechbot joined
masak relishes the sushi 13:29
moritz_++ # trac.parrot.org/parrot/ticket/1560 13:33
that's why you were talking about t/spec/S06-signature/introspection.t
13:38 M_o_C joined, M_o_C left, M_o_C joined 13:40 M_o_C left 13:52 xinming_ left 13:54 xinming joined 14:04 kfo_ joined 14:06 xinming left 14:07 kfo_ is now known as kfo 14:12 xinming joined 14:27 pmurias joined
pmurias diakopte1: hi 14:27
14:27 dual joined 14:37 snarkyboojum left 14:59 mikehh left 15:01 pmurias left 15:04 iblechbot left
colomon seems to have found the setting on mandelbrot-color.pl that makes it swap even on his big 64-bit Linux box... 15:11
15:12 finanalyst left, finanalyst joined
colomon It's using up 5.7 GB at the moment. 15:15
15:15 justatheory joined
colomon masak: ping? 15:30
15:30 mberends joined
colomon mberends++ # I think we learned a good bit using your script last night. 15:31
mberends :) current trying to port it to Windows
masak colomon: pong
colomon masak: trying to take a quick look at Tree before lunch (while watching the boy). 15:32
masak awesome.
colomon weirdly, adding does Iterable to Tree breaks everything.
15:32 mikehh joined
masak is in the middle of hunting down a very elusive but possibly disastrous rakudobug 15:32
colomon masak++ 15:33
masak the kind I didn't think we had anymore.
colomon rakudo--
masak of course, at this stage it might just be masak who deserves that --
we'll see :)
colomon rakudo++ # iterator works! 15:35
gist.github.com/370287 15:36
first time I've ever used a local sub in Perl 6, worked like a charm on the first try.
:D
of course, the does Iterable crash seems worrisome. 15:39
masak colomon++ # nice local sub :)
colomon attempting to get grep to work, but having difficulties. time for lunch, alas. 15:40
finanalyst rakudo: role A {has $.a; has $.b}; my @x of A; for ^3 { @x[$_] .= new(:a($_*2), :b($_+3))}; for @x {.perl.say 15:43
p6eval rakudo 5ab1c0: OUTPUT«Unable to parse blockoid, couldn't find final '}' at line 11␤current instr.: 'perl6;Regex;Cursor;FAILGOAL' pc 1664 (ext/nqp-rx/src/stage0/Regex-s0.pir:907)␤»
finanalyst rakudo: role A {has $.a; has $.b}; my @x of A; for ^3 { @x[$_] .= new(:a($_ * 2), :b($_ + 3))}; for @x {.perl.say} 15:44
p6eval rakudo 5ab1c0: OUTPUT«get_pmc_keyed() not implemented in class ''␤current instr.: 'perl6;Positional[::T];postcircumfix:<[ ]>' pc 10553 (src/metamodel/GrammarHOW.pir:44)␤»
finanalyst rakudo: role A {has $.a; has $.b}; my @x of A; for ^3 { @x[$_] .= new(:a($_ * 2), :b($_ + 3))}; for @x {.perl.say}
p6eval rakudo 5ab1c0: OUTPUT«get_pmc_keyed() not implemented in class ''␤current instr.: 'perl6;Positional[::T];postcircumfix:<[ ]>' pc 10553 (src/metamodel/GrammarHOW.pir:44)␤»
finanalyst rakudo: role A {has $.a; has $.b}; my @x of A; for ^3 { @x[$_] .= new(:a($_ * 2), :b($_ + 3))}; for @x {$_.b.say} 15:47
p6eval rakudo 5ab1c0: OUTPUT«get_pmc_keyed() not implemented in class ''␤current instr.: 'perl6;Positional[::T];postcircumfix:<[ ]>' pc 10553 (src/metamodel/GrammarHOW.pir:44)␤»
finanalyst rakudo: role A {has $.a; has $.b}; my @x; for ^3 { @x[$_] .= A.new(:a($_ * 2), :b($_ + 3))}; for @x {$_.b.say}
p6eval rakudo 5ab1c0: OUTPUT«Unable to parse blockoid, couldn't find final '}' at line 11␤current instr.: 'perl6;Regex;Cursor;FAILGOAL' pc 1664 (ext/nqp-rx/src/stage0/Regex-s0.pir:907)␤» 15:48
15:48 _jaldhar_ joined, jaldhar_ left
finanalyst role A {has $.a; has $.b}; my @x; for ^3 { @x.push: A.new(:a($_ * 2), :b($_ + 3)) }; for @x {$_.b.say} 15:49
rakudo: role A {has $.a; has $.b}; my @x; for ^3 { @x.push: A.new(:a($_ * 2), :b($_ + 3)) }; for @x {$_.b.say} 15:50
p6eval rakudo 5ab1c0: OUTPUT«3␤4␤5␤»
15:50 justatheory left
finanalyst Hi anyone. Is there a neater way to instantiate an array of objects other than pushing? 15:52
masak rakudo: my @x = 1, 2, 3; say @x.perl 15:53
p6eval rakudo 5ab1c0: OUTPUT«[1, 2, 3]␤»
15:53 diakopte1 is now known as diakopter
slavik rakudo: my @arr = "blah" ** 5; say @arr.perl 15:54
15:54 baisong left
masak rakudo: class A {}; my @x = A.new, A.new, A.new; say @x.perl 15:54
p6eval rakudo 5ab1c0: OUTPUT«[0]␤»
rakudo 5ab1c0: OUTPUT«[A.new(), A.new(), A.new()]␤»
slavik rakudo: my @arr = "blah" x 5; say @arr.perl
p6eval rakudo 5ab1c0: OUTPUT«["blahblahblahblahblah"]␤»
slavik that way
finanalyst slavik: see my previous code. Suppose the objects have typed structure 15:55
15:55 sorear left 15:57 Kyril joined
finanalyst rakudo: role A {has $.a; has $.b}; my @x of A; for ^4 { @x[$_] .= new(:a($_ *3), :b($_ + 2)}; say @x.perl 15:57
p6eval rakudo 5ab1c0: OUTPUT«Unable to parse blockoid, couldn't find final '}' at line 11␤current instr.: 'perl6;Regex;Cursor;FAILGOAL' pc 1664 (ext/nqp-rx/src/stage0/Regex-s0.pir:907)␤»
finanalyst rakudo: role A {has $.a; has $.b}; my @x of A; for ^4 { @x[$_] .= new(:a($_ *3), :b($_ + 2))}; say @x.perl
p6eval rakudo 5ab1c0: OUTPUT«get_pmc_keyed() not implemented in class ''␤current instr.: 'perl6;Positional[::T];postcircumfix:<[ ]>' pc 10553 (src/metamodel/GrammarHOW.pir:44)␤»
diakopter g'morn #perl6 15:58
15:58 eternaleye left
masak diakopter: \i 16:00
diakopter p/
masak :-w
*shudder* -- I think I just got a glimpse into what combination of circumstances cause this bug to happen. 16:01
diakopter do tell
masak and why no-one has discovered it until now. :(
diakopter your theory
masak my current theory involves references to lexical subs, anonymous subs, reentrancy, and deep nesting.
16:01 sorear joined
masak I'm minimizing very carefully, but the thing I started from is GGE, so it will probably take all evening or more. 16:02
diakopter ah
diakopter dealt a lot with those things last night perlesque
in perlesque
masak yeah, I saw.
heady stuff :)
diakopter is it reproducible 16:05
with a particular pattern in your gge
masak it is.
want me to push the debug branch?
diakopter what's the pattern that manifests it 16:07
or is it every pattern
masak a+ ** b
diakopter remind me what ** means plz
masak in this case, it repeats 'a+' with 'b' as a separator. 16:08
but that doesn't matter; the error occurs in the parsing of the regex.
to parse 'b', the regex parser has to call itself.
that seems to be essential to the bug.
diakopter how does it call itself 16:10
dalek kudo: 842d2b0 | pmichaud++ | docs/spectest-progress.csv:
spectest-progress.csv update: 391 files, 30643 (79.4% of 38606) pass, 2 fail

S06-multi/syntax.rakudo 21 - use multi method to override builtin lives S06-multi/syntax.rakudo 22 - use multi method to override builtin
masak ok, so the optable parser works through callbacks, right?
diakopter misses phenny b/c I want to msg pmurias
masak for example, a sub parse_quantifier that can parse '**' and such.
now, parse_quantifier needs to know where 'b' ends, because it might be a parenthesized expression or something. 16:11
so it simply calls the whole regex expression parser with a caveat to stop parsing at the term: level. 16:12
diakopter oh
masak so, when we're parsing 'b', we're in the p6regex parser calling the optable parser calling parse_quantifier calling the p6regex parser calling the optable parser calling parse_term.
masak exhales
s/ex/in/ 16:13
and somewhere there, an array that should have contained one element turns out to be empty.
diakopter any chance it's a parrotbug? 16:14
masak and this under mysterious circumstances, such as it containing the element before a call, and being empty right after the call.
diakopter: dunno, might be.
16:14 takadonet joined
takadonet morning all 16:15
masak hm, explaining this might actually help me write a minimal example from scratch :)
takadonet: \o
takadonet I see commit are happening!
takadonet does a git pull
masak: what editor do you use for creating Perl6 code? 16:19
16:19 rv2733 left
masak takadonet: vim 7.2 16:19
16:19 IllvilJa joined
masak and I write most prose in Emacs. :) 16:19
takadonet <--- emacser here 16:20
diakopter <--- Notepad++
takadonet I been installing Padre so at least I will have syntax highlighting 16:21
masak well, I have it too; it's just that it breaks sometimes. 16:22
diakopter referring to www.cs.chalmers.se/~boquist/phd/ does anyone know what he means by "forcing of closures"? 16:28
masak something that causes the call target to be unknown at compile time? :) 16:30
wouldn't that be more or less all of OO method calls out there?
all polymorphic dispatch, at least. 16:31
16:36 masak left 16:37 justatheory joined
diakopter jql: howdy 16:38
16:49 lichtkind joined
lichtkind cheers 16:51
diakopter hi 16:52
16:52 Guest61693 is now known as Trashlord, Trashlord left, Trashlord joined 16:55 levengli joined
levengli is there a library of pm for perl6 in the dist/ 17:01
17:05 envi^home left, levengli left
diakopter I admit: I don't know what dist/ levengli was asking about 17:09
17:18 hercynium joined 17:21 wknight8111 joined 17:23 Chillance joined
jnthn looks over the NPW email and is highly amused by the restaurant description "Prikið (burgery-fusion, greasygood, hangover killer meal incl. painkillers)" 17:30
Including painkillers?!
:-)
colomon jnthn: why does class Tree is Iterable blow up my code? :) 17:32
(or rather, masak++'s code that I have copied) 17:33
jnthn 'cus Iterable in Rakudo is a class last I checked.
17:34 lichtkind_ joined
colomon errr... is "is Iterable" not the proper syntax for class derivation? 17:34
17:34 lichtkind left, lichtkind_ is now known as lichtkind
colomon www.harmonyware.com/pictures/big.png 17:37
1001 x 1001 completed while we were at lunch. 17:38
took 6.4 GB of memory to compute. 17:39
17:39 rv2733 joined
wknight8111 I've learned a lesson today: If I have only 10 minutes to hack, I can't even run a failing Rakudo test case 17:57
much less figure out why it fails
18:06 Alias left, Alias joined
jnthn colomon: Well, it's certainly pretty. :-) 18:06
colomon FK 18:08
afk
back. 18:14
I'm quite pleased by the picture, actually. :) 18:15
with Tree $t: ($t.iterator Z~ 'a'...*).eager.perl.say; 18:21
("H2Oa", "Hb", "Oc", "Od")
:)
If I could just make Tree Iterable, that would be cleaner... ;) 18:22
18:23 meppl joined 18:28 iblechbot joined 18:30 hercynium left 18:34 _jaldhar_ left 18:35 _jaldhar_ joined 18:42 _jaldhar_ left 18:43 _jaldhar_ joined 18:46 Kyril left
slavik colomon: mandebrot ... perl6? 18:49
finanalyst alpha: my @a = 1,2,3,4; for @a[ 1 .. *-1 ] { .say } 18:50
p6eval alpha 30e0ed: OUTPUT«2␤3␤4␤»
finanalyst rakudo: my @a = 1,2,3,4; for @a[ 1 .. *-1 ] { .say } 18:51
p6eval rakudo 842d2b: ( no output )
finanalyst jnthn: is it known that rakudo has stopped liking * in arrays? 18:52
arnsholt Regression from alpha, IIRC 18:54
19:00 [synth] joined 19:02 synth left
colomon slavik: mandelbrot, in perl 6: github.com/colomon/mandelbrot 19:03
It is slower than a very slow thing, but it works and is stable. 19:04
slavik colomon: nice :)
colomon (errr, the color version works, the b&w version probably still needs a tiny bit of work to be ported to the current Rakudo.)
slavik err
it's like 20 lines of code ... 19:05
.map is paralell?
colomon slavik: no.
slavik :(
colomon perl6advent.wordpress.com/2009/12/1...g-snowmen/
slavik colomon: use opengl to draw it? 19:06
colomon that post describes the code, and also the changes needed to make it parallel when Rakudo supports that.
nope, it just dumps a PPM file to standard output.
slavik aww
colomon Hmmm, actually >>☃ might work now. I probably need to revisit the code more carefully at some point. But for this weekend, it's been a great timing / stability / memory usage test. 19:08
slavik colomon: cool stuff :) 19:09
colomon :) 19:10
19:22 [synth] is now known as synth 19:32 finanalyst left 19:37 ejs joined
sorear colomon: 6 GB? Why? Memory leak? 19:46
colomon sorear: either that or Parrot / Rakudo is immensely piggy. 19:47
That was actually what I predicted it would take after my tests yesterday, the script seems to need 4MB additional memory every time you increase the edge length of the fractal by 1. 19:48
sorear last time I studied the mandelbrot set, it could be computed in constant space
my favorite definition of a memory leak is "any bug which increases asymptotic space complexity" 19:49
(never-freed memory was nice until they invented GC)
19:50 nihiliad joined 19:53 molaf left 19:57 pmurias joined
pmurias diakopter: ping 19:57
diakopter: how do i submit failing test cases for perlesque? can i get a commit bit? ;) 19:58
20:16 eternaleye joined
colomon sorear: yes, constant space is right. for this algorithm, each pixel is buffered until a line is done, and then the entire thing is dumped. its memory usage should be minimal. definitely should not take 6GB. 20:29
20:32 justatheory left 20:49 _buno_ joined 20:54 _buno_ left 20:55 _buno_ joined 20:56 _buno_ left
lichtkind mberends: hai 20:56
mberends hai lichtkind 20:57
lichtkind mberends: howsetging meen
mberends lichtkind: it's going on Windows today, spelunking the mysterious corridors of the Win32::PerfLib class 20:59
21:00 ReiniUrban joined
lichtkind hm, my new system runs well, im sort of back trying go over 500 today 21:00
21:00 wknight8111 left 21:01 rurban left, ReiniUrban is now known as rurban 21:04 rv2733 left
lichtkind whats the differens between the bit and bool type? 21:06
21:11 buno_ joined
mberends probably bit is 1 or 0, bool is true or false (this is splitting hairs). numeric versus logical. 21:12
lichtkind i suppose your right 21:13
and that some conversion might also be easy
Juerd Which right? :)
lichtkind: s/your/you're/
lichtkind Juerd: right as in Bool::True
mberends Juerd: my right ;) 21:14
Juerd I hope you have multiple. :) 21:16
21:16 buno_ left 21:17 _buno_ joined
lichtkind Juerd: i know it you're im just lazy now 21:17
mberends lichtkind: s/it/it's/ s/im/I'm/ ;) 21:18
Juerd s/<after you're>/;/
s/$/./ # nit
colomon I believe Bool is numerical in theory.
say True 21:19
rakudo: say True # works better with rakudo:
p6eval rakudo 842d2b: OUTPUT«1␤»
colomon "numerical", ugh, should be "numeric".
21:25 _buno_ left 21:26 _buno_ joined, _buno_ left
IllvilJa Is 'gitpan' the future CPAN? 21:29
mberends no, CPAN will continue in the future, with additions for Perl 6 21:30
IllvilJa ok, so CPAN for released packages and gitpan for development snapshots of them? 21:31
21:31 _buno_ joined
mberends yes. predictions are hard though, particularly those about the future ;) 21:32
21:32 _buno_ left
IllvilJa Yes. Predictions are overrated, wild speculations will do fine :-D. 21:33
21:37 PacoLinux left, _buno_ joined 21:38 iblechbot left, _buno_ left 21:39 _buno_ joined 21:50 hercynium joined 21:54 _buno_ left, _buno_ joined
diakopter pmurias: hi :) 21:54
pmurias: yeah; what's your google id
21:56 _buno_ left, _buno_ joined
sorear mberends: you beleive in CPAN6? 21:57
mberends sorear: I'll believe it when I see it.
diakopter finally today finished the SF heptary Saga of Seven Suns
21:58 justatheory joined
sorear ok, we're in agreement there. 21:58
diakopter CPAN is plenty fine good 21:59
sorear IllvilJa: if you ask me, the future lies with metadata-only systems like proto and plumage 22:01
diakopter sorear: I'm staring blankly at the disassembly of ActiveState's perl-clr bridge. 22:02
sorear ooooh 22:04
someone has a reverse engineering project for me?
22:09 cosimo left 22:12 _buno_ left 22:13 _buno_ joined
lue why hello there 22:15
mberends welcome to our century 22:16
diakopter it's a dark one.
lue 21st, right?
the DW premiere was awesome. Epic title sequence, epic remodel, and awesome h2g2 reference :) 22:19
diakopter pmurias: you around? 22:20
pmurias diakopter: yes 22:21
diakopter pmurias: should I add your google email address? 22:23
mberends discovers he has been asleep for a few minutes. Goodnight o/ 22:24
pmurias [email@hidden.address]
that's my google id right?
diakopter I guess :) 22:25
mberends: :)
pmurias diakopter: does perlesque have a test suit?
diakopter no
it's only recently attained verifiable input/output status 22:26
external interaction, I mean
as opposed to hosting the tests itself
which I hate
pmurias hosting the tests itself?
that means hardcoding ok(...)? 22:27
diakopter I mean, internal tests
I think all the tests should be "black box" 22:28
to spec
what's your failing test 22:29
it's probably just NYI
pmurias reproducing...
sub int foo(int $arg0) {return 4;} fails to parse with a bizare error message 22:30
diakopter: the parser is hand coded recursive descent? 22:31
diakopter no
heh
LOL, in fact.
you obviously haven't looked at the code :P
it's got its own parser generator (which does generate a recursive descent parser, yes) 22:32
std: sub int foo(int $arg0) {return 4;} 22:33
p6eval std 30404: OUTPUT«===SORRY!===␤Malformed block at /tmp/dN6EcgZRmy line 1:␤------> sub int ⏏foo(int $arg0) {return 4;}␤ expecting any of:␤ block␤ routine_def␤ trait␤Parse failed␤FAILED 00:01 107m␤»
pmurias diakopter: the problem is with the lack of the whitespace after foo
diakopter I know, but I'm curious why std doesn't like it either
std: my $a = sub Int foo (int $arg0) {return 4;} 22:34
p6eval std 30404: OUTPUT«===SORRY!===␤Malformed block at /tmp/srPya0uZAn line 1:␤------> my $a = sub Int ⏏foo (int $arg0) {return 4;}␤ expecting any of:␤ block␤ routine_def␤ trait␤Parse failed␤FAILED 00:01 108m␤»
pmurias std: sub foo(int $arg0) {return 4;}
p6eval std 30404: OUTPUT«ok 00:01 109m␤»
pmurias std: int sub foo(int $arg0) {return 4;}
p6eval std 30404: OUTPUT«===SORRY!===␤Two terms in a row at /tmp/w6p6WacmR2 line 1:␤------> int ⏏sub foo(int $arg0) {return 4;}␤ expecting any of:␤ bracketed infix␤ infix or meta-infix␤ statement modifier loop␤Parse failed␤FAILED 00:01 110m␤»
pmurias std: sub int foo(int $arg0) {return 4;}
p6eval std 30404: OUTPUT«===SORRY!===␤Malformed block at /tmp/1T5ySTq2vW line 1:␤------> sub int ⏏foo(int $arg0) {return 4;}␤ expecting any of:␤ block␤ routine_def␤ trait␤Parse failed␤FAILED 00:01 107m␤»
diakopter TimToady: consider yourself summoned
ok; it's a bug in the grammar, yeah 22:35
I thought return type went between 'sub' and subname
diakopter goes to look at the Synopsis again 22:36
22:36 clintongormley left
pmurias diakopter: why is a sub name require in a sub declaration? 22:37
s/declaration/literal/
diakopter b/c anonymous ones are still NYI
:P
22:37 _buno_ left
pmurias named ones seem a bit unnessesary 22:39
good night 22:42
lue gnight o/
diakopter o /
22:42 pmurias left
lue afk 22:46
22:48 cosimo joined 22:53 mberends_ joined, mberends left 23:00 Kyril joined 23:09 orafu left, orafu joined 23:21 astrojp joined
diakopter got anonymous subs working. 23:21
heh
colomon std: my int sub foo(int $arg0) {return 4;} 23:22
p6eval std 30404: OUTPUT«ok 00:01 111m␤»
diakopter ahhh....
23:22 wknight8111 joined
colomon std: our int sub foo(int $arg0) {return 4;} 23:23
diakopter std: my $a = my int sub foo(int $arg0) {return 4;}
p6eval std 30404: OUTPUT«ok 00:01 109m␤»
std 30404: OUTPUT«ok 00:01 110m␤»
diakopter std: my $a = my int sub (int $arg0) {return 4;}
p6eval std 30404: OUTPUT«ok 00:01 110m␤»
diakopter std: my $a = my int sub(int $arg0) {return 4;}
p6eval std 30404: OUTPUT«===SORRY!===␤Malformed my at /tmp/B6ZEE25DRq line 1:␤------> my $a = my int ⏏sub(int $arg0) {return 4;}␤ expecting any of:␤ declarator␤ multi_declarator␤ routine_declarator␤ typename␤Parse failed␤FAILED 00:01 108m␤»
diakopter whew
perlesque: my $a = sub (int $b) { say($b); say($b); return 1; }; $a(9); say(10); 23:24
p6eval perlesque: OUTPUT«9␤9␤10␤»
diakopter o_O
I, um, um, 23:25
it's supposed to require the sub return type
wait
23:25 Grrrr joined
diakopter perlesque: my $a = sub (int $b) { say($b); say($b); }; $a(9); say(10); 23:25
p6eval perlesque: OUTPUT«9␤9␤»
diakopter O_O
weird.
perlesque: my $a = sub (int $b) { say($b); say($b); }; $a(9); say(10); 23:26
p6eval perlesque: OUTPUT«9␤9␤»
diakopter strange.
ooooooo I have an idea
oh, interesting; it was infinitely looping b/c it never returned. 23:29
will fi l8r
fix l8r
23:30 p6eval left, p6eval joined
diakopter perlesque: my $last = 4; my $sub_a = sub int foo (int $arg0) { say(my $b = $arg0 + 5); $last = $b; return $b; }; my $sub_b = $sub_a; $sub_b(4); $sub_b = sub int bar (int $arg0) { say(my $b = $arg0 + 4); if ($b>19) { $sub_b = $sub_a; }; $last = $b; return $b; }; while ($last < 100) { $sub_b($last); }; 23:30
p6eval perlesque: OUTPUT«9␤13␤17␤21␤26␤31␤36␤41␤46␤51␤56␤61␤66␤71␤76␤81␤86␤91␤96␤101␤»
diakopter perlesquel: runlast
p6eval perlesquel: OUTPUT«9␤13␤17␤21␤26␤31␤36␤41␤46␤51␤56␤61␤66␤71␤76␤81␤86␤91␤96␤101␤»
diakopter hee
perlesquel: runlast 23:31
p6eval perlesquel: OUTPUT«9␤13␤17␤21␤26␤31␤36␤41␤46␤51␤56␤61␤66␤71␤76␤81␤86␤91␤96␤101␤»
diakopter there you see the difference two compilation stages makes.
lichtkind good night your brave ones
23:33 lichtkind left
sorear what's perlesquel? 23:33
diakopter :D it just runs the last image perlesque compiled from p6eval input
perlesque: loop (my $x=0;$x<100000000;$x+=1) { }; say(939391); 23:35
p6eval perlesque: OUTPUT«939391␤»
diakopter perlesquel: runlast
p6eval perlesquel: OUTPUT«939391␤»
diakopter mmmmm raw speed
23:46 p6eval left, p6eval joined
diakopter perlesque: loop (my $x=0;$x<100000000;$x+=1) { }; say(939391); 23:47
p6eval perlesque: OUTPUT«939391␤»
diakopter perlesquel: runlast
p6eval perlesquel: OUTPUT«939391␤»
diakopter hm
23:54 p6eval left, p6eval joined
diakopter perlesquel: runlast 23:55
:/
p6eval perlesquel: OUTPUT«939391␤real 0.39␤user 0.39␤sys 0.00␤»
diakopter perlesquel: runlast
p6eval perlesquel: OUTPUT«939391␤real 0.40␤user 0.37␤sys 0.02␤»
diakopter oh ok
pugssvn r30405 | diakopter++ | [evalbot] add perlesquel pseudo-backend (it times the execution and displays the output of the last perlesque p6eval input 23:58
diakopter )
perlesquel: runlast 23:59
p6eval perlesquel: OUTPUT«4␤real 0.05␤user 0.04␤sys 0.00␤»

Powered by Cro and the Raku® Programming Language.

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

Thank you!