Parrot 3.8.0 "Magrathea" | parrot.org | Log: irclog.perlgeek.de/parrot | #parrotsketch meeting Tuesday 19:30 UTC
Set by moderator on 29 September 2011.
dalek kudo/nom: dc9ffc4 | jnthn++ | src/ops/perl6.ops:
Avoid spending so much time in realloc. Shaves another ~12% off 'for 1..1000000 { $i++ }'.
 00:19
00:37 whiteknight joined
whiteknight good evening, #parrot 00:39
cotto hi whiteknight 00:41
sorear hello whiteknight.
whiteknight hello cotto, sorear. How are you two doing tonight?
sorear I'm good 00:43
whiteknight sucks about Steve Jobs
I'm hardly an apple fanboi 00:44
00:44 schmooster joined
cotto heads home 01:01
01:38 benabik joined
benabik ~~ 01:38
cotto ~~ 02:01
benabik I find it amusing the the Planet Parrot RSS shortens my entire blog post to the TL;DR paragraph. 02:16
benabik needs to teach Vim that .md means markdown, not modula2 02:34
cotto you may think that now, but wait until you really need to edit some modula2 code 02:35
benabik … right
cotto happens all the time 02:37
dalek CT: 340008c | benabik++ | TODO.md:
TODO: Add notes about PACT languages

Having a language at each level will be instructive and useful. Also add to documentation the idea that PACT assembly is intended to be also implemented in C and used in favor of PIR.		 02:56
cotto 0 stories on hacker news about not Steve Jobs 03:03
benabik Yeah, that'll happen.
cotto not terribly surprising 03:04
dalek CT: 761bdf6 | benabik++ | nodes.md:
Remove equivocation about a layer below POST

A CFG based model is nearly a requirement for nearly any reasonably complex code analysis. Make there be One True CFG model so the wheel doesn't have to be reinvented any time someone wants to write complex optimizations. Also, ideally we'll be implementing complex optimizations so let's not start by shooting ourselves in the foot.		 03:16
CT: 0d50888 | benabik++ | stages.md:
Additional thoughts on stages
benabik I want to keep tweaking those docs, but I'm not sure how much more I can do without either feedback from others or starting to implement something. And tomorrow I should probably get started on homework and studying for my midterm. 03:33
And right now, I should sleep. 'night
cotto 'night 03:56
also, don't wait too long for feedback
plobsing cotto: I've been thinking about how dis and m0 are similar. If they're similar, is there a reason why m0 is not dis? There are other systems out there with strong similarities to m0 (as I understand it). Why not any of them? 04:26
for example, I just read about the Mite VM (rrt.sc3d.org/Software/Mite/), whose stated goals are nearly identical to m0. 04:28
04:31 nbrown joined 04:38 silug joined
cotto plobsing, that's be great. I'd love to not have us write and debug M0 if it's not necessary. 04:39
plobsing++ 04:40
plobsing I am nothing if not lazy. 04:45
05:17 alvis joined 05:18 rfw joined 05:45 SHODAN joined 06:15 l3l1p joined 07:30 baest joined 07:53 baest joined 08:04 lucian joined
dalek kudo/convergent-num-rat: 0e351b2 | moritz++ | src/core/Num.pm:
try not to loop infinitely in Num.Rat
 08:55
cotto plobsing, I'm curious how you found mite. It looks like its build depends on a tool that's no longer available ("nuweb"), but I'll plan on plundering it for ideas. 09:08
09:11 ambs joined 10:01 preflex_ joined 10:59 mj41 joined 11:37 Psyche^ joined 11:52 benabik joined
benabik o/ #parrot 11:53
plobsing cotto: I found mite while going deep into the google results for dis. The guy's thesis mentions dis a few times. 12:02
12:05 whiteknight joined 12:12 gbacon joined
plobsing cotto: as for the nuweb problem, a sourceforge project by the same name seems to have activity within the last year. 12:15
12:16 bluescreen joined 13:18 whiteknight joined
whiteknight good morning, #parrot 13:18
benabik o/ whiteknight
whiteknight Here's an interesting link about PHP: seldo.com/weblog/2011/08/11/php_nee...replace_it
hello benabik 13:20
13:27 bluescreen joined
dalek kudo/eval-throws: 817dcc2 | moritz++ | src/core/Exception.pm:
make Exception False for now
 14:02
kudo/eval-throws: f56944f | moritz++ | docs/ChangeLog:
update ChangeLog
14:49 lucian joined 14:56 bluescreen joined 15:06 dmalcolm joined 15:10 mtk joined
dalek kudo/nom: 1c2c2d4 | moritz++ | src/Perl6/Actions.pm:
do not complain about double declared anon variables
 15:13
16:45 l3l1p joined 16:51 mj41 joined 16:56 dodathome joined 17:05 fperrad joined 17:10 pyrimidine joined 17:11 pyrimidine joined 17:34 davidfetter joined, contingencyplan joined
cotto ~~ 18:10
plobsing: any luck getting it to build?
nine Why does the headerizer not pick up my new function? 18:15
whiteknight nine: have you made it angry?
nine: it might not recognize the formatting. It's pretty particular
cotto headerizer is quite picky
nine Ah, now it worked 18:18
cotto it's fun like that
whiteknight: I wasn't impressed by that article. Anyone who says anything good about pear hasn't seen cpan. 18:32
It's very interesting to think about what's made PHP so popular and how to work with those forces to produce and popularize a superior replacement. 18:34
whiteknight cotto: it's not really in competition with cpan because they're two different languages 18:35
cotto whiteknight: sure
whiteknight if cpan wasn't enough to keep the web from moving from perl to php, the relative merits of pear vs cpan don't factor in at all
cotto whiteknight: exactly
same for language elegance, modularity, consistency, etc 18:36
whiteknight nowhere in there does he even say that php is better than perl, only that it was more suited to the problem domain
for some features like unicode, PHP gives up and says it's impossible. Perl says that it's perfectly possible, but extremely hard to do right and you're better off not bothering 18:37
atrodo I like how he says that perl's features made web development insecure 18:38
whiteknight there is a certain amount of unnecessary perl bashing in the article, sure.
but there is clearly more php bashing, and a few other random drive-by targets 18:39
nine I found him too fixed on the language having to support his problem domain perfectly. That's perfectly possible using libraries and frameworks. Having them baked into the language itself does not improve things. 18:42
whiteknight the underlying point is a good one though: There aren't any good, modern programming languages that have a fundamental focus on the domain of website creation
PHP certainly targets a particular niche, but arguably PHP is a generation or two behind other modern dynamic programming languages 18:43
cotto not a hard argument
whiteknight something with PHP's focus but a feature set more similar to python, ruby, or scala would be a compelling choice for many
nine Who'd have thought? preempting a task actually seems to be easier than a task giving up on it's own free will (think sleep op). In the latter case, I can't just restrict it to the topmost runloop. 19:01
whiteknight I believe it 19:12
giving up means exiting runops and needing to perform a bunch of bookkeeping 19:13
nine no problem at all in the outermost runloop...but as soon as it's an inner one, I'm back to square one again 19:15
whiteknight help us O'M0 kenobi, you're our only hope 19:18
nested runloops are such a big goddamned problem. I really wish they had been avoided from the start 19:21
nine This whole thing would have been almost trivial if not for them. But they are a fact right now 19:23
whiteknight yeah 19:25
nine The only workaround that comes to my mind is to forget concurrency and really sleep on a sleep op in a nested runloop 19:34
Sucks and cannot possibly be a persisting solution. But may do for now to at least allow the 90% we have to be used, leaving the real solution for later (probably when M0 arrives) 19:35
whiteknight yeah, if we have to do that, do that 19:47
what's most important for us now is to get a basic idea of the kinds of behaviors we want, and get a decent enough API so we know what behavior to preserve and reproduce when we switch out the internals for something better
In short, what we most need are good ideas 19:48
19:49 bluescreen joined
nine And we are more likely to get those when people start using concurrency features and run into problems 19:53
Well, I'm off to bed for now. Good night. 19:54
tadzik o/
21:00 soh_cah_toa joined 21:01 RobertLJ joined 21:18 soh_cah_toa_ joined 21:20 l3l1p joined 21:49 Coke joined 21:58 RobertLJ left
benabik Do Parrot strings share buffers? i.e. if I have a very large string and do `$S1 = substr $S0, 128, 256` does that $S1 allocate memory or just reference $S0? 22:23
soh_cah_toa_ that's a good question... 22:31
22:33 alvis joined
benabik If not, I wonder if that's a big chunk of the memory usage of something like Rakudo's setting, since PCT nodes store the snippet of source that created them. 22:35
benabik is back to wanting to implement Ropes.
soh_cah_toa_ benabik: Ropes?
benabik aloha: ropes? 22:36
aloha benabik: I give up.
benabik aloha: ropes are a data structure to store and edit large strings.
aloha benabik: Okay.
benabik aloha: ropes are also en.wikipedia.org/wiki/Rope_(computer_science)
aloha benabik: Okay.
benabik soh_cah_toa_: ^^
cotto was it you that said that every implementation of ropes is broken?
benabik I don't think so… I find them a very interesting data structure. 22:37
soh_cah_toa_ indeed
cotto I'm not criticizing the idea. It looks like fun. 22:41
soh_cah_toa_ error:imcc:syntax error, unexpected DOT ('.') 23:07
in file '(file unknown)' line 254744173
sigh...
cotto I laugh and I cry.
soh_cah_toa_ has no energy left to hate imcc with 23:08
dalek CT: 433a6ea | benabik++ | compiling.md:
Compiling: Errors, Warnings, and multiple outputs

One of the more frustrating things about PCT as a HLL developer was always the extra verbose and non-informative messages that occurred because an exception was thrown. So include error and warning handling in the system from step one.
Also, a single linear set of stages doesn't actually reflect how a
  compiler works, so add some thoughts on the matter.		 23:18
benabik hopes someone is finding this interesting. 23:22
23:26 aloha joined 23:30 not_gerd joined
not_gerd hello, #parrot 23:30
cotto, plobsing: just took a look at Mite, but I don't think it's a good fit for parrot 23:32
in particular, I don't like that arithmetics are done with pointer-size precision and that it's stack-based (which is bad for CPS)
cotto not_gerd: stack-based is a killer 23:34
register VMs aren't all that common 23:36
not_gerd cotto: one can work around that by elevating values from stack to heap as needed (afaik the Lua VM does that when you take a closure), but it's definitely sub-optimal if you want to base your control flow on CPS 23:37
23:44 aloha 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!