'-Ofun: xrl.us/hxhk | pugscode.org | pugs.kwiki.org | paste: paste.lisp.org/new/perl6 or sial.org/pbot/perl6'
Set by rafl_ on 5 November 2005.
leo rafl: maybe folks should run nice 'the work' 00:07
autrijus sync.pl (and darcs) is now running in +20 00:08
hope that helps 00:09
leo I don't really care how long my 'nice make all test' is running, when I'm doing that, but I don't like latency in my screen/irssi session ;-) 00:11
anyway feather is really great - Juerd++ 00:12
rafl Indeed!
Juerd :)
leo: Do you usually have latency?
'cause if people nice their stuff correctly, there shouldn't be 00:13
leo no - no problem
Juerd ok
leo I'm compiling on another machine in hawaii regularly - I know what latency is ;-) 00:14
Juerd Hehe
leo but with that machine I have svn up;make realclean;perl Configure.pl;make test in ~4 mins 00:15
Juerd That's fast
leo it is
Juerd Can't you set up distcc on it? ;)
rafl We shoud setup a global ccache on feather, btw. 00:16
leo it has ccache alreay
Juerd rafl: There is
leo but distcc could work too
Juerd Something ccache like for ghc would be sweet.
eric256 juerd...any ideas why the feather build seems to be failing? the .err file is empty. ;(
leo it's a dual opteron 250 machine 00:17
eric256 trying a manual build now
rafl Juerd: a global one?
SamB Juerd: ccache?
leo nope
compiler cache
rafl Juerd: Could be possible. At least for the work ghc does using cc.
Juerd rafl: Oh, no, it had that, but things broke randomly and I lacked fixing fu
eric256: No idea.
rafl Juerd: I also tried it and it didn't work out. Thought I was to dumb for it. 00:18
Juerd eric256: That's why I wanted someone else to do this :)
rafl: Anyway, ccache per user is a big save already
rafl Juerd: Yes. A global one would be even better because most people are compiling the same stuff there, afaik. 00:19
Juerd Yes, but as said, that fails misteriously
mysteriously.
eric256 juerd. no prob, looking into it ;) 00:20
svnbot6 r7865 | autrijus++ | * repo cp Pugs.Compile to Pugs.Compile.PIL2 to begin the new compilation saga. 00:22
leo Juerd: a global ccache is imho not the best idea 00:23
rafl leo: Why?
leo tell folks how to use it locally
I don't think it scales and in trouble case you can't clean it
Juerd leo: IIRC, it's used by default 00:24
leo indeed 00:25
Juerd Then telling people isn't necessary :)
leo ok - as long as it works
Juerd Haven't had complaints
Well, since the global cache was removed, anyway
leo I had some nasty compile errors that went away after clearing the cache, not often, but ... 00:26
Juerd You just bumped the counter to 1
I'll take action when it's at 5 :)
leo ah - I see - the .ccache is local - fine then
autrijus hm, I'd like to change -CPerl5 to -CPIL-Perl5. 00:33
svnbot6 r7866 | autrijus++ | * New compiler backend targets around PIL2:
r7866 | autrijus++ | ./pugs -C PIL2-Perl5
r7866 | autrijus++ | ./pugs -C PIL2-JSON
r7866 | autrijus++ | ./pugs -C PIL2-Binary
r7866 | autrijus++ | The dash is optional (all nonalphanum is dropped), and
r7866 | autrijus++ | it's case insensitive. Eventually these will replace
r7866 | autrijus++ | the PIL1 variants when our p5/js codegens finishes the
r7866 | autrijus++ | migration.
autrijus reserving -CPerl5 for the true "compile to perl5", not "compile to pil, dump as perl5"
Juerd autrijus: Do you know if there's anything that lets ghc cache things? 00:34
autrijus: So compilation doesn't happen again for unchanged files?
autrijus Juerd: GHC should automatically do that. 00:35
it's the linking that is mandatory
Juerd Oh
But this doesn't work across make cleans, right?
Ah well, it's no big deal anyway 00:36
feather's fine as it is
eric256: One piece of advice: when running autobuild.sh manually, disable the cron thingy. Two concurrent instances will mess things up. 00:37
eric256 how do i disable the cron? comment it out? 00:39
autrijus Juerd: well, we can spec so that clean doesn't clean dist/build 00:40
in Makefile.PL
in which case it will survive clean
Juerd eric256: Yes
rafl Pugs.cabal contains /usr/lib/blib/lib for me now which seems to be wrong. Does anyone know which changes introduced that?
Juerd autrijus: Then what does it clean? I think this would be counterintuitive 00:41
autrijus doesn't say that here
Juerd: k
Juerd My dillemma is that for an automated building thing, you'd want make clean, but not to recompile.
But cleaning up does improve the success ratio a lot :)
ccache fixes this for C
rafl autrijus: On feather? 00:42
eric256 could make clean on the hour, but between hours just regular make ?
Juerd eric256: Oh, if a build fails, then from then on, it will not build at all (and thus keep err empty) until there's a new svk revision.
eric256: When testing, comment out the condition
rafl eric256: Could you please stop your autobuild job? I'd like to try how it works for the debian packages.
eric256 yea i'm not testing the script right now, just trying to get it to build at all 00:43
rafl Thank you.
eric256 um ralf the build isn't working at all yet so you might want to make
make clean; perl Makefile.PL; make unoptimized; doesn't seem to work. trying a build right now 00:44
theorbtwo is back.
rafl eric256: I'm rafl, btw. Not ralf. :-) 00:45
eric256 hehe sorry. daughter started chocking...on nothing apparently. fun how they do that. ;) 00:46
svnbot6 r7867 | autrijus++ | * ./pugs -C Perl5 (and -C JSON, -C Binary) are deprecated. 00:51
r7867 | autrijus++ | Use "-C PIL1-Perl5" instead; it doesn't have to be a dash;
r7867 | autrijus++ | any nonalphanumerical character would do:
r7867 | autrijus++ | ./pugs -CPIL1.Perl5 -e 'say 123'
r7867 | autrijus++ | This is because "-C Perl5" will be reserved to mean the
r7867 | autrijus++ | backend that actually compiles to Perl 5.
rafl Is there a way to do a "lexically scoped" chdir? 00:54
eric256 wonders why pugs is taking SOOO long to build on feather now... hmmm 00:55
autrijus rafl: why, yes, 00:59
temp $*CWD = '/path'
if you mean p5, File::chdir on cpan does the same. 01:00
I think I'll sleep now -- tomorrow should see the beginning of -CPerl6.
autrijus waves &
Juerd afk 01:02
z
eric256 rafl: you around? 01:13
wanna try your debian package stuff? build passed finaly ;)
rafl eric256: Yes. 01:15
eric256: First tell me how to do anon closures in p6. :-) 01:16
Khisanth $foo = { ...; } ? 01:17
rafl sub { ... } seems to work. 01:19
Khisanth both work :)
unless your sub looks like a hash to pugs 01:20
rafl Look at perlcabal.org/~rafl/autobuild.p6 -- in recompilation_is_needed @list.elems says 2 so it should return bool::false. It returns something true nevertheness. 01:22
Using an explicit return $ret in work_in fixes that. Why doesn't --> $ret work? 01:24
dduncan so, for those of you using svk to commit pugs offline, do you set svk up to first mirror only the latest pugs rev, or the whole pugs history, or do something else 01:25
Khisanth rafl: perhaps not implemented yet? :)
rafl THought is was.. 01:26
eric256 rafl i would guess it doesn't work yet.
?eval sub test (--> $ret) { $ret = "hello";"nope" }; print test();
evalbot_7867 Error: unexpected "-" expecting formal parameter, ":" or ")"
eric256 ?eval sub test ($str --> $ret) { $ret = "hello";"nope" }; print test("a"); 01:27
evalbot_7867 Error: Undeclared variable: "$ret"
eric256 thinks he doesn't understand --> ;)
rafl ?eval sub foo ($a, $b --> $c) { $c = $a + $b; "I'll return $c anyway!".say } foo(1, 2);
evalbot_7867 Error: Undeclared variable: "$c"
rafl ?eval sub foo ($a, $b --> $c) { my $c = $a + $b; "I'll return $c anyway!".say } foo(1, 2);
evalbot_7867 I'll return 3 anyway! bool::true
rafl :-( 01:28
OK.. please do some commits on parrot and pugs so I can verify it works ;-)
Or do we already have a getopt lib?
theorbtwo wonders if it's strange that Jet Lee films now always remind him of Autrijus. 01:29
Khisanth rafl: why do you expect $ret to be returned? 01:30
rafl Khisanth: Because perl 6 can do named returns 01:31
eric256 will be able too...doesn't yet ;) /me goes to give daughter bath...back in 30 01:33
rafl What's wrong with this version: perlcabal.org/~rafl/autobuild.p6 01:34
It doesn't get parsed right.
Khisanth hrm but according to S6 that --> is only to indicate the return type 01:36
theorbtwo ?eval sub test(--> Int) {3.14}; test(); 01:38
evalbot_7867 Error: unexpected "-" expecting formal parameter, ":" or ")"
theorbtwo ?eval sub test($a --> Int) {3.14}; test();
evalbot_7867 Error: No compatible subroutine found: "&test"
theorbtwo ?eval sub test($a --> Int) {3.14}; test(1); 01:39
evalbot_7867 157/50
theorbtwo I don't think it does anything. 01:40
rafl How about the parse error in perlcabal.org/~rafl/autobuild.p6 01:43
I don't see what's wrong.
Khisanth rafl: but what is the error? 01:45
message
rafl unexpected "@" 01:46
expecting block construct, ":", word character, "::", term postfix, operator, ",", "!!", postfix conditional, postfix loop, postfix iteration, ";" or "}"
at autobuild.p6 line 20, column 16
Khisanth perhaps you want @ARGS[0] or @*ARGS[0]? 01:47
rafl That doesn't change anything. 01:50
eric256: OK, auto building/installing is done via dpkg now. 02:06
eric256: Only the haskell interface to pugs isn't installed.
azuroth afternoon, all 02:20
rafl Hi azuroth 02:22
azuroth how's it all going?
rafl Quite well. I'm happy that feather now uses the Debian packages of pugs and parrot and that I can upload parrot officially to the Debian archive tomorrow. :-) 02:24
azuroth oh, excellent! 02:25
how about ubuntu..? 02:27
rafl I'm not an ubuntu developer. 02:28
azuroth ahh 02:32
eric256 rafl: do you have a script for me to use? or is it running on your account? 02:34
rafl It runs on my machine currently. You have the right's to install debian packages using sudo withoug a password as well. You can steal ~rafl/autobuild if you want. 02:37
eric256 actualy i can't cause i can't get into that directory ;)
if you have it setup to cron thats fine with me 02:38
your building pugs and parrot?
rafl I set up cron and I build parrot and pugs (with p5 and parrot embedding)
eric256 sounds fine to me. i'll still build and smoke every 6 hours. 02:39
rafl OK 02:40
eric256 passes an only recently acquired torch on to rafl ;)
?eval ('a*' ~~ /<p6rule>/) 02:41
evalbot_7867 Error: cannot cast from VUndef to Pugs.AST.Internals.VCode (VCode)
eric256 wishes eval bot used the same version of pugs as the general feather one 02:42
azuroth haha. that's crazy 02:43
eric256 azuroth?? 02:44
azuroth rule called p6rule 02:45
eric256 yea. its pretty cool. gives you a hash representation of the match object...kinda hard to explain but if you try it on feather you'll see the output. ;) 02:46
i can't wait tlil we have the p6code rule ;) or a p5code rule. be able to parse code on the fly ;) 02:48
azuroth :D 02:49
what will the benefits of that over eval be, though? ;-p
eric256 source filters that work on an actual syntax tree? or how about building your own grammar for a script language? its the hard part of making your own programming language already done! ;) 02:50
rafl eric256: Why doesn't it do that? 02:51
eric256: wrt evalbot and the pugs executable
azuroth ohh
eric256 easy safe.. just scan the parsed tree for unsafe calls. etc. color coding a perl6 source file, super smart search and replace, editors tahn can refactor
and more. ;)
rafl: no idea, been looking for ages for the owner of that bot, it should also stop logging on and off every time someone commits, eventualys its going to hit some rate limit on IRC ;) 02:52
rafl eric256: Yes! 02:53
eric256 in case azuroth didn't notice eric256 can think of 101 uses for parsing perl and thats just the start ;)
rafl eric256: It runs on feather. fibonaci (Luke) runs it.
eric256 luke!!! i saw fibonaci but never put 2 and 2 together. ;) 02:54
rafl finger fibonaci.. whois fibonaci
azuroth syntax highlighting would be awesome
rafl goes to bed. 4am already.. bye! 02:55
azuroth night, rafl
eric256 oddly that bit with the rule doesn't work on the command line at all. but it does in interactive.. 02:56
is there some difference between those two i'm not aware of?
azuroth you make me strong, you're my vitamin 03:00
eric256 gaal: ping 03:24
nm. i had a question about your article. but it makes sense after all ;) 03:25
Khisanth ?eval my $h = hash( a=>1, b=>2, c=>3 ); 03:45
evalbot_7867 Error: cannot cast from VUndef to Handle (VHandle)
Khisanth ?eval my $h = {a=>1, b=>2, c=>3}; $h.perl
evalbot_7867 "\{(\"a\" => 1), (\"b\" => 2), (\"c\" => 3)}"
eric256 ?eval my $h = hash(( a=>1, b=>2, c=>3 ));
evalbot_7867 Error: cannot cast from VUndef to Handle (VHandle)
Khisanth ?eval my $h = {a=>1, b=>2, c=>3}; $h<a b c>;
evalbot_7867 [1, 2, 3]
Khisanth hrm
?eval my $h = [{a=>1, b=>2, c=>3}]; $h[0]<a b c>; 03:46
evalbot_7867 [1, undef, undef]
Khisanth ?eval my $h = [{a=>1, b=>2, c=>3}]; $h[0].<a b c>;
evalbot_7867 [1, undef, undef]
azuroth ?eval (2,4) == (5,7)
evalbot_7867 bool::true
azuroth ?eval (2,4) ~~ (5,7) 03:47
evalbot_7867 bool::false
azuroth ?eval my @a = (2,4), @b = (5,7); @a == @b 03:48
evalbot_7867 Error: Undeclared variable: "@b"
azuroth ?eval my @a = (2,4); my @b = (5,7); @a == @b
evalbot_7867 bool::true
Khisanth an overloaded ==?
azuroth ?eval my @a = (2,4); my @b = (5,7); *@a == *@b 03:49
evalbot_7867 bool::true
azuroth ?eval my @a = (2,4); my @b = (5,7); ~@a == ~@b
evalbot_7867 bool::false
Khisanth ?eval my @a = (2,4); my @b = (5,7); my $a = @a; my $b = @b; $a == $b
evalbot_7867 bool::true
azuroth it was happening on feather too, so I wanted to check evalbot
Khisanth ?eval my @a = (2,4); my @b = (5,7); my $a = @a; say $a.perl, $b.perl
evalbot_7867 Error: Undeclared variable: "$b"
Khisanth ?eval my @a = (2,4); my @b = (5,7); my $a = @a; my $b=@b; say $a.perl, $b.perl 03:50
evalbot_7867 [2, 4][5, 7] bool::true
Khisanth why are they equal?
azuroth ?eval my %a = {:a(5), :b(2)}; my %b = {:a(3), :b(0)}; %a == %b 03:51
evalbot_7867 bool::true
azuroth ?eval my %a = {:a(5), :b(2)}; my %b = {:a(3), :c(0)}; %a == %b
evalbot_7867 bool::true
azuroth ?eval my %a = {:a(5), :b(2)}; my %b = {:a(3), :b(0)}; %a ~~ %b
evalbot_7867 bool::false
azuroth ?eval my %a = {:a(5), :b(2)}; my %b = {:a(3), :d(0)}; %a ~~ %b
evalbot_7867 bool::false
Khisanth ?eval my @a = (2,4); my @b = (5,7,8); @a == @b
evalbot_7867 bool::false
Khisanth ?eval my @a = (2,4); my @b = (5,7,8); @a > @b 03:52
evalbot_7867 bool::false
azuroth hmm, ok. I thought smart match was supposed to check the keys, not vals. must be mistaken
Khisanth oops
?eval my @a = (2,4); my @b = (5,7,8); @a < @b
evalbot_7867 bool::true
azuroth so it's acting as if it's using numeric context? 03:53
Khisanth I guess
jdv79 seen stevan 05:20
jabbot jdv79: stevan was seen 1 days 8 hours 7 minutes 56 seconds ago
gaal eric256: pong + pong**-1 :) 05:42
eric256 ;) 05:43
map being lazy is...confusing. ;) at least after p5
btw i think the class sigil is now ^ , i've seen larry use it a couple times now ;) sub test (^class $test) {}; 05:44
Khisanth that was also used in the Synopsis 05:48
gaal map being lazy is *wonderful*
hmm, thanks for the class tip! this really shortens the unicode list 05:49
eric256 i dunno about wonderfull.... map being able to be lazy ++...defaulting to lazy...i dunno. ;)
btw on your split().join() example why not split on space, join with - or something. just so it seems less pointless. ;) 05:50
?eval "Hello, world!".split(" ").join("-").say;
evalbot_7867 Hello,-world! bool::true
eric256 ?eval "Hello, world!".split(" ").join("\t").say;
evalbot_7867 Hello,world! bool::true
gaal hee, well the point was that a list was an invocant. you can't do that in p5. that block should probably go elsewhere where i have space to elaborate 05:51
eric256 and the equvilant p5 might be more like...print join("-", split(" ", "hello world")),"\n"; just so you are comparing apples to apples
gaal you have the usual p5 chain:
$out = join split $in # pseudocode 05:52
but these are builtins
.join and .split are methods
?eval <3 1 4 1 5>.join("") 05:53
evalbot_7867 "31415"
gaal can't do that in p6
errr
in p5
eric256 10 01?eval <3 1 4 1 5>.join()
10 01?eval <3 1 4 1 5>.join;
?eval <3 1 4 1 5>.join;
evalbot_7867 "31415"
eric256 stupid space before ? lol.
point just is that in p6 you don't need to pass empty string, so either don't make it empty or don't pass it, because it looks like you are writing p5.5 ;) 05:54
any i should sleep. article look great BTW!
gaal anyone got a link to the ruling on the cent sigil? 05:55
thanks!
aaah, thanks, good point
eric256++
we need a mailing list interface where, like in the spirit of anonpan, people can flag emails with the 'ruling' bit. 05:58
.oO( hmmmm, i think i may leave join("") for symmetry with split(""). )
 06:01
eric256 ?eval "hello".split;
evalbot_7867 ("hello",)
eric256 hmmm interesting 06:02
gaal not special: whitespace based split, like p5.
eric256 S02/Names and Variables 06:03
gaal ahh, thanks.
eric256 ?eval "hello world".split.join;
evalbot_7867 "helloworld"
eric256 ?eval "hello world".split.join("-");
evalbot_7867 "hello-world"
eric256 ?eval "hello world".split("").join("-");
evalbot_7867 "h-e-l-l-o- -w-o-r-l-d"
eric256 [~] "hello".split(""); 06:04
gaal makes perfect sense when you think of it. just like p5
eric256 ?eval [~] "hello".split("");
evalbot_7867 "hello"
eric256 i'm not sure p5 ever realy makes perfect sense. ;) and i'm not sure spliting on ' ' implicity is DWIM. guess its splitting on \s+ not ' '..anyway realy must sleep. later 06:05
gaal it's a little more DWIMmier than that, it throws away trailing records that are empty :) but i have to say that often that was *exactly* what I wanted. 06:06
night :)
azuroth why is it called golfing system, anyway? 06:11
svnbot6 r7868 | gaal++ | tpr.pod: remove obsolete class sigil from unicode list, add eric256++ 06:14
gaal azuroth: it helps you get to your goal with fewer kestrokes 06:24
work & 06:27
azuroth ahh
dduncan well, I seem to have gotten svk installed, so tomorrow I can start using it 06:38
svnbot6 r7869 | stevan++ | Perl6::ObjectSpace - nil now has &is_not_nil; so does type; more converting of Boostrap.pm to s-expressions about 80% done now (see docs/sexp_meta_model.pod)
dduncan unfortunately, it spews a bunch of warnings from Class/Autouse.pm under the system perl, though the latest perl doesn't have those problems ... but I can't easily get it to work with that newer perl in a nonstandard location, so I'm stuck with two-dozen warnings every time I invoke svk 06:40
azuroth ?eval sub a($a,$b){say $a ~ $b; }; a 1 2; 06:44
evalbot_7869 Error: unexpected "2" expecting operator, ":", ",", term postfix, postfix conditional, postfix loop, postfix iteration, ";" or end of input
azuroth ?eval sub a($a,$b){say $a ~ $b; }; a 1, 2;
evalbot_7869 12 bool::true
azuroth hm, ok. I didn't see optionality of ()s for subs in perl 6 essentials - just methods 06:45
dbrock I thought it was the other way around 06:47
azuroth ?eval "hello world".split "o" 06:48
evalbot_7869 Error: unexpected "\"" expecting operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
azuroth ?eval "hello world".split("o")
evalbot_7869 ("hell", " w", "rld")
dbrock :-)
azuroth hah. that's interesting...
?eval split "hello world": "o" 06:49
evalbot_7869 ("hell", " w", "rld")
azuroth hmm, okay, maybe it was only talking about that
dbrock I don't remember or understand why allowing <"hello world".split "o"> is problematic
but I'm sure there is a very good reason :-) 06:50
azuroth hmm. yeah
dbrock and I'm sure it's done that way so that we can enjoy other cool and convenient niceties
?eval split "hello world", "o" 06:52
evalbot_7869 ("o",)
dbrock ?eval split "o" 06:53
evalbot_7869 ("o",)
dbrock what does that list syntax mean? 06:54
?eval ("o",)
evalbot_7869 ("o",)
dbrock ?eval ("o", undef)
evalbot_7869 ("o", undef)
dbrock ?eval ("o",,,)
evalbot_7869 ("o",)
azuroth ?eval split 06:55
gaal method calls require parens if they have params
evalbot_7869 Error: No compatible subroutine found: "&split"
gaal ?eval "hello world".split("o")
evalbot_7869 ("hell", " w", "rld")
gaal but i'm not here :)
azuroth you're having fun working, right? 06:56
gaal no, i'm having fun being late for work :)
azuroth haha
gaal &
azuroth wonders about porting rails to perl 6 08:25
r0nny yo 08:27
azuroth hey r0nny
is there any ported database stuff yet? 08:50
r0nny i dont think there is such stuff atm 08:53
azuroth ;-)
wolverian azuroth, use perl5:DBI; 09:19
azuroth or maybe I could port DBI, or something with a similar goal at least? 09:20
wolverian right. whatever is fun :) 09:23
azuroth :D 09:24
webmind morning 11:03
hm is there anything like sysread for perl6 ?
r0nny_ yo 11:04
i dont know if sys_read is implemented 11:05
rafl webmind: There's a sysread in the S29 draft.
Why must I use Pipe::open instead of Pipe.open which seems more natural to me? 11:07
webmind S29 ? 11:09
gaal rafl: Pipe.open "should" work to the best of my knowledge. 11:10
that is, i think that's what the spec says, but haven't seen it actually working that way recently. It did use to. 11:11
rafl gaal: It doesn't as it seems. No compatible subroutine found: "&Pipe" 11:12
r0nny_ hmm - i think i found a bug - i did try{eval "use foo"} and pugs exited from cmdline mode with a error instead of catching the try stuff
btw - is there a way, i can do eval<foo> "bar" intead of eval "bar", :lang<foo> 11:15
rafl pugs compile error: src/Pugs/PIL2.hs: file name does not match module name `Main' 11:17
Can someone confirm this?
r0nny_ moment
rafl It's a newly checked out tree. 11:18
From feathers svk mirror.
webmind hmm, sysread is pending in S29 ? 11:20
gaal rafl: looks like a bug to me. but i'm $working so i can't look into this now. sorry :/ 11:21
rafl gaal: The Pipe or PIL2 storry? 11:23
btw: working on sunday? Poor gaal..
r0nny_ re 11:51
rafl: for me it works fine
gaal rafl: Pipe.open ought to work IMO (I may be wrong, check with the L people...) 12:00
rafl: aw, not to bad: I bet you do work on Fridays :) 12:01
rafl gaal: I do work once a week if I'm in mood for it. ;-) 12:04
autrijus rafl: I'll look into the PIL2.hs borkage 12:18
bbiab
svnbot6 r7870 | autrijus++ | * drift.pl - don't corrupt the PIL1.hs and PIL2.hs 12:58
r7870 | autrijus++ | when DrIFT is not installed.
rafl autrijus: That is what made me wonder if pugs depends on drift yesterday. With drift installed it worked. 13:06
svnbot6 r7871 | iblech++ | * Usual svn props, 13:10
r7871 | iblech++ | * pugs::run: Documented the new names for the backends.
geoffb rafl: it seems it *did* depend on drift -- it just wasn't supposed to. :-) 13:16
rafl syntax error at util/drift.pl line 14, near "$in;" 13:18
svnbot6 r7872 | iblech++ | Pugs.Pretty: Correctly prettyprint one-elem-arrays (e.g. (3,) should be 13:21
r7872 | iblech++ | prettyprinted as (3,), not as (3)).
r7873 | iblech++ | * t/subroutines/loopscope.t: Added some more tests and s/"a".."c"/1..3/ (as 13:56
r7873 | iblech++ | this test tests $_ in loops and not .. on chars (which PIL2JS doesn't support
r7873 | iblech++ | yet)).
r7873 | iblech++ | * t/operators/reduce-metaop.t:
r7873 | iblech++ | * Added try {...} blocks around in-PIL2JS-inflooping tests.
r7873 | iblech++ | * Fixed the [=] test (eval_is can't find lexicals) and added some
r7873 | iblech++ | more [=] tests.
r7873 | iblech++ | * PIL2JS: Prelude::JS::Operators:
r7873 | iblech++ | * Unbroke [=>] -- PIL2JS's => currently captures containers, i.e.
r7873 | iblech++ | my $pair = ($key => $val);
r7873 | iblech++ | $pair.val = 42;
r7873 | iblech++ | say $val; # 42; correct? 13:57
r7873 | iblech++ | * Implemented [=] and [,] (with all tests passing).
r7873 | iblech++ | * New t/pugsbugs/arity-of-blocks-with-placeholders.t:
r7873 | iblech++ | While fixing [=>], I noticed that .arity calculation with blocks with
r7873 | iblech++ | placeholders doesn't work in all circumstances:
r7873 | iblech++ | { $^a,$^b }.arity # 2
r7873 | iblech++ | { my $foo; $^a,$^b }.arity # 1!
r7873 | iblech++ | { $^a,$^b; my $foo }.arity # 2
xerox haha, .arity is the best.
Juerd iblech: When would you want to use [,]? 14:02
iblech Juerd: obfu? :)
Juerd Good enoug.h
s/.h/h./
iblech (&prefix:«[,]» ::= &list)
Juerd The list operator makes no sense. Why do we still have it? 14:03
xerox What is [,] ? .-)
Rebuilds the list?
Juerd xerox: Do you know what [+] does?
xerox Yes.
Juerd xerox: Then you know what [,] does.
xerox Nothing, more or less.
iblech TIMTOWTDI probably -- if you don't want to type ($a,$b,$c), you can type list $a,$b,$c (but yes, it's redundant) 14:04
xerox: Yep. Even the containers are kept -- ([,] $a,$b)[0] =:= $a is true 14:05
Juerd iblech: The thing is, that in all cases where "list" actually returns a list, the entire keyword could be left out.
What remains is a "list" operator that practically doesn't have anything to do with lists. 14:06
Juerd already knows his first Perl 6 best practice: Never use the "list" operator :)
iblech Yep, but consider: my $arrayref = list $a <=> my $copy_of_a = $a 14:07
Juerd I think the former is very bad style.
iblech It's probably mostly useful in situations where you don't want to write ($single_element,)
Juerd It should be: my @arrayref; @arrayref[0] := $a
Which happens to be *@arrayref := $a 14:08
Iff you really need this, which I think you will not, ever.
iblech I used this a few times in PIL2JS's Prelude. It's important to know the differences between my $arrayref = ($a,) and $arrayref = [$a], of course (the former does not create new containers, while the latter does) 14:10
Juerd Then still, "list" is obfuscation 14:12
And why do you use $arrayref = ($a,) instead of *@array := $a? 14:13
iblech Well, some people (wolverian?) prefer to type list $a instead of ($a,) because they don't like ($a,)'s appearance 14:15
Because using subroutine signatures as LHS of := is not yet implemented and because I may want to have an arrayref, not an array.
Juerd Wanting a ref instead of an array is weird, in Perl 6. 14:16
Because the moment you use the real array as a ref, it will behave like one.
Juerd thinks the documentational value of the @ is very valuable 14:18
svnbot6 r7874 | iblech++ | util/drift.pl: drift.pl should no longer print to STDOUT, but to @ARGV[1].
Juerd And even a good reason to write "my @foo := %bar<baz>;" instead of "my $foo_ref = %bar<baz>;" 14:19
Juerd thinks := will be very popular 14:20
svnbot6 r7875 | iblech++ | t/operators/binding/{arrays.t,hashes.t}: 14:29
r7875 | iblech++ | @array := $arrayref should *not* mean @array := ($arrayref,) (as it would with
r7875 | iblech++ | normal assignment) (consider
r7875 | iblech++ | sub foo (@array) {...}
r7875 | iblech++ | foo $arrayref; # internally, @array := $arrayref is called).
r7875 | iblech++ | Juerd++ for raising this topic.
iblech Juerd: Obviously, I agree :) (on := going to be very popular) 14:30
Juerd And references with $ sigils not :) 14:33
iblech Yep :) I think there are situations where not dereferencing refs is any(useful,more clear) 14:37
Juerd Yes, but I do doubt that those are situations where you'd copy the ref.
When binding, you're not really dereferencing. 14:38
Or referencing.
You may be changing the method of accessing the array
@a := $aref doesn't really deref $aref. It's actually more like @a := @$aref 14:39
It binds directly
(Okay, this can be called dereferencing, of course. I should find a wider vocabulary.)
I might even not be making any sense right now 14:40
Forget this :)
wolverian damn, Messenger sucks. it can't connect, gives a problem ID that MS's help pages don't know about. nice support there. this is what most people tolerate? sigh.
Juerd It's not just what they tolerate 14:41
It's what they love.
wolverian I can understand that. it's purty. but tolerate as well?
Juerd People love weakness.
Isn't $cute := $weak, in a way?
Eh, :=: 14:42
Or was that =:=?
wolverian right. .needs_protection ;)
Juerd I keep forgetting.
Right
iblech Juerd: [@a := @$aref] Yep, right.
Juerd Don't we all love the English language?
It's fragile, makes no sense, but that's a reason in itself to keep it around.
Heck, Perl 5 is as fragile as cute. 14:43
We protect it and love it
wolverian gaim++ # spellchecks many many languages
gaim-- # the languages aren't in alphabetical order when selecting which dictionaries to install
Juerd I don't get why people love Google so much
Perhaps the near empty default page makes it look much more fragile than it is. 14:44
"We're too poor to hire designers, so we let the techies do that kind of stuff"
Sure, *we* know the techies are right, and this is a good interface.
Non-geek people just find it looks silly. Yet they love it. 14:45
Okay, I do get why people love Google now.
webmind heh 14:48
theorbtwo Eurooscon article on BBC news: news.bbc.co.uk/1/hi/programmes/clic...07742.stm. 15:03
Juerd theorbtwo: Small, nitpicky advice: if you want to end your sentence properly, while the last atom is a URL, enclose it in <> 15:14
geoffb theorbtwo, or just put a space before the sentence-ending punctuation . . . . 15:20
Juerd autrijus: Is there any audio or a/v record of your Pugs talk? 15:30
r0nny_ any changes for thread primitives ? 15:59
Simonff Hi - I have a Pugs install problem. Anyone online? 16:02
rafl Just ask.
Simonff After the Pugs compilation is done: 16:03
Pugs build should not get here at ../../inc/Module/Install.pm line 29.
After: cd perl5/Perl6-Container && perl Makefile.PL && /usr/bin/make
This is on Fedora Core 3.
rafl What revision are you using? 16:04
Simonff Pugs 6.2.10
rafl The release? Please try using svn HEAD. 16:05
Juerd Releases are nice for changelogs
They're not things you should actually use ;)
r0nny_ hehe 16:06
Simonff I downloaded a CPAN release. Should I be using SVN?
Juerd They're points in time when stuff is cleaned up, and synchronized.
Simonff: Yes.
rafl Juerd: If they were complete. Try to summarize some thousand commits in a few changelog lines. :-)
Juerd rafl: I prefer incomplete to complete, re this.
rafl: See linux kernels. Nowadays, they just dump a bunch of logs on you and call that changelog. 16:07
In the 2.4 series, you'd get summaries. Comprehensible stuff, albeit incomplete.
rafl Juerd: I see your point. 16:08
Juerd No raise?
rafl Well.. if you like it? 16:09
svnbot6 r7876 | liz++ | grammatical nit 16:24
r7877 | liz++ | some remarks in the article marked with (EM) 16:30
r7878 | liz++ | Added Kane and myself to AUTHORS 16:36
Simonff Okay, I just tried to compile the current SVN version of Pugs. Same problem: 16:39
Pugs build should not get here at ../../inc/Module/Install.pm line 29.
webmind rafl, could you help me out with this debian packaging stuff I said I would do? 16:46
rafl webmind: What's the problem?
webmind wondering what I'd need to do 16:47
don't have much experience with packaging
rafl, uhm -hilight- ? 16:56
rafl webmind: Let me explain this tomorrow, OK? I'm kind of busy currently. 16:57
webmind rafl, alright 16:58
svnbot6 r7879 | autrijus++ | * Empty p6 rules now raises exceptions. 17:12
r7879 | autrijus++ | Reported by Patrick Michaud.
rafl autrijus: Is there a mechanism to glue arbitrary haskell interfaces into pugs? 17:13
clkao lol on the ja pugs presentation :D 17:22
even funnier now i can read some more japanese
gaal rafl: you know about eval :lang<haskell> yes? 17:26
rafl Btw: fresh debian packages for parrot and pugs every 15 mins: perlcabal.org/~rafl/debian/
gaal: Yes, but that would require to write the glue myself.
gaal it's a start. :) 17:27
theorbtwo Nifty, rafl! 17:31
rafl gaal: I'm lame, you know? :-) 17:38
svnbot6 r7880 | liz++ | Braindump of coro meeting 17:48
wolverian filepaths would be nice in the commit messages.. 17:50
rafl Yeah, like svk log -v does 17:52
gaal wolverian: svn log -v. I think it's too long for pasting in the channel. 17:53
wolverian hm. I mostly meant manually typing them when relevant. 17:55
but that works too. thanks :)
gaal ah, you mean a developer convention where people give context about what they're doing? I'm never sure if I should do that or not 17:56
we have trouble enough remembering the convention of prefixing things we think should go into the changelog with "* ". 17:57
theorbtwo If the svnbot was smart enough, it could summerize the files that got hit, possibly. 17:58
gaal how? 17:59
theorbtwo Hold on a sec, let me see what a normal spread is like.
gaal like logjam factors out common prefixes in journals?
this idea, only with paths? logjam.danga.com/tour/usejournal.png 18:00
because i don't think that'll be useful here :) 18:01
theorbtwo Yes, lots like that.
Hmm, looking at the log, it looks pretty useful ot me. 18:02
gaal it must be smart enough to protect against Great Renamings where millions of files change 18:03
theorbtwo Yeah. 18:04
For that matter, to looks like many commits only hit a single *file*. 18:05
(And a directory, for some strange reason.)
gaal svn propchanges probably. 18:06
svnbot6 r7881 | autrijus++ | * add Allison to AUTHORS 18:12
r7882 | liz++ | Junctive Autothreading and Hyper Operations fart 18:13
iblech liz6: Your last change to AUTHORS removed many people -- intentional or $EDITOR having problems with Unicode? 18:17
svnbot6 r7883 | iblech++ | S17draft.pod: My thoughts -- coroutine parameters should never be rebound.
r7883 | iblech++ | Instead, yield(...)s return value is an Arglist object containing the new
r7883 | iblech++ | arguments.
theorbtwo Interesting idea, iblech.
liz6 iblech: it did? argh 18:18
iblech theorbtwo: Old discussions about coroutines on p6l wanted this too -- but at this time, Arglist didn't exist 18:19
theorbtwo This is actually the first I've heard about it, but I'm quite glad it exists.
iblech liz6: See rt.openfoundry.org/Foundry/Project/...;rev2=7878
Yep, it's really great :) See t/data_types/arglist.t for examples 18:20
autrijus iblech: so if you are to rebind
you'll go like
coro ($x) { $x := yield($x*2); $x := yield($x*3) } 18:21
right?
except you don't want to do that when it's not called with something
xerox should have done more for pugs.. :(
iblech Hm, probably you'll need to splat -- $x := *yield($x*2), as otherwise $x would be bound to the Arglist object. But yes, that's the idea 18:22
liz6 iblech: looks more like a borked installation somehow to me...
autrijus hm, I think liz6's svk copy of pugs is broken
svnbot6 r7884 | autrijus++ | * return() now resets the coroutine pointer, as S17 will
r7884 | autrijus++ | very soon spec. This makes all of t/unspecced/coro.t pass.
r7884 | autrijus++ | (the test doesn't yet cover rebinding of arguments -- which
r7884 | autrijus++ | will happen when you call the coro again with arguments, but
r7884 | autrijus++ | not if you call it as a nullary -- gnomes welcome)
autrijus it has old revisions of AUTHORS
I'll help her fix it after dinner
liz6 autri: tx
autrijus iblech: my idea is that a continuing coro becomes multi 18:26
the argless form means continue
and arful form means rebind.
I think have the user manually rebind yield() will be asking a bit much 18:27
but I'm being called to dinner
bbiab
wolverian wow, coldfusion is really bad. really really really bad.
iblech Yep, but then you, as an coroutine author, can't be sure that your variables stay unchanged around yield() calls (but you know that, of course). Not sure... 18:28
svnbot6 r7885 | autrijus++ | * repair the AUTHORS borkedness due to corrupt svk repo
theorbtwo That was a quick dinner... 18:30
xerox Anybody has ACM access?
clkao autrijus: what happened to her svk repo? 18:47
r0nny_ re 18:55
autrijus iblech: actually you can easily do that 19:18
coro foo ($x is copy) { ... }
rebinding can't touch it then (at least we can spec it so)
iblech autrijus: Ah! I'd be fine with this :) 19:19
autrijus because it's like -- coro foo ($x) { my $xx = $x; ... }
except implicit in the trait
oh cool!
liz6: add that do S17? :)
s/do/to/ 19:20
Khisanth 3hours 20minutes to build pugs >_<
no wonder it feels like it takes a long time :P
autrijus the way for "is copy" to prevent that would mean that rebinding still happens, but the rebinding takes place in an outer scope, and the "is copy" is doing a ($x = $OUTER::x) 19:21
conceptually
iblech yep
autrijus except it's not really OUTER -- user doesn't see it
so it is a bit murky
you can of course do it by hand: 19:22
coro foo ($x) { { $x := $OUTER::x; ... } }
which would do away with an assignment
iblech right. 19:25
autrijus I'm not sure -- I like yield() being symmetrical but the extra shuffling care is a bit too much. I think we keep the multi idea, and let the rebinding be implemented by the postcircumfix:<()> handler
and you can reasonably subclass Coro or use a trait to get different rebinding behaviour.
because the same handler also controls what yield() will get back 19:26
19:29 Lopo_ is now known as Lopo
autrijus iblech: otherwise, how do you like the idea of explicit .start form? 19:30
iblech autrijus: [subclass Coro or trait] fine with me (and note that I don't have a strong opinion on the coro stuff) 19:32
I like .start very much. That way you can even pass coros around to other modules in a safe way 19:33
autrijus yup
liz6 clkao: the initial svk sync was interrupted 19:34
clkao: a restart after that seemed to have borked my repo
clkao: autri is giving me her copy 19:35
Juerd liz6: What was this with Wendy calling him Audrey all of a sudden? Or did I misunderstand that? 19:37
Juerd hadn't asked before. Kind of forgot. "Autri" ringed a bell. 19:38
liz6 juerd: wasn't that "autri" ?
Juerd Then I didn't hear it correctly :) 19:39
I thought maybe it was referring to something that happened while I wasn't paying attention, or something like that.
liz6 I don't think you missed anything...
Juerd Okay, then I misheard Autri as Audrey :) 19:40
liz6 I guess so... ;-)
autrijus iblech: oh, can you think a bit about two bad ideas of mine? 20:05
iblech autrijus: sure :)
autrijus class Hash { &postcircumfix:<( )> { derefing a hash } } 20:06
er
class Hash { &postcircumfix:<( )> { ... } }; %ENV()
that's one
bad idea two
$x = 3; # Scalar.new.infix:<=>(3)
azuroth go go go speed racer
autrijus basically, make the Assign form a meth call on container obj, and make the Apply form a meth call on code obj 20:07
iblech Thought about these, too
Yep
Implementation-wise it'd be relatively easy
What about :=? 20:08
autrijus this is so that everything is an object, for real; it also strenghtens the container/value difference
you think we can pull off Arglist.infix:<:=> ?
iblech I think there's a problem with making := really $lhs.infix:<:=>($rhs), as subroutines can't rebind the vars of their caller 20:09
i.e.:
autrijus $OUTER::x := 3;
why not?
$CALLER::x := 3; # even 20:10
iblech my $foo; sub bar ($var) { $var := $some_other_var }; bar($foo); # $foo not rebound
hm
autrijus er but it is not Container.infix:<:=>
it's Arglist.infix:<:=>
we are promoting $var into :($var)
coercion, even
when using :=
(am I making sense?) 20:11
iblech You are making sense :)
autrijus good. so Arglist is like chip's PadInfo
it dispatches the binding forms
iblech I'd have guessed that rebinding $CALLER::x causes the binding of $CALLER::x -> caller's $x to get destroyed (similar like rebinding %hash<key>) 20:12
autrijus (I thought larry called Arglist as "Signature")
sure, but that's also what happens with regular $MY::x
I don't know... I'm still thinking 20:13
3 := 3; # this would actually be noop
right?
if we are to go with arglist-coercion road
because it's like saying
Juerd Can an rvalue be on :='s LHS then?
autrijus (sub (3) {}).(3)
(sub ($ where { $_ ~~ 3 }) {}).(3) 20:14
which would be noop
iblech Juerd: It always could (consider @array[$idx]<key>.blarb() := $foo)
autrijus I think that is too surprising :)
so maybe binding is still a special form.
Juerd iblech: I fail to understand this
autrijus turning it into something overridable, while very "io"ish, doesn't seem to do much good 20:15
iblech Juerd: := was never restricted to only take variables as LHS. You were always able to use expressions (and subroutine signatures) as LHS.
Juerd Oh.
Signatures I knew.
But I fail to see how 5 := $foo can ever make sense.
autrijus Juerd: (5) is a Signature. 20:16
that means ($ where { $_ ~~ 5 })
Juerd Yes, but what would the semantics of this be?
autrijus multi fac (5) { 6 }
multi fact ($_) { $_ * ($_-1) }
er, s/fact/fac/
Juerd I understand that it makes sense for pattern matching
But what would it do with :=?
integral throw an exception then $foo isn't 5 presumably... 20:17
autrijus because supposedly, the first thing subroutine does is to run := on their Signature with their input
exactly what integral said.
but I also agree it doesn't seem to make much sense.
integral it's the same as: sub fac (5) { 6 }
autrijus which calls for the treatment that makes := a special form still, not a method on Signature.
integral autrijus: "with their input" is there a function that returns this input, so one can say: (... signature here ...) := functionArgs; ? sort of @_, but not an array 20:18
iblech I agree... 3 := foo should be a compile-time error
Juerd autrijus: Well, := is like argument binding for non-multi subs.
iblech integral: sub foo (\$arglist) { $arglist }?
Juerd autrijus: Of course, there are differences between multis and nonmultis 20:19
integral o_O wow
iblech autrijus: It's "Arglist" according to r6522 of S03.pod
autrijus integral: it's the shiny new s03
iblech: yeah, I noticed
liz6: svn.perl.org/perl6/doc/trunk/design/syn/ 20:20
integral cool
iblech Juerd: See docs/quickref/namespace and t/operators/binding/* for examples (of using complex things as :='s LHS) 20:21
autrijus $x := 3; # sugar for \($x) := 3 then?
iblech I think so.
Juerd This is what I've been waiting for. A reference to a quickref document :)
autrijus and we can spec so value-based where{} forms is only recognized for subroutines
Juerd Thanks :)
I will have a look at the current revisions 20:22
autrijus so \(3) will be utterly rejected
if so, we can move back so that := becomes merely Arglist.infix:<:=>
except it's not really arglist is it... it's now on the rhs
er I mean lhs 20:23
arglist in S03 is something that you feed to subs
ingy hi autrijus
autrijus not something you read into subs
Siglist?
iblech Oh, right, of course.
autrijus ingy: hey 20:24
ingy has many new and exciting projects :)
autrijus ingy: eggcellent
iblech: and if I remember correctly you construct them with :()
so :() is sigs, and \() is args 20:25
iblech autrijus: Yep, although :() was never official, was it?
ingy autrijus: maybe you can teach me some new and exciting languages to program them in
autrijus iblech: yeah, but I think it's starting to make more sense
iblech autrijus: yep
autrijus ingy: try io, maybe
so \(3) makes no sense, and the only thing that makes sub (3) working is a macro 20:26
that turns it into \($ where {$_~~3})
and it is just for binding 20:27
as assignment is handled by something else altogether
($x?, $y?) = 9; # doesn't look right anyway 20:28
iblech *nod*
autrijus but consider this form 20:29
($x, $y) = (1,2);
is this List.infix:<=> ? 20:30
(which then redistributes the args)
into two Scalar.infix:<=> calls
it's how it is implemented in hs runcore anyway
but I think it's worth exposing to userland so user can *gasp* override them.
iblech I'd think so. In current PIL2JS, it's &infix:<,> returning a magical rw proxy object 20:31
yep.
liz6 smells string obfu possibilities
r0nny_ hello ?
iblech Should specifying Siglist objects as LHS of := be allowed? $siglist := ... would surely rebind $siglist (per the "$foo := desugaring into :($foo) :="-rule) 20:32
r0nny_ are conditional variables aviable ?
autrijus sub List infix:<,> is assoc<list> (\$arglist) {}
iblech Probably directly using a siglist as LHS is too evil
autrijus :() := ...
is completely fine as it does something utterly equivalent 20:33
and if you want to do on $siglist you do
$siglist.infix:<:=>(...)
which would surely call into the siglist obj living in the $siglist Scalar.
iblech Yep. But this is not specially syntax-support
Right.
autrijus so yay, we deformed^Wreformed everything into a method call. 20:34
;)
iblech :)
autrijus would you have some tuits to t/unspecced/ them?
that way I can focus on PIL2 and not being able to lie to myself :)
and once it's reasonably working we can S03 them.
(after running them thru p6l of course) 20:35
iblech Not today unfortunately (school starts again tomorrow); but from Wed on, I'll probably have some free time
autrijus okay, that's cool
iblech yep :) 20:36
r0nny_ autrijus: where is a good place too look for thread-controlling stuff 20:37
Juerd r0nny_: An operating system's kernel.
autrijus r0nny_: you mean in current pugs? I don't know ... examples/network/http-server.p6
otherwise see S17 for what will come
r0nny_ hmm
iblech Need to go off now, bye all! 20:38
autrijus iblech++ # general sanity 20:39
mm svk did it to mugwump again 20:49
and to svnbot6.
(huge commit log generated by push) 20:51
mugwump Oh, dear
I'm on svk 1.05
autrijus it's automatically clkao's fault, not youre
mugwump can anyone svn ps --revprop it away? 20:52
autrijus I can do that 20:53
r0nny_ DARN
S17 is basically a list of headlines 20:54
autrijus mugwump: I edited revprops with vi 20:57
;)
should be fine now
mugwump cool 20:58
obra 'afternoon 20:59
autrijus yo obra.
mugwump morning obra
obra how's things, guys?
mugwump: are you in auckland, wellington or other? 21:00
mugwump Wellington
obra nods
sadly, not much chance of getting there when I'm in auckland in december
mugwump How long are you there for? and which dates?
autrijus obra: I just consulted with iblech and we are going to make p6 very much rubyish ;)
obra likely about a week over newyear
autrijus: oh? how?
autrijus (with = and := overridable by user) 21:01
actually that goes beyond ruby.
oh, and .() as well.
r0nny_ autrijus: is there any way to get more information about a thread ?
autrijus r0nny_: beyong S17, not much -- detach/kill/join is all you can do at this moment with pugs 21:02
you can read up on "perldoc perlthrtut"
obra autrijus: have you implemented your S17 proposal? 21:03
mugwump New Years is a great time to be in NZ ... lots of parties going on 21:04
I'll be at rhythmandvines.co.nz ...
r0nny_ autrijus: what exactly does async return ? 21:08
liz6 r0nny: good question:
I think autri and I disagree on that...
I think it shouldn't return anything
if you need threads, use threads
in P5 ithreads, async { } is just syntactic sugar for threads->new( ) 21:09
which I think is bad
r0nny_ hmm 21:10
u did "use threads;" now it complains about missing modules
works with perl5:threads 21:12
should i use it, or is it a killer ?
liz6 I would not recommend using P5 ithreads in any way, shape of form, really 21:14
well, not in production anyway
because it's the worst of both worlds: forking and threading 21:15
r0nny_ then how do i create a thread in perl6 ? 21:17
liz6: ? 21:18
liz6 obra: I think autri is preparing for travel early tomorrow morning (about 7 hour from now)
obra no worries
r0nny_ and waht the hell does async return ? 21:19
liz6 r0nny: hmmm... not sure at this stage of the game (I only got drafted by autri into pugs this week ;-)
r0nny: in P5 an instantiated threads object 21:20
which you can later use to do a join() on
r0nny_ any propper thread stuff ?
liz6 r0nny: no idea, as I said, please try 21:21
r0nny_ odd 21:22
async returns a thread 21:23
hmm 21:24
SamB r0nny_: that doesn't sound particularly odd to me...
r0nny_ but there is no meta method 21:25
autrijus liz6: ping 21:27
liz6: turns out I've written something to explain the coro too, so if it makes sense, maybe work it in
liz6 autri: pong
r0nny_ coro ?
liz6 coroutine
obra how close is S17 to "ready for review"? 21:28
autrijus obra: another 10hr or so
obra :)
autrijus (of raw work, not of wallclock)
r0nny_ cool :)
autrijus liz6: use.perl.org/~autrijus/journal/27490
I'm still writing it, but thought the coro part may be of use 21:29
r0nny_ autrijus: why does the meta method fail on Thread ?
obra wonders how hard the parrot threading spec will be after this 21:30
autrijus r0nny_: I'm quite sure .meta is waiting on MM/PIL2 :) 21:32
r0nny_ so .meta isnt implemented ? 21:33
autrijus no it is not.
r0nny_ hmm
guess now all my reasons to stay awake are gone 21:34
autrijus good night :) 21:35
obra autrijus: where to next?
autrijus obra: helsinki, to visit aevil+jhi
then after a day, taipei
obra nice. 21:37
.oO {Are you talking to jhi about 6pan or unicode?}
liz6 obra: I think autri mentioned 6pan 21:40
autrijus 6pan mostly, I think, as I think unicode is pretty well understood
obra *nod* 21:42
though I bet he has nice cautionary tales ;)
autrijus very dense braindump journal finished -- hopefully still making some sort of sense 21:46
shower &
dduncan ... dainbramaged 21:55
liz6 dduncan: after a while you don't feel it anymore... ;-) 21:57
dduncan I wasn't talking about anyone in particular ... seeing the word 'brain' reminded me of that word ... 21:58
liz6 I guess we should refer to it as "drainbump" then? 21:59
autrijus paindumbr 22:00
dduncan more fitting in this situation
rong 22:01
anyway, I'm starting to read the latest journals now
on a separate note, my first relational database on pugs will be 'Native', a pure perl reference implementation of the Rosetta API that supports all of its features, but with the conscious decision to save on complexity by allowing it to be slow ... 22:03
correct but slow ... 22:04
this exists mainly to help guide the re-plan of the API and help test it, as well as have *something* working on Pugs, but isn't intended for heavy duty use 22:05
obra using your DBI port? 22:06
dduncan it will be slow mainly because it will maintain the whole database in memory and write it all to disk on a commit, rather than just the changed parts, so it will only perform reasonably on small data sets
fyi, I am doing both at once 22:07
obra but could be fast if you made flushing optional?
r0nny_ dduncan: what exactly is the Rosetta api ?
dduncan saving it to disk will be optional
it is essentially an in-memory database that can optionally be saved to disk as well
the saving is necessary to be ACID compliant ... when a commit returns without error, you know the data is saved 22:08
the in memory only version would be more as scratch space for an app, not persistant storage
using this will loosely resemble using SQLite, but it will implement features that SQLite doesn't, and SQLite is fast 22:09
the whole db is in a single on disk file that can be stored anywhere
obra why not just implement an SQLite binding?
dduncan I will do that later ...
I'm doing my own first because I want to have a solid testing example for Rosetta, and SQLite doesn't support a number of its important features 22:10
obra Rather: Implementing your own database seems like a possibly painful operation that could be short-circuited.
what's SQLite missing that rosetta needs?
dduncan it would only be painful if I tried to make it fast
obra (Or DBD::mysqlPP)
is there already an SQL parser in perl6?
dduncan one key difference is that no SQL strings have to be involved in the backend 22:11
Rosetta natively takes an AST, and my db would natively use that
no work in generating SQL to be parsed again
obra ah
r0nny_ so Rosetta is basically like CDBI, just __more__ sweet ?
dduncan also, the same SQL::Routine that is used to store the AST when an app builds SQL statements can be used to natively store my db's schema 22:12
my db will support multiple schemas in a db, tables, foreign keys, views, stored routines ... SQLite doesn't support native updateable views or most stored routines 22:13
despite appearances, I think that this will actually be quite easy 22:14
obra That would be really cool
dduncan I will follow the SQLite model of having only one writer to the db at once, which has exclusive access, but there can be multiple readers 22:15
all transactions have serialized concurrency
also for simplicity
obra will the writer's locking be opportunistic or will it have to explicitly relinquish control?
dduncan the bottom line is ...
liz6 wonders how Rosetta could fit in with "is atomic" functionality
dduncan the bottom line is that, it is easier for me to conceptualize an API if the database testing it works in exactly the same way, rather than my having to mess around with my mind about how various dbs act differently and how to compensate 22:16
remember that my db is just a reference implementation, so I only care about correctness, not speed
though it will still be designed efficiently within certain bounds 22:17
liz6, in answer to your question ...
assuming the normal case of a database stored on disk, so it is persistant ... 22:18
it will all be stored in a single file, and the operating system's ordinary file locking mechanism will govern access ...
liz6 dduncan: ah, ok...
dduncan when a process starts a transaction that wants to write, it will block until it can get an exclusive lock 22:19
liz6 it's just that "is atomic" functionality has a lot in common with commit/rollback functionality
dduncan it will then perform, in memory, the steps of the transaction
on a rollback, the file is reread into memory, overwriting any changes 22:20
or, on a commit, the changes are written into the file
to deal with interruuptions ...
before writing any changes to the file, a copy of the pristine file will be made on disk first under a known file name, and synced 22:21
then the changes will be written to the original file, and synced
then the copy will be deleted
liz6 well, for a reference implementation of "is atomic" that might make sense...
but maybe just serialization would also do the trick... 22:22
r0nny_ so its basically not thread-save, and emulates this by big locks ?
dduncan what I mentioned is meant to guard against power outages or OS failures or program crash
to save on complexity, the whole db is locked at once, not just the smaller parts ... that's also what SQLite does 22:23
obra you really want your reference implementation of a DB API to be clogged with this stuff?
dduncan threading is a separate issue
obra I can see that as a great 2.0 feature
dduncan I consider ACID to be a fundamental feature 22:24
of a database
if its supposed to be reliable
r0nny_ as ling as it cant deal with threads correct, it wont be used for heavy load parralel stuff
ACID ?
dduncan well, this reference is for the API, its not actually meant to be used in production
ACID is an acronym that essentially means 'reliable' ... 22:25
if a database is ACID compliant, it guarantees certain features ...
r0nny_ hmm
dduncan 1. either all parts of a transaction succeed or none of it has any effect 22:26
liz6 en.wikipedia.org/wiki/ACID
r0nny_ well - i hope your api wont break if real multithreading is introduced
obra it sounds like you have conflicting goals, darren.
dduncan I'll consider threads to be the thing I put off for version 2
obra 1) clean API to implement against
2) full ACID compliance
3) quick to implement and easy to understand 22:27
dduncan this is accomplished partly by sacrificing speed and non-serialized concurrency
and multi-threading
I consider this database to be a test for Rosetta, to help hash it out, as Pugs is a test for Perl 6 22:28
r0nny_ btw - u could save the entire db by using the perl method, and reload it by using eval
dduncan I was going to
obra oh. I thought it was a test for DBI6 22:29
dduncan part of the simplicity is that the file will simply be a text file containing perl code, that when evaluated produced the original data structure
r0nny_ hmm
liz6 dduncan: so how different is that from Storable?
dduncan simpler
liz6 like Data::Dumper? 22:30
r0nny_ well - not so sweet formated
dduncan said perl code will just be a single hash declaration, whose elements are whatever
no mention of objects or anything complex
just a hash of arrays of ...
liz6 ah, ok
r0nny_ hmm
btw - how does the perl method deal with cyclic stuff
dduncan fyi, the older SQL::Routine can already serialize itself in that manner, in only about 10 lines of code
theorbtwo ?eval $a = []; $a[0] = $a; $a 22:31
evalbot_7889 Error: Undeclared variable: "$a"
mugwump pixie is very similar to that, dduncan
theorbtwo ?eval my $a = []; $a[0] = $a; $a
evalbot_7889 [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
dduncan the perl objects will first "serialize" themselves into plain perl structures, hashes of arrays etc, and from that the perl code is generated
theorbtwo r0: Not very well.
dduncan SQL::Routine already does this too
see ...
mugwump also, if you extend the concept to have "mementos", and lazy loading, then you have pixie 22:32
which is a bit like MLDBM, but allowing nodes to store other nodes
r0nny_ i think i found a bug
the perl method cant deal with cyclic data 22:33
dduncan see the get_all_properties_as_perl_str() function on CPAN, and the 2-3 utility functions it uses
in SQL::Routine
it breaks all the cycles
get_all_properties() is the core, spitting out a cycle-broken hash-of-arrays 22:34
the other func wraps that and is like perl()
the output of the first can be fed right back into build_container() to recreate everything
r0nny_ well - i wish someone could write a data-serialisation/deserialisation framework wich deals with all this stuff 22:35
dduncan that already works in my pre-rewrite version ... the new one will be similar
the cycle breaking is built into the object because it knows best about how to do it for its own circumstances
the actual serializing could be farmed out, but I didn't at the time 22:36
r0nny_ requests a Data::Dumper and Data::Reader can do perl code, yaml, xml, and some other stuff 22:37
dduncan question ... when people say they want DBI in Perl 6, what do they *actually* want?
r0nny_ who knows
guess they want enought, to keep theis scripts working, and easy ported 22:38
dduncan as far as I'm concerned, if I get Rosetta working and people can access databases with it (besides my test impl) then their needs have been met
and a DBI-emulating wrapper over Rosetta is already part of its roadmap
r0nny_ darn
dduncan for those older scripts
liz6 dduncan: to an extent, that's true
dduncan and Rosetta is sort of a DBI-v2 in design 22:39
r0nny_ the.perl method is allready lost on simple cycles, but in complex ones
dduncan it may not look it, but you should be able to use it much the same way as DBI
liz6 but I think a lot of people would just want to be able to access their data in existing tables, e.g. in MySQL or PostgreSQL
dduncan its overall design is the same ... common front end with connection handles and statement handles, and multiple specialized backends
the basics like, "just get that data" will be among the first things implemented 22:40
the only main prerequisite is the ability to connect to the database, over which connection other stuff happens
to talk to SQLite in Pugs, we need to be able to invoke its C library API 22:41
to talk to MySQL or Postgres, we either need to talk to their client C library API, or code a network client in Perl
to do either will take a bit of time 22:42
liz6 isn't there a pure perl implementation of DBD::mysql ?
dduncan that's written in perl 5, and has several other perl 5 modules as dependencies 22:43
all would need to be ported or rewritten
if you just want a "get the data", all that really needs porting is "Net::MySQL"
but that is more like the MySQL native API and not the DBI API
mugwump The first thing I want to see is pictures of the data model used for schemas, tables, expressions, etc. 22:44
dduncan that will be forthcoming
SQL::Routine::Language and friends define version 1, and the rewritten version I'm doing now will be loosely similar 22:45
essentially, a catalog has zero or more schemas, has zero or more of any of: tables, views, procedures ... 22:46
liz6 is calling it a day...
dduncan domains, sequences ...
the API of a table and a view are both defined by a row, which is an associative array of names and data types 22:47
a table has one column for each row field
a table also has constraints on it, such as not null, unique value, foreign key
mugwump great. draw it up in Umbrello 22:48
It should be very close to the first thing people see when they look at the framework ... it will explain so much, so quickly.
obra Umbrello? 22:49
dduncan yes indeed
I assume its like what Autrijus uses for his pictures
mugwump well, umbrello is specifically a UML modeller
it's not the best CASE tool on the planet, but it is better than dia when it comes to drawing UML diagrams
dduncan I can make something similar in ASCII or a drawing program as a first draft
mugwump start with dia or umbrello, you'll save yourself a lot of pain. 22:50
dduncan I don't have any specific uml programs
but if they're free for mac os x, I can look ...
mugwump should be. dia runs standalone, umbrello is kde 22:51
dduncan I don't have kde installed
in any event, it will take a few days before I come back with this
on a side note, an official DBIv2 is being hashed out among the usual people ... I'm focusing on Rosetta first partly because it implements a number of features I'd like to see in the actual DBIv2 ... and my implementation will let people actually see the suggestions in action before they commit to them ... also, who knows how long they will take, and meanwhile my doing this will put some functionality on Pugs "now" 22:56
eric256 rafl: around? 23:30

Powered by Cro and the Raku® Programming Language.

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

Thank you!