perl6-projects.org/ | nopaste: sial.org/pbot/perl6 | evalbot: 'perl6: say 3;' | irclog: irc.pugscode.org/ | UTF-8 is your friend!
Set by moritz_ on 4 May 2009.
00:16 mikehh_ left 00:17 amoc joined 00:18 mikehh_ joined 00:19 r0bby left 00:20 ZuLuuuuuu left, r0bby joined 00:21 bacek__ joined 00:22 mikehh_ left, mikehh_ joined 00:30 mikehh_ left 00:33 mikehh_ joined 00:37 mib_8hwlwu joined, mib_8hwlwu left 00:43 wollmers left 00:44 mikehh_ left, mikehh_ joined 00:55 cls_bsd joined
meppl good night 01:02
01:03 meppl left 01:10 cognominal left 01:32 eternaleye joined 01:40 nbrown joined 01:55 yary joined, Patterner left 01:56 yary left 01:59 Psyche^ joined, Psyche^ is now known as Patterner 02:05 Whiteknight left 02:13 r0bby left, r0bby joined 02:27 alester joined 02:44 nihiliad joined 03:04 amoc left 03:06 amoc joined 03:07 japhb left 03:14 Eevee left 03:16 bacek joined, japhb joined 03:18 bacek_ left 03:20 donaldh left, donaldh joined 03:23 mikehh_ left 03:30 nihiliad left 03:38 nihiliad joined 03:39 orafu left, orafu joined 03:46 mikehh_ joined 03:47 mikehh_ is now known as mikehh 03:54 cdarroch left 04:06 SamB left 04:22 [particle]2 left 04:26 skids left 04:27 amoc left 04:32 amoc joined 04:36 [particle] joined 04:58 cognominal joined 05:11 mikehh left 05:14 charsbar_ left, charsbar joined 05:22 cognominal left 05:29 finanalyst joined 05:30 hcchien left, clkao left 05:37 clkao joined 05:49 DemoFreak joined 05:51 sephee left 05:52 sephee joined 05:54 c9s joined 06:11 justatheory left, nihiliad left, hcchien joined 06:14 alester left 06:33 Maghnus joined
moritz_ good morning 06:35
06:39 redicaps joined 06:42 iblechbot joined 06:43 japhb_ joined, japhb left 06:47 redicaps left 06:51 ejs joined
moritz_ rakudo: say sin(3.14) 06:53
p6eval rakudo 6db88b: OUTPUT«Could not find non-existent sub sin␤current instr.: '_block14' pc 60 (EVAL_16:47)␤»
moritz_ rakudo: use Num :Trig; say sin(3.14)
p6eval rakudo 6db88b: OUTPUT«0.00159265291648683␤»
dalek kudo: 966ea72 | moritz++ | docs/ChangeLog:
[docs] update ChangeLog
 06:58
Matt-W Morning 07:05
07:09 mberends joined
mberends hi 07:12
07:13 wollmers joined
wollmers Morning 07:14
Matt-W Hey 07:16
07:20 donaldh left, payload joined, donaldh joined
mberends wishes Rakudo's 'Use of uninitialized value' was more explicit 07:23
finanalyst mberends: i just tried running ./proto all and got:Unrecognized subcommand 'all'.
lambdabot finanalyst: You have 1 new message. '/msg lambdabot @messages' to read it.
mberends finanalyst: that was intended for './proto update all' 07:24
finanalyst oh. shame. 07:25
mberends the response could be clearer, I'll look at doing that in a few minutes.
finanalyst i wanted to install all the modules to review perl6 idioms
by the way, there is a spelling mistake in README. Just a mo, will find it 07:26
mberends finanalyst: './proto install all' might do the trick
07:28 bacek__ left
finanalyst mberends: in PIONEER, "This is what proto was designed for;" I think this should be ':' at the end, not ';' 07:29
mberends yes, you're right. fixed locally, pushing soon. thanx. 07:30
07:30 ejs left
finanalyst mberends: now can find the minor error I saw before. trying install all 07:30
moritz_ james-iry.blogspot.com/2009/05/brie...wrong.html very funny read 07:31
mberends moritz_: excellent, from the first :) 07:33
wollmers moritz_: missing two important languages in this article: APL and 'R' (later badly implemented as SQL) 07:44
moritz_ wollmers: there are many debatable choices in there, but foremost it's funny :-) 07:45
wollmers history of IT is always funny - pragmatic (re-inventions of the wheel) versus never used academic solutions 07:50
Matt-W lol 07:52
I love Haskell jokes
I'm sure they aren't as funny if you don't know Haskell
wollmers Don't know Haskel;-) 07:53
*Haskell
Matt-W Haskell is great :)
Although Perl 6 is more generally practical
wollmers Perl 5 is practical, Perl 6 I don't know good enough at the moment 07:56
Matt-W Perl 6 is much more practical 07:58
Or at least, it will be when we've got more of it
wollmers Will see, if I still will miss the possibilities of APL and Prolog for some (special) problems 08:01
moritz_ the nice thing about Perl 6 is that it's *very* extensible 08:03
08:03 kane_ joined
Matt-W Adding a prolog-style unification system might be pushing the extensibility a little... 08:12
but it's probably possible
wollmers Or implement Prolog on Parrot, but Prolog seems to be nearly dead, very few users 08:16
Matt-W Prolog is too weird for most programmers to swallow, I think 08:17
I noticed that at university
it was hard enough getting people to swallow Haskell, and that's nowhere near as weird as Prolog
where 'weird' indicates a deviation from the imperative programming style which is the most common thing these days 08:18
08:18 sephee left
wollmers Prolog was very hard to learn for me after six years using assembly language 08:19
But mathematicians learn Prolog very fast. 08:21
I teached them Prolog in 4 hours. 08:23
08:25 amoc left
Matt-W Traditional programmers tend to get set in their ways, it seems 08:28
wollmers Of course. At the beginning I used Perl like an procedural language, in structured style. 08:31
moritz_ going from basic and C to perl was quite a change
I took me some time, but the string operations were just fascinating 08:32
Matt-W yeah that would've been an interesting transition
I learned Perl on the back of C++, Pascal and Java
moritz_ I wrote a very simple substitution cypher in basic, in about 50 lines. It was one in perl (tr///). 08:33
wollmers My first Perl script was something like while ($line=<>) { if ($line =~ /.../) { print $line; } } 08:38
moritz_ aka "grep" :-) 08:40
finanalyst mberends: tried ./proto install all and got a parrot error: too few arguments passed (0) - 1 params expected 08:42
mberends looks
wollmers yeah, grep. But I had to use Win.
08:47 fridim joined
mberends finanalyst: thanks, you found a bug. fixed and testing... 08:49
finanalyst++: three more related bugs found and fixed. 08:52
finanalyst i am using to perl6 to generate data that I would like to plot. GNUPLOT is used as a plotting backend for several CPAN modules. So where do I start in writing perl6 wrappers for GNUPLOT? 08:53
moritz_ where? in your head?
I don't quite understand the question 08:54
or do you mean where to get inspiration?
finanalyst where. in my head surely. where should i start looking for how to do this. I havent written wrappers for C libraries before. 09:00
but i cant ask someone else to do it
so if i need gnuplot, i have to write the interfaces
moritz_ you can simply emit gnuplot scripts 09:01
and pipe them into gnuplot
no need to interface C libraries, IMHO
finanalyst can i start a pipe in perl6 ? 09:02
has that been implemented?
moritz_ no
but you can 'run("gnuplot < tempfile")' 09:03
finanalyst thanx. will try that route. very glad i dont have to interface with C libraries 09:04
moritz_ I think it's much easier to debug :-)
09:06 payload left
Matt-W also I don't think we've got an interface for C libraries yet have we? 09:08
Except writing the binding in Parrot
09:08 cognominal joined
moritz_ right 09:09
Matt-W and that's kind of nasty
mberends finanalyst: './proto install all' is fixed and pushed to github.
jnthn H H
Matt-W HAI jnthn
mberends O AI jnthn 09:10
moritz_ jnthn: any news on your hague grant application? 09:11
jnthn moritz_: Didn't finish doing email yet... :-) 09:14
moritz_ :-)
jnthn moritz_: No news that I can see. 09:18
moritz_: Ah well, it's Vienna.pm Rakudo Day today anyways. :-)
moritz_ :-)
wollmers jnthn: you mean the funding by Vienna.pm? 09:28
jnthn wollmers: Yes - I'm funded by them to hack a day a week on Rakudo. :-) 09:29
wollmers Vienna.pm itself seems inactive 09:30
jnthn wollmers: But I have more than one free day a week to do Perl 6 stuff, so I've got an application in for a Hague Grant also, instead of going looking for $other_paid_work
Inactive?
I went to a Vienna.pm meeting like, last month.
moritz_ wollmers: vienna.pm made the YAPC::EU last year
wollmers I know, many of Erlangen.pm visited it 09:31
moritz_ yes. Sadly I wasn't one of them :(
jnthn wollmers: They have meetings monthly or so. 09:32
wollmers: I live not so far away from Vienna, so I attend now and then.
wollmers couriously, I was not there, because of $payed-work. I still have my appartement in Vienna.
jnthn makes coffee and grabs the RT queue 09:41
finanalyst i was looking for the equivalent of the perl5 heredoc << syntax . there seem to be quoting adverbs but rakudo does not implement them. How to stuff lots of lines into a variable? 09:42
mberends finanalyst: just open quotes, enter several lines, close quotes 09:44
finanalyst but if lines contain quotes 09:46
?
mberends finanalyst: qq[ ...several lines... ] is handy to put quotes inside 09:47
finanalyst: example gitorious.org/projects/http-daemon/.../bin/httpd 09:48
finanalyst thanx 09:49
jnthn std: sub foo { }; sub foo { }; 10:10
p6eval std 26720: OUTPUT«ok 00:02 35m␤»
10:29 wollmers left 10:37 pmurias joined 10:45 iblechbot left 11:20 donaldh left, dalek left, dalek joined, donaldh joined
pugs_svn r26721 | jnthn++ | [t/spec] Un-todo 3 tests that we now pass. 11:21
11:23 payload joined 11:39 dalek left, dalek joined
moritz_ rakudo: sub a { }; sub a { }; 11:43
p6eval rakudo dc680c: OUTPUT«Redefinition of routine a␤»
moritz_ rakudo: sub a { }; { sub a { }; };
p6eval rakudo dc680c: ( no output )
moritz_ jnthn++
jnthn moritz_: Not completely sure we're right on that second answer. 11:44
moritz_ rakudo: sub a { "outer" }; { sub a { "inner" }; say a }; say a
p6eval rakudo dc680c: OUTPUT«outer␤outer␤»
moritz_ neither am I
rakudo: sub a { "outer" }; { my sub a { "inner" }; say a }; say a
jnthn moritz_: Since default scope declarator is package, not lexical.
p6eval rakudo dc680c: OUTPUT«outer␤outer␤»
jnthn I knew that my patch wouldn't catch that case though...
moritz_ that for sure isn't right 11:45
jnthn Huh? Lexical subs should work...
OH
*sigh* 11:46
11:46 iblechbot joined
jnthn rakudo: { my sub a { "inner" }; say a }; say a 11:46
p6eval rakudo dc680c: OUTPUT«inner␤Could not find non-existent sub a␤current instr.: '_block14' pc 69 (EVAL_18:53)␤»
jnthn Yeah. That.
11:46 masak joined
jnthn That's a Parrot bug. 11:46
masak Rakudo Day!
moritz_ rakudo: my sub a { "outer" }; { my sub a { "inner" }; say a }; say a
p6eval rakudo dc680c: OUTPUT«inner␤outer␤» 11:47
moritz_ masak: and a bug for you to submit (if you scroll back a little)
jnthn (It says "oh, there's an a in our package, let's skip the lookup and directly reference the package one"
masak moritz_: will do.
moritz_ jnthn: which is the wrong way, for rakudo
jnthn moritz_: I've seen (and analysed) that bug before, or pm has analysed it and we've discussed it for sure. I'd be surprised if we didn't have a ticket already (but it's possible).
moritz_: Yeah, it's an optimization fail for sure. :-( 11:48
moritz_ jnthn: don't know about the parrot side, but I'd be surprise if there were one for the rakudo side
and IMHO we need a rakudo ticket, at least to make sure it gets proper test coverage
jnthn moritz_: Yeah, I think we may have a rakudo ticket because this has come up before. 11:49
jnthn tries to fix up nested role composition to have the Perl 6 semantics 11:51
masak rakudo: { my sub a { "inner" }; say a }; say a 11:53
11:53 bacek left
p6eval rakudo dc680c: OUTPUT«inner␤Could not find non-existent sub a␤current instr.: '_block14' pc 69 (EVAL_18:53)␤» 11:53
masak what do you expect that to say? 11:54
moritz_ exactly that
masak ok, I think I must have missed the bug, then.
rakudo: sub a { "outer" }; { my sub a { "inner" }; say a }; say a 11:55
that one?
moritz_ yes, that one
jnthn yes
p6eval rakudo dc680c: OUTPUT«outer␤outer␤»
jnthn fail :-(
.oO( thanks Parrot... )
moritz_ rakudo: fail
p6eval rakudo dc680c: OUTPUT«No exception handler and no message␤current instr.: 'return' pc 16207 (src/builtins/control.pir:39)␤»
moritz_ fail fail
:-) 11:56
masak hah :)
jnthn We fail at failing.
.oO( so we win? )
masak meta-failure.
11:58 bacek joined
finanalyst mberends: my ./proto install all has just run. It appears to have installed all the modules, but it ends in a failure, viz., Downloading all...Use of uninitialized value followed by some lines and a parrot failure message. 12:03
12:03 eternaleye left 12:04 eternaleye joined
masak finanalyst: does it actually say 'downloading all...'? 12:04
I don't remember it doing that. 12:05
finanalyst i can nopaste you all the output ok?
masak yes, please.
pugs_svn r26722 | moritz++ | [t/spec] RT #64098, methods on enums 12:06
pasteling "finanalyst" at 193.110.4.137 pasted "output from 'proto install all'" (68 lines, 3K) at sial.org/pbot/36505 12:07
masak finanalyst: sir, you just found a bug. :) care to leave a bug report? github.com/masak/proto/issues 12:09
just re-pasting the nopaste content should be enough, actually.
(strange, this used to work.) 12:10
jnthn makes role composition infinite loop and use all his memory. Oops.
finanalyst that's the second time today when running ./proto install all. :( mberends fixed one earlier. 12:11
filing bug
masak yes, I see that from the logs.
sorry about lingering bugs; proto is a bit unpolished still. 12:12
jnthn oh, argh. 12:14
finanalyst but its still won the Nobel prize :) 12:15
moritz_ probably the first software ever to do that 12:16
jnthn The tests that claim to be for 63330 are actually testing 63332
oh no, actually...one tests one, one tests the other...
masak finanalyst: I should lie more often. proto has a wonderful rep nowadays. :) 12:17
moritz_ any bug admins around? RT #65492 is spam 12:18
jnthn moritz_: Is fudge not meant to comment out a whole block if you put a skip above a block? 12:21
masak finanalyst: fixed, thanks. 12:22
jnthn moritz_: oh, never mind...it has to be an outer-most block, not a nested one...then it works.
masak finanalyst: can't find your bug report, so I can't close it. :/ 12:23
jnthn And it passes now. w00t. masak can do diamonds. :-) 12:25
masak yay!
applying CS on Rakudo is a nice way to bring out bugs :)
pugs_svn r26723 | moritz++ | [t/spec] RT #64104, constructing a Range in an action stub
jnthn Yeah. I have a slight fear I may have broken some spectests, but we'll see... 12:27
12:27 PhatEddy joined
masak if it ain't broke, break it. 12:28
jnthn rakudo: role A { }; role B { }; class C is A is B { }
p6eval rakudo dc680c: OUTPUT«The class 'C' already has a parent class ''. It may have been supplied by a role.␤current instr.: '!meta_trait' pc -5324611 ((unknown file):-1)␤»
jnthn Yeah, it's nothing to do with diamonds. 12:29
oh ffs please say Parrot ain't going on string names for checking...
rakudo: role A { }; role B { }; class C is A{ }
rakudo: role A { }; role B { }; class C is A { }
PhatEddy I rejected the spam RT. If there is any other action I should take please let me know ...
p6eval rakudo dc680c: OUTPUT«Unable to parse class definition at line 1, near ""␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)␤»
rakudo dc680c: ( no output )
moritz_ PhatEddy: I meant deletion, not rejecting 12:30
PhatEddy: but my privs don't allow that
jnthn omfg 12:32
/* throw an exception if we already have this parent */
if (Parrot_str_equal(interp, current_name, parent_name))
12:32 ruoso joined
masak o_O 12:33
rakudo: class A::C {}; class B::C {}; class D is A::C is B::C {}
p6eval rakudo dc680c: ( no output )
masak hm. so that's not the problem... 12:34
jnthn masak: I'm patching Parrot.
masak jnthn++
ruoso hello! 12:36
jnthn hi ruoso
jnthn smokes his Parrot fix 12:37
moritz_ in t/spec/S02-whitespace_and_comments/unspace.t
I get a Null PMC access in get_attr_str()
jnthn moritz_: New bug?
(failure)?
erm, regression even... :-) 12:38
moritz_ jnthn: yes, I think so
and t/spec/S04-statements/for.t segfaults after 50 tests 12:39
(with parrot r38602, which is from today)
ruoso wonders if that rumour about we getting contextual variables today would be confirmed... 12:40
finanalyst masak: i cant find my bug report (which u have fixed already). i labelled it 'install all....' in case it mysteriously appears
masak oki.
there appears to be issues with githubs issue-tracker :)
jnthn masak: github has issues. 12:41
;-)
finanalyst first time i used it, so i thought is had got it wrong
masak jnthn: that's a better way of putting it. :)
finanalyst s/is/it
s/it/I/ !!!!!!! 12:42
12:42 kidd_ joined
jnthn Damm. I did break t\spec\S14-roles\crony.t :-( 12:45
otoh my Parrot patch passes all Parrot sanity tests. 12:46
s/sanity///
masak jnthn: gotta go study chinese, then have dinner with some friends, then watch the new Star Trek movie. good luck with the rest of Rakudo Day!
jnthn masak: Dude, you'd think it was you living in the country that has today as national holiday. :-P 12:47
masak (I got halfway hacking in '.=flip' last night. will try again later.)
jnthn Have lots of fun (sounds hard not to with that lot...)
masak jnthn: just working half-time, that's all.
masak bows, walking out backwards
12:47 masak left 12:48 iblechbot left
jnthn ruoso: (context vars) maybe, if pmichaud wants to do them today and wants my help on some bits related to them too, I'll happily join in with working on that 12:48
ruoso cool...
ruoso anxious to have faz working as expected...
jnthn :-) 12:49
ruoso which we then will probably be able to plug into masak's web stuff 12:50
jnthn, I think I already asked this before... but I think I forgot the answer... does rakudo already support defining custom grammars with interleave with Perl 6 grammar? 12:52
jnthn Such that we'd parse parts of our program differently? I don't think so. 12:53
ruoso maybe the inverser
*inverse
a grammar that has parts of it in Perl 6
i.e.: 12:54
12:54 nihiliad joined
ruoso controller Foo { action bar { perl 6 code here } }; 12:54
12:55 amoc joined
jnthn Oh, that might not be so hard. 12:57
ruoso hmm...
jnthn Thing is that it may not know when to terminate.
ruoso what do you mean?
jnthn That is, it may try and parse the last } } and say they are unmatched.
ruoso hmm...right
jnthn Rather than handing back over to your language.
ruoso maybe we could start from the method rule
jnthn True. 12:58
ruoso isntead of comp_unit
jnthn Aye, that may work.
ruoso is it possible to hand that code later to the emitter?
moritz_ rakudo: say "(foo)" ~~ m/\( ~ \) <Perl6::Grammar::identifier> / && say $/
p6eval rakudo dc680c: OUTPUT«(foo)␤1␤»
12:59 LylePerl joined
jnthn ruoso: You thinking of code-generating and eval'ing, in which case I guess what moritz_ just wrote... 13:00
ruoso jnthn, and what if I want to get that compiled down to bytecode?
jnthn $/.ast should give you the PAST, and I guess then you'd need to get hold of something that transforms that to bytecode. 13:01
Oh, you mean you want to eval it and then save it as bytecode?
13:02 payload left
ruoso jnthn, well... what I want is that code to be compiled regularly... 13:02
jnthn What do you mean by "regularly"? 13:03
ruoso as if it was regular Perl 6 code 13:04
actually... I think it could even be expressed as macros
moritz_ which are NYI, and quite a bit down the roadmap 13:06
jnthn Aye, macros won't be for a while.
ruoso hmm... maybe doing it as a different HLL in parrot?
13:07 kidd_ left
moritz_ I don't think it's that easy :( 13:07
jnthn Aye.
ruoso hmmm... any reasonable way of doing it?
jnthn ruoso: pmichaud is much more the person to ask about this stuff.
ruoso right... 13:08
moritz_ ruoso: I fear you have to set up the compiling environment etc.
ruoso: for now it's probably easiest to use string eval
jnthn I'm not familiar with PGE guts and I don't know how hard the "knowing when to stop parsing" issue is.
Or if that's a solved problem.
(I fear not though.)
As moritz_ said, the eval route - if you can get the parser to do what you want - should Just Work. 13:09
13:09 payload joined
jnthn Your other option is getting the PAST (I guess you need to somehow get the action grammar in place though). 13:09
moritz_ rakudo: say '(say "hi";)' ~~ /\( ~ \) <Perl6::Grammar::TOP> /
p6eval rakudo dc680c: OUTPUT«Syntax error at line 1, near ")"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)␤»
jnthn And building up PAST nodes from your other surrounding rules that call into Perl 6.
moritz_: Yeah, that's the issue I was expecting. 13:10
moritz_ jnthn: me too, but I had hopes for magic :-)
jnthn moritz_: I don't understand enough to know how tractable that is as a problem. It could be anywhere from "easy" to "ouch". :-)
ruoso rakudo: my $x = 'action foo { say "Hello!"; }'; $x ~~ / action <Perl6::Grammar::block> /; say $/.ast
p6eval rakudo dc680c: OUTPUT«␤»
ruoso rakudo: my $x = 'action foo { say "Hello!"; }'; $x ~~ / action <Perl6::Grammar::block> /; say $/; 13:11
p6eval rakudo dc680c: OUTPUT«␤»
moritz_ rafl: if block is a token or rule, it won't backtrack to search for it
ruoso rakudo: my $x = 'action foo { say "Hello!"; }'; $x ~~ / action { ~ } <Perl6::Grammar::statementlist> /; say $/;
moritz_ sorry, mean ruoso
p6eval rakudo dc680c: OUTPUT«Statement not terminated properly at line 1, near "~ } <Perl6"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)␤»
ruoso rakudo: my $x = 'action foo { say "Hello!"; }'; $x ~~ / action \{ ~ \} <Perl6::Grammar::statementlist> /; say $/; 13:12
p6eval rakudo dc680c: OUTPUT«␤»
ruoso rakudo: my $x = 'action foo { say "Hello!"; }'; $x ~~ / action \s foo \s \{ ~ \} <Perl6::Grammar::statementlist> /; say $/;
p6eval rakudo dc680c: OUTPUT«action foo { say "Hello!"; }␤»
jnthn aha! 13:13
rakudo++
moritz_ indeed
and pmichaud++, just to be sure :-)
jnthn And pmichaud++ too.
oh :-)
Well, karma deserved.
jnthn thinks he's got the two tricky RT tickets on role composition errors fixed 13:14
I've put those off for weeks...turned out not to be quite so bad after all.
moritz_: Did you say you had problems with Rakudo + latest Parrot? 13:15
ruoso I guess writing a DSL compiler in rakudo to be ran in this files is a possible way of doing it
compiling from that to Perl 6 13:16
moritz_ jnthn: yes, two or three segfaults (one known, slice.t)
PhatEddy A proposed rakudo independent workaround to spec tests checking for Null PMC: regex fairly_conclusive_platform_error {:i \N*<<Null?>>} 13:17
moritz_ jnthn: but I didn't do realclean in Parrot
jnthn moritz_: slice.t is a known, yeah
moritz_: I did some analysis on it and field a ticket with details...no ideas for a solution yet though. 13:19
moritz_ jnthn: I know, I've seen it
13:34 sri_kraih joined 13:36 exodist joined 13:40 sri_kraih_ left
dalek kudo: 14cd976 | jnthn++ | src/ (2 files):
Refactor the way we handle the composition of role cronies to be more in line with the spec. Fixes RT#63330 and almost fixes RT#63332 (that needs us bumping up to a fixed Parrot).
 13:45
pugs_svn r26724 | jnthn++ | [t/spec] Unfudge test for RT#63330. 13:55
13:55 hcchien left, r0bby left, betterworld left, finanalyst left 13:56 hcchien joined
ruoso pmurias, hi 13:58
13:58 r0bby joined 14:00 iblechbot joined, twigil joined, Casan left 14:01 betterworld joined
pmurias ruoso: hi 14:04
ruoso I'm working on getting p6opaque running
PhatEddy giving one more chance to review my package test file (20 tests) before I commit/add: gist.github.com/108785 14:06
pmurias ruoso: ok
pugs_svn r26725 | ruoso++ | [smop][nagc] provides a version of the release code that doesnt call "free" for objects that need a more complex destruction process. 14:11
r26725 | ruoso++ | [smop][tools/ri] provides a nagc.nofree attribute for choosing the more complex destruction
r26725 | ruoso++ | [smop][p6opaque] uses the nagc.nofree option
r26725 | ruoso++ | [smop][p6opaque] placeholder for dispatching the actual DESTROYALL method
ruoso pmurias, now we just need to figure out a way to dispatch DESTROYALL on the object and still free the object later...
jnthn rakudo: multi foo(Bool :$baz = Bool::False, *@vals) { say "foo" }; foo(:baz(Bool::True), 1, 2, 3); 14:12
p6eval rakudo 14cd97: OUTPUT«No applicable candidates found to dispatch to for 'foo'␤current instr.: '_block14' pc 91 (EVAL_18:57)␤»
14:13 skids joined
pmurias we could have a LOST frame (factory) which would call DESTROYALL on an object and the free it 14:13
ruoso pmurias, yeah... I was thinking about it also.. 14:14
14:15 PacoLinux joined
pmurias ruoso: should i write it? 14:15
14:16 jan_ left
ruoso pmurias, please... I'm still getting the last pieces of p6opaque together 14:16
pugs_svn r26726 | jnthn++ | [t/spec] Unfudge another now-passing role composition test after bug fixes.
pmichaud good morning perl6 14:19
dalek kudo: 52419be | jnthn++ | build/PARROT_REVISION:
Bump us up to Parrot revision with the string comparison to determine class identity in inheritance ripped out.
pmichaud 52419be++!!!
pmurias ruoso: how should the module which does that be called? 14:20
jnthn pmichaud: Yeah, I finally got onto two of the roles composition bugs I've had in my queue for ages, and found this was part of it...
ruoso pmurias, I'm not sure it needs to be a module
pmurias, maybe just a part of the p6opaque module
pmurias yes, that would be better 14:21
pmichaud reads a couple of days of scrollback 14:24
jnthn ouch.
ruoso ouch indeed
pugs_svn r26727 | ruoso++ | [smop][nagc] add smop_nagc_free as a visible function. 14:29
r26727 | ruoso++ | [smop][tools/ri] call smop_nagc_free in the end of DESTROYALL if nagc.nofree is set - the idea is that you should return from DESTROYALL to interrupt the process and call it again when you really want it to be destroyed...
r26727 | ruoso++ | [smop][p6opaque] the actual dispatching was missing...
PerlJam Have any of you other language people seen this? msdn.microsoft.com/en-us/oslo/dd548667.aspx 14:30
That M language looks an awful lot like Perl 6 Regex
(well, the syntax is completely different, but the same semantics appear to be there) 14:31
14:33 payload left
PerlJam the more I look at it, the more Perl 6 I see. 14:34
ruoso it really look a lot like it...
14:36 [particle] left 14:39 mizioumt joined
pmurias svn-- # insanity 14:40
Matt-W Perl 6 doens't have interleave patterns 14:41
which seem liable to induce headaches
but then, parsing always does anyway
PerlJam Matt-W: yes it does.
interleave is <ws>
syntax + interleave in M is the same as a rule in p6
Matt-W hmm 14:42
pmurias ruoso: what's the point of not setting nagc.nofree?
Matt-W I still don't understand the differences between regex, rule and token
ruoso pmurias, of not setting? 14:43
pmurias yes
Matt-W I just use regex for everything because I know it's not goign to do anything weird
ruoso pmurias, what do you mean?
PerlJam Matt-W: regex is just "normal" regex, token is regex with no backtracking, rule is regex with not backtracking and with implicit whitespace
Matt-W hmm
implicit whitespace where 14:44
pmurias if the nagc.nofree is not set the object call smop_nagc_free itself and free doesn't get called from RELEASE
14:44 twigil left
PerlJam Matt-W: whitespace in your pattern matches (smartly) whitespace in the input 14:44
pmurias s/is not set/is set/
* calls
ruoso the idea is that you interrupt DESTROYALL do whatever you need and call it again 14:45
so you take advantage from the generated destruction code for attributes and stuff
14:46 jiing joined
ruoso pmurias, note the "return" in p6oaque DESTROYALL cod 14:46
PerlJam Matt-W: regex foo { "a" "b" } only matches "ab". rule foo { "a" "b" } matches " a b " or "a b" but not "ab"
14:46 FurnaceBoy joined
pmurias ruoso: what i mean it that we could always make DESTROYALL in charge of calling smop_nagc_free 14:49
s/it/is/
ruoso pmurias, yeah... we could... but now you can choose when you set up the RI 14:50
Matt-W PerlJam: okay 14:54
14:56 Eevee joined
PerlJam Hmm. 15:02
rakudo: rule foo { 'a' 'b' }
p6eval rakudo 52419b: ( no output )
PerlJam rakudo: rule foo { "a" "b" }
p6eval rakudo 52419b: OUTPUT«Statement not terminated properly at line 1, near "a\" \"b\" }"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)␤»
jnthn thought double quotes were implemented recently? 15:03
pmurias ruoso: i have to catch a train, so i'll write the LOST frame much later today/tommorow
pmichaud implemented but patch not applied yet.
It's on my todo list for today.
jnthn Ah, ok.
pmurias but having switched to git-svn should avoid getting sidetracked by version control client bugs
jnthn pmichaud: It's national holiday here and I've been invited out for dinner, so going to go do that soon. Will be back later on today though for more Rakudo stuff.
pmichaud jnthn: that's fine. I'm not sure what I'm working on today -- have a few things to catch up on from yesterday. 15:04
Happy $national-holiday though :-)
jnthn Yeah, I'm not entirely sure what this one's for.
But the sun is shining and there's opportunity to enjoy food and a beer or two outside. :-)
15:05 donaldh left
jnthn Just spectesting then hopefully applying a fix for 63920 first. 15:06
([TODO] default type for params in pointy blocks should be Object) 15:07
pugs_svn r26728 | ron++ | Add S10-packages/basic.t. Currently 20 tests including tests for 14 active RTs.
15:09 PhatEddy left, PhatEddy_ joined, PhatEddy_ is now known as PhatEddy
pmichaud jnthn: (Parrot uses strings for testing class membership) -- I've been carping about that particular design misfeature for months. :-) 15:10
15:10 justatheory joined
pmichaud not only is it wrong, but it's slow, too. 15:10
15:10 mikehh joined
jnthn pmichaud: I literally replaced a string comp with a pointer comp. 15:10
pmichaud that's far more correct, yes. :-) 15:11
and what I would expect.
jnthn All Parrot tests passed, another Rakudo test passed...
Added a Parrot test too.
pmichaud there are more instances of Parrot doing that in the code, iirc.
as in, a lot more instances.
jnthn awww...I broke some of autothreding.t...
Yeah, I know.
15:12 pmurias left, meteorjay joined
pmichaud (from scrollback) -- to compile a Perl 6 string, as opposed to eval'ing it, in Rakudo one can do 15:15
my $code = Perl6::Compiler.compiler('...');
rakudo: my $code = Perl6::Compiler.compile('say <hello>'); $code()
p6eval rakudo 52419b: OUTPUT«hello␤»
moritz_ hilights ruoso 15:16
pmichaud I _think_ it might also work to pass a Perl6::Grammar parse tree there (to avoid the duplicate parse), but I'm not sure that works yet.
actually, I suspect it doesn't. 15:17
15:17 FurnaceBoy left
moritz_ currently tests the patch in RT #64868 15:17
looks mostly good so far
though I might upper-case some warnings
pmichaud reviewing patch 15:18
moritz_ the `system(qw(make clean))' 15:19
bothers me
not every make utility is called `make'
jnthn Mine sure ain't.
pmichaud agreed.
I suspect the "make clean" should be done as part of Rakudo. 15:20
moritz_ but changing that $config{'make'} should work
pmichaud yes, that would help
s/Rakudo/Configure.pl/ above
I don't want gen-parrot to do the rakudo "make clean" -- it's mixing pieces the wrong way.
moritz_ the problem with sticking it into the Makefile is that it'll fail epically on parallel builds 15:21
at least I think so
pmichaud so, Configure.pl then
15:21 Kisu left
pmichaud Oh. 15:22
I'd like this patch significantly refactored, actually.
15:22 Kisu joined
pmichaud "check-parrot-revision" belongs in gen_parrot.pl, not in Rakudo's Configure.pl 15:22
otherwise we end up with the parrot-version-checking code in two places. 15:23
moritz_ so do just want to always call gen_parrot.pl, potentially with options?
pmichaud (gen_parrot.pl already checks Parrot revisions, and does the handling of build/PARROT_REVISION)
yes
moritz_ we could also factor out a module
pmichaud I'm not sure that buys much
moritz_ you're the boss ;-) 15:24
pmichaud in some sense...
15:24 [particle] joined
PhatEddy that Perl6::Compiler.compile example from earlier core dumps for me on cygwin (but not native windows). anyone interested in verifying ... 15:24
moritz_ pmichaud: most of all I trust that you have much more experience with build systems than I have ;-)
pmichaud gen_parrot could become the generic "do all of the parrot-related stuff" tool.
I don't know about "much more", but I've written a couple. 15:25
jnthn PhatEddy: Quite believable, we do have other issues on cygwin I think.
15:25 alanhaggai_ joined
moritz_ as opposed to 0 ;-) 15:26
pmichaud oh, I think I know what I want. 15:27
I think I want my own "get the parrot config" wrapper.
that also optionally builds and tests the parrot version
and returns that result back to Configure.pl, which then generates the makefile.
(this would extend/replace the existing gen_parrot.pl)
that way Configure.pl doesn't have to worry about "how do I get/build Parrot" -- that's handled by the other wrapper. 15:28
moritz_ and if you want to make the "return" not a string, but a data structure, it will be in a module
15:28 zamolxes joined
moritz_ that's where my idea came from 15:28
pmichaud sure
but really the "get the parrot config" wrapper only needs to return the path to the correct 'parrot_config' to use.
TimToady rehai 15:29
pmichaud and Configure.pl simply call it for the configuration information it needs.
good morning, TimToady
jnthn: I saw the earlier conversation about "type methods" -- where did that end up? 15:30
TimToady I've been thinking about that a lot
jnthn pmichaud: It ended up with my brain exploding.
pmichaud I'm curious how one writes "class methods" if we do that.
15:31 alanhaggai_ is now known as alanhaggai
jnthn Depends what sort of class methods you mean. 15:31
pmichaud The kind where I can have Int.foo and 3.foo both invoke the same method.
15:31 clintongormley joined
jnthn Anyway, I'd also missed out on something fairly major in the way smop does things. 15:31
TimToady they used a fallback to normal dispatch for that, but I'm not sure I like that
pmichaud TimToady: which one falls back, though?
I thought the fallback was backwards there.
TimToady exactly
jnthn Which basically called a halt fo the refactoring I was starting digging in to. 15:32
[particle] role, role, role your boat...
TimToady on the other hand, what motivated it was ACCEPTS, which needs to check validity first
pmichaud well, it's not just ACCEPTS, at least to my view.
we also have various postcircumfixes and other method-like thingies
jnthn Indeed. Hash{ ... } and %foo{ ... } are different. 15:33
pmichaud there are a number of ways in which a type object responds differently from a normal instance, and I'm not sure that all of them are syntactic.
jnthn ACCEPTS is sure not syntactic.
Anyway what I've essentially figured is that smop doesn't really do the "type objects and instances are the same type" any more literally than we currently do it in Rakudo. 15:34
Differently, but still...
pmichaud Yes, I often felt it was the to-may-to to-mah-to sort of thing. We just pushed the "are you a type" bit into different places.
jnthn I'm not convinced type object = empty instance + role mixed in is a wrong way to do it. 15:35
pmichaud Rakudo chose to distinguish types by a role. Which really is a form of "are you a type bit" at its core.
jnthn *nod*
pmichaud The question is whether using a role brings along other baggage that is fundamentally incorrect.
TimToady part of the problem is that we're using type to mean several different things here
jnthn We need to be a little careful where this gets us on representation vs metaclass.
pmichaud or perhaps not "whether", but "to what degree" it brings along other baggage that we then have to fixup or work around. 15:36
yes, in jnthn's earlier list of things that a "type" is used for, I didn't see "package namespace handle"
TimToady when we say ~~ $x we're using a type as a subset meaning "all the possible values"
not as a type meaning "no valid value" 15:37
pmichaud okay. I was just checking to see if I followed the conversation, it appears that I did. 15:38
I don't have any new insights to add right now beyond what I've just written (or things I've written previously).
15:39 jan_ joined
TimToady I've thought in various directions on it, but not entirely happy with any of those yet 15:39
but my gut feeling is that there's something going on with values being subset types
jnthn OK, I need to go for a bit. Be back later.
pmichaud jnthn: sure, thanks.
TimToady laters 15:40
I've also been working over the notion of language braids
pmichaud TimToady: I think there's something there also, but I haven't been able to get beyond the "it sure seems like..." proposition to anything more concrete that fits well with the rest of what I know of Perl 6 (implementation)
TimToady but I'm wary of dispatchers that reorder 15:41
pmichaud agreed.
TimToady the subset idea leads me to think that it's more like eligible vs ineligible, keeping the same order 15:42
15:42 nihiliad left
pmichaud I tend to fall back to "if I have X and Y of the same type but X responds differently to various method-like things, then X must have an additional role, trait, or other thing applied that causes it to act differently" 15:42
15:42 nihiliad joined
TimToady but that also gets tangled 15:42
there's a figure ground problem with that
since roles are always added in
pmichaud agreed.
TimToady but sometime you want to see the concrete and sometimes the abstract as the figure 15:43
or both, for $x.new
pmichaud (and other "class methods")
TimToady well, that's a bit of a mislabel too, if class method means +abstract-concrete 15:44
pmichaud I think of "class method" in the traditional sense, where it can be invoked by either the abstract or concrete instances
TimToady currently our methods are ++, but maybe we need "meta" "metamethod" and "method", or some such
pmichaud (at least, that's "traditional" in my java and C++ version of OO stuff)
15:45 nihiliad left
TimToady meta meaning "I can handle an invalid value" 15:45
pmichaud hmmmm
TimToady and "method" meaning I can handle an ordinary value
a valid valud
pmichaud that has some good promise, I think.
TimToady or something like that--seems too important to let traits deal with it 15:46
pmichaud at first look, I kind of like the meta, metamethod, and method distinction
that feels... "clean"
TimToady well, that's the direction I'm leaning, anyway, and it defaults normal methods to not seeing type objets
[particle] would there be syntactic difference for calling those?
pmichaud [particle]1: for defining them, I think. 15:47
TimToady no
pmichaud calling would be the same.
Type.foo would call the "foo"'s that are meta or metamethod
$obj.foo would call the foo's that are metamethod or method
[particle] ok, so no self!method or !^meta kinda stuff
pmichaud (assuming that $obj isn't a type object)
TimToady no, this is trying to get the semantics of Plain Old Dispatch correct 15:48
[particle] can you replace 'meta' with 'type'?
pmichaud having those as declarators would also mean that it becomes much clearer about when a method would be valid on a type object (as opposed to an instance)
[particle] meta is so generic
maybe typemeta, typemethod 15:49
TimToady type isn't any better
[particle] well, we have type objects
at least it's not protomethod...
TimToady besides, we're labeling a method, and "type" would be misleading
since ACCEPTS isn't a type
pmichaud i.e., if I say method foo() { say $.x; } then it's pretty clear that Type.foo would result in a "method not found" error, instead of something where we have to explain that Type is not an instance 15:50
(and detect that because there's not an instantiated self)
if I want Type.foo to also work, I'd have to declare it as a metamethod
[particle] so, these are decorated upon definition
how does only/multi factor in? 15:51
15:51 Psyche^ joined
TimToady orthogonal 15:51
[particle] only metamethod foo {...}; multi method foo {...}; #okay
pmichaud anyway, from a definition/clarity side, I like it. I'm not entirely sure what it implies for dispatch (either Perl 6 or Parrot) 15:52
TimToady only/multi is just saying whether we overload the name in the current namespace
this is more like an extra bit on the binding constraints 15:53
pmichaud (I should've used "$!x" in my example above, sorry.)
[particle] so, my example wouldn't be okay, then.
since the metamethod and method of the same name have the same sig, and one is marked only 15:54
TimToady correct, you can't do only foo vs multi foo
[particle] sensical.
TimToady multi meta foo vs multi method foo is an interesting thing though, especially when you mention &foo to mean either 15:55
[particle] is it a junction, then? 15:56
pmichaud okay, my brane exploded.
15:56 zamolxes left
pmichaud (it was about to do that anyway -- I'm trying to hold too many projects in at once) 15:56
TimToady well, we'll leave that one to cook for a bit longer
clintongormley welcomes pmichaud to the other side of the brane
TimToady another brane exploder is the language braid thing
a braid is really its own little namespace running through a compilation 15:57
[particle] i've got to figure out this braid thing...
TimToady I'd been emulating a braid with separate variables
[particle] consults 6perl notes
TimToady $?Q vs $?Regex vs $?Quasi etc
but then there's no way to treat them as a group
it's not a package namespace though, more like a hash in the compiler 15:58
so instead of identifying sublangs as $?Regex we'll probably given strands of the braid their own twigil namespace 15:59
probably $~Regex means the currently defined regex language, for instance
pmichaud at first instance that also sounds sane to me
TimToady and we also get a new declarator to talk about that
[particle] a group meaning the current working set of DSLs?
TimToady yes
[particle] ok.
TimToady and since it has its own namespace that is specifcally *not* a package 16:00
[particle] there's no ascii char that looks like a braid, save $
TimToady we can use augment/supersede to mix into the current DSL
so we get things like
pmichaud I'm not sure that twigil needs to be ascii.
TimToady I though ~ looks like a strand of a braid, is all
[particle] all twigils need ascii variants imo
TimToady *thought
things like: 16:01
[particle] yes, ~'s possible. # is a weave
^ is a toupee
TimToady augment slang Regex { token backslash:sym<Y> {...} }
and I like slang as an abbrev for sublanguage
[particle] slang++
TimToady because slang is a subset of the current language, not a lang of its own 16:02
and it is invented for temporary use
and may pass out of use after this era
[particle] as opposed to creole
moritz_
.oO( you can solve any problem by inventing more syntax for it )
TimToady yes, all the other words for lingos and such imply an entire language
more than they do a subset of the current language
pmichaud ooc, where does "dialect" fit in? 16:03
TimToady well, there are linguistic terms, but they're unworkable, like "register"
a dialect is also a complete patois
[particle] dialect isn't a subset.
pmichaud okay.
TimToady *patoís (sp?)
pmichaud it will be interesting to see when things graduate from being "slang" to being "official Perl 6" :-) 16:04
TimToady anyway, I think slang will do
yes, it works well culturally as well
pmichaud (yes, I know, that all slang is officially Perl 6...)
ruoso back from lunch
TimToady it's really the right word
pmichaud agreed.
ruoso TimToady, the problem with using the name 'meta' is that it wrongly points to the metaclass (HOW)
clintongormley the difference between a language and a dialect? an army and a navy
[particle] std is the perl equivalent of oed.
moritz_ lisp has s-expressions, perl has s-languages ;-) 16:05
ruoso TimToady, this is not a meta method, because meta methods are the one defined in the metaclass already
pmichaud actually, I think std is more like "the Queen's English"
TimToady and then we get to use the augment/supersede distinction to decide whether we're mixing in the slang or replacing the strand entirely
ruoso TimToady, they are simply methods that are treated differently, as submethods are 16:06
TimToady ruoso: but it's semi-meta in that it's primarily used as a carrier of .HOW information
ruoso not really...
TimToady well, I have other names up my sleeve too
ruoso specially in the way rakudo is pointing now (which is to work as smop)
TimToady a generic method is a policy 16:07
so we could have policy, policymethod, and method
16:07 Patterner left, Psyche^ is now known as Patterner
ruoso we're now pointing to storing the methods in the type object, and no longer in the HOW 16:07
so the HOW is generic to all things that behave like that
i.e. ClassHOW
pmichaud ruoso: note that jnthn++ may be reconsidering rakudo's smop-ness 16:08
or, more precisely, that rakudo today may not be so far away from smop as initially thought
ruoso but even then... the HOW defines how things in general work 16:09
i.e. how classes work
how roles work
metamethods are part of the HOW protocol
pmichaud 15:34 <jnthn> Anyway what I've essentially figured is that smop doesn't really do the "type objects and instances are the same type" any more literally than we currently do it in Rakudo.
TimToady though getting people to write "policy new" will be an interesting trick
and requiring "policymethod" for something that responds either way will certainly discourage that practice 16:10
ruoso whispers "static"
16:10 iblechbot left
TimToady shoots ruoso after carefully installing the silencer 16:10
pmichaud thinks he might've heard something, somewhere. Probably not. 16:11
moritz_ enjoys the silence in the wood, where no tree has fallen
[particle] virtual silence.
ruoso TimToady, I guess they are not really methods 16:12
pmichaud this would end up being interesting from a PGE/Rakudo perspective, though, where most tokens and rules expect to be handled more as "policymethods" than "methods"
ruoso since a method is something you call on the object
TimToady I guess I missed 16:13
[particle] bullets for sale!
ruoso bullet-proof-jacket++ 16:14
TimToady maybe method is the middle one, and we need a concrete name
[particle] landmine new {...}
moritz_ [particle]++
TimToady so abstraction vs method vs concretion
ruoso method being the middle one is cool
because that is what most perl users expect 16:15
16:15 Woody2143 left
TimToady there a madness to it 16:15
[particle] because that is what most perl users expect
16:15 Woody2143 joined
pmichaud it does mean that "method foo() { say $!x }" still has to figure out how to respond to Type.foo .... but I'd probably be able to live with that. 16:15
(since we're already looking at living with that in the present state of things) 16:16
ruoso just as in p5
TimToady it just sets both notional bits, I respond to abstract/concrete calls
16:16 iblechbot joined
pmichaud do we have abstract/concrete submethods? 16:17
or does that not make any sense?
TimToady submethods are particularly about partial objects 16:18
and partially valid objects
ruoso it's another axis
TimToady they're infrastructure
not ultrastructure
ruoso feels a glimpse of marxism...
[particle] hyperdimensional spaces are groovy
pmichaud [particle]1: I always thought of them as being more stringy :-) 16:19
TimToady THE FROOT FLAIS ARE COMMING!!!
pmichaud PLZ TO KEEP THE FROOT FLAIS AWAY FROM MY FROOT LOOPS.
moritz_ what are we doing today, then? string theory or science?;-) 16:20
TimToady the two are not mutually inclusive
moritz_ no, but only rarely overlapping, it seems 16:21
16:21 a3r0 joined 16:22 bejuryu joined
TimToady let's throw out method, and go for solid/liquid/gas, and for the really academic type stuff, plasma 16:23
liquid new () {...}
gas ACCEPTS
triplepoint WHAT 16:24
moritz_ and if we want more aggregate types, we can revert to BEC (bose einstein condensate)
TimToady bose BUILD
or maybe more like quantumfoam BUILD 16:25
16:26 rindolf joined
TimToady and we measure everything with units of ℎ instead of bits 16:26
ruoso opens a wormhole to bring everyone back to the point 16:27
pugs_svn r26729 | lwall++ | [STD] add in slang declarator and twigil 16:28
TimToady feels singularly good
16:28 nihiliad joined
[particle] timtoady is a boson 16:29
...just when you think the spec has settled down... 16:30
TimToady that's obvious to the most casual observer, oh drat...!@#$!(@*#($*(!@*#$*#(!$@#$
I haven't specced it yet, just implemented it. :)
[particle] ...along comes an implementation to prove that wrong. 16:31
TimToady you shouldn't have observed the design until you were in the correct universe
ruoso considers this is not a change with huge impact 16:33
the change in the way capture works was much heavier
TimToady which change are you referring to, dispatch or slangs? 16:34
ruoso dispatch
16:34 clintongormley left
TimToady well, the real design question is whether method should default to both or to concrete only 16:35
ruoso IMNSHO, method should be both
TimToady and it's in the defaults where we'll make most of our "new mistakes" 16:36
ruoso Perl developers are already used for methods to have this dual-life
[particle] perl 6 has proven again and again that perl developers are often wrong
TimToady and swear at it occasionally
[particle] although perl 6 hasn't proven that perl 6 developers are often right 16:37
TimToady anyway, "P5 does it that way" is not a strong argument around here these days :)
ruoso even in Perl 6 we use this dual life...
TimToady What Would Perl 5 Do?
ruoso $x.new
TimToady yes, but what if that ability had to be requested specially? 16:38
ruoso the point is that the code in new can work with both an abstract and a concrete object
TimToady and would setting the default more strictly help the optimizer 16:39
yes, I'm not arguing with that
the question is whether that should be the defalut
*default
16:40 kst joined
TimToady is it worthwhile to say that a normal method can never be called with an invalid object, to put it another way 16:40
in a Haskellian sense, do we build in Just typing a little harder than we have 16:41
ruoso TimToady, point taken
ruoso .oO(what is a slang anyway?) 16:42
TimToady backlog
about 50 minutes ago 16:43
just after pmichaud's brane exploded
ruoso foudn 16:44
*found
TimToady but the short answer, a current sublanguage such as Q or Regex 16:45
ruoso so, we agree that there should be the different dispatching for type objects vs instances 16:46
and instead of fallbacks, we have methods that are visible in both lookups
16:47 samlh left
TimToady we agree that the dispatcher should be able to restrict itself to abstract or concrete instances when it wants to 16:47
that's not quite right either
16:47 a3r0 left
ruoso it's the method that chooses when to be seen, not the object... right? 16:48
TimToady an instance's valid bit may be used by some candidates to disqualify themselves
yes
16:49 iblechbot left
TimToady that being said, an optimizer might refactor that into multiple candidate lists 16:49
ruoso but does that only applies to multi methods?
I think that conceptually you have two candidate lists from start 16:50
TimToady multi is orthogonal
ruoso right...
TimToady no, if you have only two lists, you cna't easily dispatch to both at once
ruoso it's just that candidate is usually used in the context of the multi dispatcher
TimToady I use "candidate" for much more than that 16:51
ruoso right...
TimToady all the single dipsatch candidates are also in a list
wrapped routines are in a candidate list
anything where "nextsame" means "Ignore me, keep trying" 16:52
ruoso right
TimToady again, only/multi is namespace management more than candidate list management 16:53
ruoso Ok... 16:54
TimToady whether the name is a Highlander Name 16:55
or whether we need to consider longnames
ruoso so... I think it's a matter of visibility...
TimToady or some other form of non-uniqueness
16:55 cdarroch joined
ruoso we have methods that are visible only to abstract objects, only to instances, and to both 16:55
it's in an axis near the my/has/our axis 16:56
16:56 alanhaggai left
ruoso but it's still a different axis 16:56
TimToady anyway, still thinking about defaults and huffman 16:57
pmichaud lunch.
TimToady biab & 16:58
17:14 PhatEddy left 17:27 cavelife^ joined
TimToady it's starting to feel a lot more like a constraint on the invocant to me 17:29
(which could theoretically be a constraint on any parameter)
so whatever the notation, it desugars to ($self where { .STATE === BUILT }: ...) or some such 17:30
assuming .STATE is a .WHAT-like macro
or maybe it's .WHEN, as in when are we in this object's lifecycle 17:31
and there could be more state than just BUILT vs UNBUILT
we could say, for instance, that only submethods may be called on objects in BUILDING or DESTROYING state 17:32
ruoso right... but that would require a multi for you to have different implementations for the two states
17:33 DemoPhreak joined
TimToady which is why it's orthogonal 17:33
and why it's really a sig constraint
ruoso but I mean...
you might want to implement that on "only" methods
TimToady that?
ruoso different versions of the method, one for BUILT and other for UNBUILT 17:34
TimToady then you're distinguishin on the sig, and it's a multi
ruoso I see that's what you meant 17:35
TimToady I think we could use a concise syntax for saying that a given parameter must be built
use on an invocate is a degenerate case then 17:36
*invocant
unfortunately we've already used ? and ! in parameters
ruoso I think we shouldn't promote asking that instead of .defined for other objects 17:37
TimToady maybe defined === built
ruoso by default it is
TimToady and we don't give people a choce
*choice
ruoso it already is
hmm... 17:38
being able to override was precisely the reason I said that we shouldn't promote it
but my point is that checking the invocant of a method is not intrusive.... 17:39
checking other parameters seems intrusive
TimToady how is it any more intrusive than any other constraint? 17:40
I don't understand the harm
ruoso it's not the constraint that is intrusive
is asking something that can't be overriden that is
TimToady something has to be primitive; ain't gonna apologize for that 17:41
[particle] can you replace the dispatcher?
TimToady certainly
ruoso TimToady, the point is that if we restrict that checking to the invocant 17:42
that checking can be made in the dispatcher
TimToady you do it every time you say .* instead of .
ruoso which is in the HOW
TimToady the dispatcher already throws out candidates that cannot bind
ruoso and the HOW is the one that talks to the primitives
TimToady if the dispatcher pays attention to this constraint, it's an *optimization* 17:43
ruoso my point is about keeping that kind of check in the object's HOW
ok... let me take one step back... I'm actually wondering if doing that in the signature is actually a good idea...
when the HOW is already using primitives and is able to do it fairly easily 17:45
we would only need a way for the method to define its visibility concerning the invocant state 17:46
17:48 DemoFreak left
TimToady a dispatcher can hoist any declared (immutable) constraints out of the declaration for optimization, but I still think the right place to hang a constraint on a parameter is in the sig, in general 17:48
and the invocant is an ordinary parameter these days 17:49
ruoso it is, but it isn't
TimToady it's only special to the dispatcher
ruoso it is in the capture with no special handling for it
exactly
TimToady ^^
ruoso that's my point
TimToady so optimizing outside the dispatcher is *premature* 17:50
ruoso sorry... i misread your previous line 17:51
it's not special only to the dispatcher
it's special in the sense that it's already extracted from the capture before sending to the dispatcher
TimToady no
ruoso it needs to be, in order to find the HOW
in order to call the dispatcher 17:52
(extracted, but not removed
TimToady we're calling two different things the dispatcher here, I think
ruoso maybe
we have the dotty op
which knows how to call the dispatcher 17:53
and we have the dispatcher
which resides in the HOW
so... 17:54
$foo.bar
is actually
$foo.^dispatch($foo, 'bar', \($foo))
sorry
$foo.HOW.dispatch($foo, 'bar', \($foo))
or...
$foo.^dispatch('bar', \($foo))
but the last one is really a syntax sugar for the longer 17:55
TimToady that's still the outer dispatcher 17:56
ruoso exactly...
the dispatcher itself is in the HOW
TimToady there's an inner dispatcher loop that the outer dispatcher can factor things out of
like which candidates might match
ruoso hm?
which inner dispatcher you mean
?
TimToady the loop that nextsame interates
*iterates 17:57
ruoso I was considering that to be inside the dispatch metamethod
TimToady which is the same for all dispatchers
ruoso nextsame just throws a control exception that the metamethod knows how to handle... (or something like that) 17:58
TimToady $obj.@candidates is calling the low-level dispatcher, I think 17:59
ruoso hmm... 18:00
TimToady so you can view any high-level dispatcher as setting up @candidates and calling that
ruoso I always thougth that as invoking subs as methods...
hmm...
TimToady that's what a low-level dispatcher does!
methods as subs, rather
ruoso you mean $obj.@candidates enables nextsame?
TimToady yes 18:01
18:01 alester joined
ruoso hmm... 18:01
TimToady it's the minimal dispatcher loop over a (lazy) list that handles nextsame
ruoso I wasn't aware of that
TimToady with the candidate list generation factored out
that's the intent
ruoso I thought it was just a syntax sugar 18:02
TimToady outside of that is some kind of WALK thing
that generates @candidates
it's notional, of course, and a given outer dispatcher can cheat
but fundamentally, here's a list of candidates, try calling them in order till one works
ruoso hmmm 18:03
didn't $foo.@candidates mean call all of @candidates?
TimToady and the fundamential difference between single and multi dispatch is where we look for the outer dispatcher
that would be .*@candidates
a different fundamental loop 18:04
which is missing a "last" after the first success
ruoso ok... 18:05
anyway... that doesn't change my first question
TimToady we'll have to ask your mama what your first question was
ruoso ;)
18:06 nbrown_ joined
ruoso the HOW can build a list according to the state of the invocant 18:06
I mean... it doesn't need to be in the signature 18:07
it can simply be an attribute of the method, not of the method's signature
the visibility of the method, according to the invocant's state
TimToady and then what happens if you call it as a sub outside of a dispatcher, it suddenly allows bad data?
18:08 M_o_C joined
TimToady signature are already designed to apply constraints 18:08
ruoso there are a lot of bad things that can happen by using methods as subs anyway 18:09
TimToady duplicating that functionality is the job of an optimizer
that's why we distinguish the call syntaxes so that doesn't happen, so no argument
ruoso exactly... so... as long as you use methods as methods, everytihng is ok 18:10
even if the signature doesn't check it
if you use methods as subs, you should know what you're doing
TimToady if the optimizer can cheat without getting caught, that's fine :)
but the very dispatchers are relying on the fact that methods can be called as subs, in my view 18:11
18:11 iblechbot joined
ruoso but then can still be called as subs 18:11
s/then/they/
TimToady the *sub* dispatcher should not be penalized for the *method* dispatcher's cheating 18:12
the constraint should be enforced by the sig when the sub dispatcher call it
(but we have the same issue with other invocant type checks too) 18:13
ruoso alright... I rest my case ;)...
TimToady rakudo: class A { method x () { say "oops" }; method test { &x.(42) } }; 18:16
p6eval rakudo 52419b: ( no output )
TimToady rakudo: class A { method x () { say "oops" }; method test { &x.(42) } }; A.new.test
p6eval rakudo 52419b: OUTPUT«oops␤»
TimToady should probably fail the sub call
since 42 isn't an A 18:17
ruoso that's because the dispatcher is looking at the invocant ;) not the signature ;)
18:19 meppl joined
TimToady in pure type theory the A is a constraint on the invocant, and the single dispatch order merely falls out of that 18:19
so a sig test is redundant, but in type theory that doesn't matter
so I think the dispatcher doing part of the signature's work should be considered an optmization, presuming we can disable the redundant check in the sig 18:20
but the check is notionally part of the sig
every invocant is notionally constrained to the class it's declared in 18:21
and I think the sub call should enforce that too
18:21 mizioumt1 joined
TimToady unless the dispatcher specifically disables it because it already knows 18:22
ruoso alright... that's indeed a design decision...
TimToady bows and knocks his head on the desk 18:23
18:23 nbrown left, nbrown_ is now known as nbrown
ruoso hopes it didn't hurt 18:24
TimToady all that being said, the build status can be a special constraint; doesn't have to really be a role or a where
18:24 hudnix left
ruoso I think it doesn't even need to be .WHEN 18:24
but simply .BUILT
that returns true or false
TimToady well, the parent part of an object can think it's built before the child part thinks it 18:25
ruoso hm?
TimToady BUILDALL isn't atomic 18:26
when there's derivation
ruoso well... sure... but... inside buildall there is one operation that is
which is the one that initializes the instance storage
ah... ok
TimToady but that's when you start building, not when the data is there
ruoso I've been thinking of BUILT as "having instance storage"
TimToady it goes through states to get from empty to full 18:27
ruoso right...
TimToady and it's not clear which of those states might be intersting
ruoso actually..
it goes building from least-to-most specific
TimToady can you call a .BUILT method on the parent part of an object that is partly built, for instance
does each parent class keep its own "built" status? 18:29
ruoso but it shouldn't be considered built anyhow before the end
because a subclass might override
almost any behavior
TimToady should be able to rely on Liskov
ruoso so it should only be BUILT in the end of BUILDALL
TimToady the child's BUILD should be able to treat the already built object as a valid instance of the parent object, it would seem 18:30
otherwise it wouldn't matter what order we did the BUILDs 18:31
they don't necesarily have to track all their build states, as long as we know where the transition is
ruoso well... 18:32
actually...
18:33 rindolf left
TimToady one could almost view BUILD as taking a parent object and transmuting it into a "me" object 18:33
ruoso trying to figure out how that fits together 18:34
18:34 mizioumt left
TimToady so any virtual method in a BUILD actually starts looking in the parent class (assuming C3 for siblings) 18:34
ruoso geez... that looks uterly complicated 18:35
TimToady eh? the only difference from what we have now is the low-level blessing at the end of BUILD from the parent to the actual class
ruoso plus the C3 thing
TimToady that's already there 18:36
ruoso is it
?
TimToady implied in the BUILDALL semantics
how the heck are you going to do MI without constructing all the parents?
ruoso that's not the problem
TimToady and C3 is already mandated
ruoso the problem is preserving C3 semantics during BUIDL 18:37
TimToady I think of C3 as a way of making MI look like SI
that's all
ruoso in a A - B, C - D heart-shaped MI
we build in the following order
TimToady and for purposes of BUILDALL, that's all you need, I think
ruoso A, B, C, D
so,
in the end of A's BUILD it low-level blesses it as A 18:38
in the end of B it low-level blesses it as B
and has A as parent
in the end of C, it low-level blesses it as C
which has A as parent, and B is now out of the picture
TimToady alternately, the BUILDING state means trust anything that's not me 18:39
ruoso then, during D's BUILD
TimToady so we low-level bless into new class, but $.foo doesn't dispatch to me, only to all my parents 18:40
until BUILD is done
well, something like that.
we need to be able to define the new thing in terms of more primitive things 18:41
and it would be nice not to call anything that requires BUILT on the part of the object that ain't yet
ruoso maybe we just assume that we bless to the current class before calling BUILD 18:42
TimToady so probably three-way constraint, unbuilt, building, and built is sufficient for BUILDALL policy
ruoso: sure 18:43
18:43 hudnix joined
TimToady but that tends to make me want to say that by default "method" requires built 18:43
ruoso which would mean that BUILD sees the object already as its own
TimToady and we need to know more to allow on something that building 18:44
that is building
hmm, but then parent methods won't work, unless we know which class we're building
we can assume that if our method is not in the class that the object is already blessed into, 18:45
then that part is built
I think
so state == building implies an additional check to see if $obj.WHAT === $?CLASS 18:46
built state is "yes/no/maybe" :)
or yes/no/it depends
[particle] or abstract/method/concrete :) 18:47
TimToady yes/no/anyone but Joe
biab &
ruoso what about .BUILT(type) 18:49
18:52 alester left 18:53 alester joined
TimToady more like BUILDING == type, once it's BUILT we don't care, since ordinary inheritance is used to make type decisions 18:54
were you thinking of type as the $?CLASS, or the type being matched against $?CLASS 18:55
if .BUILT($?CLASS), then yes, if BUILDING state examines the parameter and the other states don't 18:56
gah, noon, and still haven't backlogged... 18:57
19:00 exodist left 19:01 amoc left 19:06 DemoFreak joined 19:12 ZuLuuuuuu joined 19:16 nbrown left 19:17 nbrown joined
PerlJam DHH mentioned perl 6 at railsconf 19:20
19:20 DemoPhreak left
PerlJam he didn't want rails 3 to "get sucked into perl6ism" 19:20
Infinoid we has a famous 19:21
19:24 Trashlord joined 19:31 ruoso left 20:03 nihiliad left 20:07 alester left 20:15 icwiener joined 20:25 alester joined
dalek kudo: a579f31 | pmichaud++ | docs/spectest-progress.csv:
spectest-progress.csv update: 383 files, 11032 passing, 0 failing
 20:37
20:42 nbrown_ joined 20:43 Trashlord left 20:59 nbrown left, nbrown_ is now known as nbrown 21:01 alialaasami joined 21:02 alialaasami left, bejuryu left 21:04 frew|work left 21:08 alester left 21:09 frew|work joined 21:14 payload joined 21:15 mberends left 21:19 Whiteknight joined 21:22 skids left
TimToady n 21:32
21:33 simcop2387 left 21:34 pmurias joined
jnthn is back 21:35
21:36 simcop2387 joined
jnthn omfg....OK, I read that backscroll tomorrow. 21:37
pmichaud good $national-holiday ?
jnthn pmichaud: Yes, in fact I had a nice evening with nice food and beer too.
Which was kinda what I needed.
pmichaud Excellent. Those are always good to have.
I'm working on some build system cleanups at the moment... getting ready to test + commit. 21:38
er, test + push.
jnthn didn't sleep too weel last night because he couldn't stop thinking about the Perl 6 object model and metaclasses and representation polymorphism and smop guts and Parrot guts and Rakudo guts and...uff. :-)
*well
Build system cleanup is a Good Thing. 21:39
pmichaud I didn't sleep well because other house occupants kept waking me. 21:40
:-)
jnthn Ah, that's a different problem. 21:41
:-)
jnthn doesn't have other house occupants, and it's both a good and a bad thing.
I feel like I've spent the week trying to work out how the whole metaclasses and representation stuff is going to play out, and I'm not convinced I'm much further forward than at the start of the week. :-( 21:43
21:44 simcop2387 left
pmichaud I think we clarified the issues a lot, though. 21:44
jnthn True.
pmichaud That's progress, even if it's not measurable directly in code yet. 21:45
jnthn Sure.
I think for a while I had the idea that smop had a lot more things worked out that it actually really does.
21:45 icwiener_ joined
jnthn Which isn't a criticism, just not quite how it felt before I dug deeper. 21:46
meppl good night
21:46 simcop2387 joined
jnthn I thought there was a lot more stuff I could take from there than I think we really can. 21:46
The model seems to be "the most derived class owns the representation of the attributes". 21:47
I don't think that's going to fly. 21:48
21:49 meppl left
pmichaud afk for a bit 21:49
21:52 donaldh joined 21:55 hudnix left 22:00 icwiener left
pugs_svn r26730 | jnthn++ | [t/spec] Unfudge block default parameter type tests for Rakudo. 22:05
22:07 mizioumt1 left, simcop2387-vnc joined
dalek kudo: f01d5d6 | jnthn++ | src/ (2 files):
Make the default parameter type of routines be Any, and the default type of pointy blocks and others be Object. Resolves RT#63920 and brings us in line with the spec.
 22:07
22:09 payload left 22:10 ZuLuuuuuu left 22:14 wknight8111 joined
jnthn pmichaud: I'm going to do more Rakudo stuff tomorrow rather than tonight. It'll work out better. :-) 22:15
Consider it a distributed Rakudo day.
22:16 Whiteknight left
pmichaud jnthn: that sounds good. I expect to be around most of the day tomorrow, also. 22:16
jnthn Cool.
pmichaud Unless I get called out for other household tasks (but I'm expecting not).
dalek kudo: 43f257d | pmichaud++ | :
Merge branch 'master' of [email@hidden.address]
 22:19
kudo: 1f4ec5d | pmichaud++ | Configure.pl:
Configure now verifies sufficient parrot revision (or dies).
22:19 M_o_C left 22:23 amoc joined, simcop2387 left
pugs_svn r26731 | lwall++ | [t/spec] fix an "is also" that is giving STD fits 22:27
22:28 DemoFreak left 22:32 frew|work left 22:34 DemoFreak joined 22:37 iblechbot left 22:40 pmurias left
pugs_svn r26732 | lwall++ | [STD] %*LANG now holds language braid 22:43
r26732 | lwall++ | unify $*PARSER with %*LANG<MAIN>
r26732 | lwall++ | don't complain on role redefinition
22:43 cdarroch left 22:56 icwiener_ left 22:58 skids joined 23:07 [particle] left, donaldh left
pugs_svn r26733 | lwall++ | [S02] take stab at documenting how interwoven languages work 23:24
23:28 [particle] joined
jnthn I'll have to re-read the backlog again tomorrow to fully get it 23:39
But something like a .STATE === BUILT kinda distinction sounds very sane to me. Best solution I've seen so far. 23:40
OK, sleep, night all
23:44 antiphase joined 23:46 kate21de 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!