»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg p6eval perl6: ... | irclog: irc.perl6.org/ | UTF-8 is our friend!
Set by sorear on 4 February 2011.
00:04 takadonet joined 00:15 tokuhiro_ joined 00:19 thelazydeveloper left 00:30 benabik left, benabik joined 00:32 adu joined 00:33 adu left 00:42 gongyiliao joined 01:02 Chillance left 01:23 Coleoid joined 01:49 drbean left 01:50 drbean joined 01:51 Transformer joined, Transformer left 02:05 orafu left 02:06 echo__ joined 02:07 dbr left, orafu joined 02:08 echo__ left 02:26 cognominal left 02:28 benabik left 02:39 Patterner left 02:40 benabik joined
skids r: role E [ $g = "foo" ] { method h ($g = $OUTER::g) { $g.say } }; class D does E { }; my D $d .= new(); $d.h; # how to do this without renaming one of the $g's ? 02:40
p6eval rakudo 843534: OUTPUT«Nominal type check failed for parameter '$g'; expected Any but got Mu instead␤ in method h at /tmp/u2Pvb_didm:1␤ in block at /tmp/u2Pvb_didm:1␤␤»
02:41 Psyche^ joined, Psyche^ is now known as Patterner
skids ...and what is the lexical "scope" of runtime parameter binding considered to be, that of inside the Callable? 02:42
moritz yes 02:44
well, starting from its declaration to the end of the block
you can say sub f($x, $y = $x * 2 ) { ... }
so $x is actually visible in the rest of the parameter list too
but you can't assign to $g in your example above 02:46
oh wait, I'm not fully awake
I think $OUTER::g should work
diakopter moritz: ur up very late or very early
moritz diakopter: yes. I'll try to sleep again :-) 02:47
02:57 xinming_ joined 03:00 xinming left 03:03 soooga joined 03:04 soooga left 03:06 soooga joined
skids moritz++ thanks. It would suck for cascading default values to need different names. 03:34
TimToady rosettacode.org/wiki/Sorting_algori...ort#Perl_6 03:51
03:54 kaare_ joined 04:01 birdwindupbird joined 04:06 birdwind1pbird joined 04:08 birdwindupbird left 04:10 kaxing left
quietfanatic Instead of: my @buckets = @list.classify( *.substr($r,1) ).sort: *.key; 04:10
why don't you say: my @buckets; @list.classify( *.substr($r,1) ).map: {@buckets[.key] = $_ } 04:11
04:11 kaxing joined
quietfanatic oh wait, you can't map a hash, can you 04:12
.pairs.map then.
but whatever
or just @list.map: { @buckets[.substr($r, 1)] = $_ } 04:13
or thereabouts
TimToady doesn't work for the '-' bucket 04:14
quietfanatic @list.map: { my $i = .substr($r, 1); @buckets[$i] = $i => $_ } 04:15
04:22 telex left 04:29 dalek left 04:30 telex joined, dalek joined, ChanServ sets mode: +v dalek, p6eval left, Util_ left 04:31 Util joined, masak left, dalek left 04:32 Juerd left, tadzik left 04:33 dalek joined, ChanServ sets mode: +v dalek, Juerd joined 04:34 p6eval joined, ChanServ sets mode: +v p6eval 04:35 Util left 04:36 Util joined 04:37 masak joined 04:46 crab2313 joined 04:49 UncleFester6 left 04:59 crab2313 left
TimToady rpn: sub foo (&infix:<op>, $a, $b) { say $a op $b }; foo &[*], 42, 7; 05:26
p6eval pugs: OUTPUT«*** ␤ Unexpected "op"␤ expecting operator, ":", "," or "}"␤ at /tmp/sfoybqgtL9 line 1, column 40␤»
..rakudo 843534: OUTPUT«===SORRY!===␤Unable to parse blockoid, couldn't find final '}' at line 2, near "say $a op "␤»
..niecza v19-15-g051783d: OUTPUT«294␤»
TimToady rpn: sub foo (&infix:<op>, $a, $b) { say $a op $b }; foo &[/], 42, 7;
p6eval pugs: OUTPUT«*** ␤ Unexpected "op"␤ expecting operator, ":", "," or "}"␤ at /tmp/k4bR28KwsJ line 1, column 40␤»
..niecza v19-15-g051783d: OUTPUT«6␤»
..rakudo 843534: OUTPUT«===SORRY!===␤Unable to parse blockoid, couldn't find final '}' at line 2, near "say $a op "␤»
TimToady coulda sworn binding to &infix:<op> worked already in rakudo
05:27 betterworld left, araujo left 05:35 kaxing left
TimToady anyway, rosettacode.org/wiki/Parsing/RPN_ca...thm#Perl_6 only works with niecza 05:36
05:37 raiph joined
raiph is Boyer-Moore relevant to a regex implementation? 05:43
en.wikipedia.org/wiki/Boyer%E2%80%9..._algorithm
relevant to P6's regex engine?
s/P6's/a P6/ 05:44
Timbus well, yeah. im pretty sure perl 5 uses that 05:45
you can use it to find initial matches and stuff 05:46
geekosaur remebers when ilyaz hooked BM into the perl5 regex engine 05:49
05:50 betterworld joined
raiph Timbus, geekosaur: thx 05:50
(now looking at ilyaz + bm + perl google) 05:51
geekosaur hm now I wonder if that predated perl5 even. been a long time ;} 05:52
05:53 risou_ left
raiph i recall reading about bm in the 80s. i think it was in an issue of scientific american. 05:54
05:54 jimmy1980 joined
geekosaur also "ilyaz" may not be appropriate for a google search because that was his cpan id but not necessarily mentioned in his email. full name: Ilya Zakharevich 05:55
(hah, remembered it correctly without needing to google...)
raiph geekosaur: memory's interesting. i haven't had reason to recall bm in over a decade, but thought to myself "boyer", i recall "boyer"... 05:57
geekosaur boyer-moore I remember, don't have to think about it much but it still comes up occasionally 05:58
raiph "string matching algorithms that are more efficient than DFA's (like Fast Boyer Moore) can be utilized [and then] the real engine used to verify that the match is valid. In many cases this can result in much faster match times than a DFA can provide, even when the verifier is an NFA with backtracking. " 05:59
(from www.perlmonks.org/?node_id=597262) 06:00
geekosaur interesting thing is, that's how BM was being used originally... although later developments probably complicated that a bit 06:07
and as pointed out later in that thread, in fact still does ;} 06:10
TimToady um, I put BM into the regex engine 06:11
Ilya may have improved it, of course 06:12
look for fbminstr in 4.036, for instance 06:30
06:44 tadzik joined
TimToady rosettacode.org/wiki/Remove_lines_f...ile#Perl_6 06:49
06:49 JimmyZ joined
JimmyZ r: say ( 600 - 490 ) / 600 * 100 , '%' 06:54
p6eval rakudo 843534: OUTPUT«18.333333%␤»
JimmyZ memory down to 490M in pirt branch 06:56
06:58 GlitchMr joined 07:02 ranguard_ left, ranguard joined
TimToady geekosaur: actually, now that I check, I had already added fbminstr in Perl 2.0 07:09
raiph TimToady++ # rosettacode 07:18
geekosaur TimToady, I know some form of it was there early, but Ilya did some work to integrate it better into regexen at one point. on the warpath over people using index() instead of regex matching because it was faster... 07:19
TimToady I put it in to find the initial match positions; iirc Ilya extended it to be able to look for the first thing after a .*? 07:20
since .*? is a lot like the initial scan that a regex does 07:22
don't want to take away from all the tweaking that Ilya did do, of course 07:23
geekosaur yeh, but that tweaking is the main thing I remember boyer-moore for with respect to perl 07:24
think there was a big bikeshed over it too for some silly reason
raiph TimToady: does a similar approach apply to P6 regex declarative prefixes or whatever you call the fixed bits?
TimToady geekosaur: the bikeshed was, I think, because it was a lot of work for diminishing returns, and it was data-dependent 07:26
filtering the initial match positions through fbm was already getting most of the benefit for most regexes 07:27
raiph: well, it could, for matches that scan, but grammars generally don't scan, but match right were the previous thing left off 07:28
certainly the engine tracks literal prefixes, so such an optimization be beneficial in some cases 07:29
though the optimization tends to work well only for longer constant strings 07:30
especially with the cache characteristics of modern CPUs, which will tend to have the next few characters handy
of course, some of the bikeshedding back then on p5p was just personality conflicts too... 07:32
raiph: but to answer your question, bm only applies to fixed strings, which is going to be a subset of declarative prefixes, which include anything you can write a DFA for 07:36
geekosaur yes, yes it was. I think that incident was a large part of why I bailed from p5p 07:37
TimToady ooh, when did it suddenly become tomorrow? 07:39
TimToady starts to wander off toward bed
geekosaur been tomorrow here for a while, my body is just being annoying 07:40
(so still up as 0400 approaches)
07:43 crab2313 joined 07:54 adu joined 07:55 fgomez left 08:03 SamuraiJack joined 08:34 sergot joined
sergot hi o/ 08:34
08:34 huf left 08:42 crab2313 left, NABUCFR joined 08:58 sftp left 09:00 sftp joined 09:02 MayDaniel joined 09:05 UncleFester6 joined 09:10 raiph left 09:13 pmurias joined
pmurias what does QAST::Want do? 09:13
masak pmurias: it handles coercion. see the three usages in src/QAST/Compiler.nqp 09:24
(morning, #perl6)
pmurias masak: hi 09:30
masak: QAST::WVal contains a perl6 land value? 09:31
can I extract the Int from a QAST::WVal in nqp land?
jnthn .value gets the object itself 09:33
It's a bit more sutble than "a perl6 land value" - QAST::WVal is for objects that are in the serialization context for the compilation unit.
The right way in QAST to ask about stuff with known compile time values is that you can call .has_compile_time_value on any node, and the if that gives you back something true, you may call .compile_time_value to get hold of it. 09:34
JimmyZ jnthn: memory use from 600M down to 490M in pirt branch, fyi
jnthn Nice!
JimmyZ jnthn++
r: say ( 600 - 490 ) / 600 * 100 , '%' # 09:35
p6eval rakudo 843534: OUTPUT«18.333333%␤»
masak brilliant! we should have thought of eliminating PCT compilation phases much earlier. :P 09:37
jnthn masak: Sure, but which things would you have done later? :P 09:41
JimmyZ so, pct will be give up eventually? 09:44
masak PCT is alive and kicking, as far as I know.
dalek href="https://perl6.org:">perl6.org: a165385 | GlitchMr++ | source/archive/rfc/230.html:
Readd HTML version of RFC 230
 09:50
09:51 huf joined
GlitchMr Somehow I wonder why it didn't existed before 09:51
09:52 daniel-s left, araujo joined, araujo left, araujo joined 10:06 birdwind1pbird left
GlitchMr perl6: 'abc'.chop.print 10:13
p6eval rakudo 843534, niecza v19-15-g051783d: OUTPUT«ab»
10:14 rcmalleb joined
arnsholt I get a failure in t/spec/S16-filehandles/filestat.t with a pretty recent rakudo. Is that expected? 10:14
not ok 8 - IO.accessed should be updated when contents of file is read
GlitchMr ok 8 - IO.accessed should be updated when contents of file is read 10:15
Probably not 10:16
jnthn arnsholt: I've heard @other report issues in that file. It seems platform specific. 10:17
arnsholt Right. I'm on RedHat Linux
10:17 JimmyZ left
GlitchMr I'm at a4b8ad8cf33e3954b22f5221685e8423de77e303... but I don't think that revision exists (it's Rakudo + working run() function) 10:17
arnsholt I also have two passing TODOs in S02-types/type.rakudo but that's because of my patch =D 10:18
GlitchMr I'll download newest version to check, ok?
dalek kudo/nom: 2652a19 | (Arne Skjærholt)++ | src/Perl6/Actions.pm:
Keep track of return type info in --> signature syntax.
kudo/nom: 3ec91e5 | (Arne Skjærholt)++ | src/Perl6/Actions.pm:
Make create_signature_object take a signature info hash instead of param list.
kudo/nom: 82c4cb1 | (Arne Skjærholt)++ | src/Perl6/ (2 files):
Pass signature info hashes to World.create_signature.
kudo/nom: 161ab28 | (Arne Skjærholt)++ | src/Perl6/World.pm:
Actually use signature info in World.create_signature.
kudo/nom: 8cce367 | (Arne Skjærholt)++ | / (33 files):
Merge branch 'nom' of github.com:rakudo/rakudo into nom

Conflicts: src/Perl6/Actions.pm
kudo/nom: a251275 | (Arne Skjærholt)++ | / (3 files):
Merge branch 'nom' of github.com:rakudo/rakudo into nom
GlitchMr what?
10:19 daniel-s joined
arnsholt Oh, I pushed anyways =) 10:19
jnthn arnsholt++
arnsholt Didn't see your question 'till after
GlitchMr That is long log... 10:21
10:21 adu left
frettled 5 10:26
10:28 spider-mario joined
jnthn r: say chr(120) 10:28
p6eval rakudo 843534: OUTPUT«x␤»
jnthn r: say chr(95) 10:29
p6eval rakudo 843534: OUTPUT«_␤»
dalek ast: ac096e5 | (Arne Skjærholt)++ | S02-types/type.t:
Rakudo unfudge.
 10:30
arnsholt jnthn: The Zavolaj test suite errors out with "elements() not implemented in class 'Parcel'" all of a sudden. Sound familiar? 10:36
jnthn arnsholt: Not particularly. 10:37
Sounds like something that wants to be an RPA is getting turned into a Perl 6 type at some point...but I couldn't guess where.
arnsholt Oh, hang on. I might have the wrong NQP for my branch 10:38
Yeah, I think that might be it 10:39
Incidentally, I think this is my longest yak shave so far 10:40
As well as my first proper dig into Rakudo's internals
jnthn And you survived the experience! \o/ 10:41
arnsholt Indeed ^_^
I have to admit after I was done with the Rakudo hack I did have to stop for a second to execute the context switch back to what I was actually doing =D 10:42
10:43 birdwindupbird joined 10:45 jlaire left
dalek p/pirt: 302b444 | jnthn++ | src/QAST/PIRT.nqp:
Save some work (and potentially memory) during the final string concatenation.
 10:45
p/pirt: 570712c | jnthn++ | src/QRegex/NFA.nqp:
Fix thinko in NFA building, which unbusts 2 out of the 3 failing S05 tests with PIRT.
kudo/pirt: eea3e60 | jnthn++ | src/Perl6/Actions.pm:
Remove a hack we no longer need, fixing one of the other regressions.
 10:46
10:47 jlaire joined, thelazydeveloper joined 10:54 birdwindupbird left
moritz r: sub f(&x:(--> Int)) { }; f -> --> Int { } 10:55
p6eval rakudo 843534: OUTPUT«===SORRY!===␤invoke() not implemented in class 'QAST::WVal'␤»
moritz r: sub f(&x:(--> Int)) { }; 10:56
p6eval rakudo 843534: OUTPUT«===SORRY!===␤invoke() not implemented in class 'QAST::WVal'␤»
moritz r: sub f(&x:(Str --> Int)) { };
p6eval rakudo 843534: OUTPUT«===SORRY!===␤invoke() not implemented in class 'QAST::WVal'␤»
moritz r: sub f(--> Int) { 'str' }; f 10:57
p6eval rakudo 843534: ( no output )
moritz oh, rakudo is still too old
jnthn
.oO( Ракудо стар )
 10:58
masak :P 11:01
moritz --> Int works here locally, but the &x:(...) still gives "invoke() not implemented in class 'QAST::WVal" 11:02
masak phenny: "Ракудо стар"?
phenny masak: "Rakudo star" (sr to en, translate.google.com)
masak phenny: ru en "Ракудо стар"?
phenny masak: "Rakudo old" (ru to en, translate.google.com)
masak :P
tadzik oh, the cross-language puns 11:03
moritz jnthn: i guess error-reporting and callframe.t depend on annotations? 11:04
jnthn moritz: Yeah, afaik that's all that's left. 11:06
moritz: Am trying to untangle how POST did it at the moment... :) 11:07
11:07 lue left, rcmalleb left
arnsholt moritz: I think it has to be &cb(...) (no colon) 11:07
jnthn I think really the colon syntax should also work, but arnsholt++ has probably saved enough yaks for the moment ;-) 11:09
moritz arnsholt: ah, with subsignatures
masak phenny: en pl "old"?
moritz arnsholt: the &cb:(...) should also work, as part of the longname
phenny masak: "stary" (en to pl, translate.google.com)
masak tadzik: seems to work in Polish as well...
moritz arnsholt: but that's fine
tadzik :) 11:10
r: say 0.272 * 3934, " MB" 11:11
p6eval rakudo a25127: OUTPUT«1070.048 MB␤»
tadzik rakudo/pirt build on 64-bit
jnthn tadzik: Can you remember what it was for you before any of the QAST stuff? 11:12
11:12 GlitchMr left
tadzik jnthn: I remember the reason why I bought another 2 GBs of ram :P 11:13
but that was long ago
ISTR it being about 1.5 GBs
11:15 lue joined
moritz yes, 1.5GB it was on my machine too 11:20
1.7G before the bs merge
then dropped to 1.3 or so 11:21
and then slowly increased again to 1.5
11:22 SamuraiJack left, cognominal joined 11:24 jimmy1980 left
dalek p/pirt: 614dc7a | jnthn++ | src/QAST/PIRT.nqp:
Add back annotations.
 11:26
jnthn OK, spectesting and chesetesting of pirt now welcomed. :)
And memory improvement reports :)
moritz r: sub f(--> Int) { Str }; f 11:28
p6eval rakudo a25127: OUTPUT«Type check failed for return value; expected 'Int' but got 'Str'␤ in sub f at /tmp/zZZuIVUc21:1␤ in block at /tmp/zZZuIVUc21:1␤␤»
moritz \o/ 11:29
arnsholt++
jnthn merges latest nom stuff into pirt also, to ease testing 11:31
11:32 jimmy1980 joined 11:33 GlitchMr joined
dalek ast: 25cc0f0 | moritz++ | S32-str/substr.t:
remove .chrs usaeg from substr.t

After yesterday's discussion I think it will go away, and it's not what we're actually testing here anyway.		 11:34
moritz jnthn: spectesting 11:37
jnthn moritz: ok
Looks fine after the merge. 11:39
dalek kudo/pirt: 2652a19 | (Arne Skjærholt)++ | src/Perl6/Actions.pm:
Keep track of return type info in --> signature syntax.
rakudo/pirt: 3ec91e5 | (Arne Skjærholt)++ | src/Perl6/Actions.pm:
rakudo/pirt: Make create_signature_object take a signature info hash instead of param list.
11:39 dalek left
jnthn BOOM! 11:39
11:39 dalek joined, ChanServ sets mode: +v dalek 11:42 phenny left 11:43 phenny joined
jnthn away for a bit 11:45
11:50 takadonet left
moritz jnthn: seems that line numbers aren't correct 11:52
jnthn: they are all 1 higher than they should be, at first glance 11:53
12:00 drbean left
jnthn my $line := HLL::Compiler.lineof($!node.orig(), $!node.from(), :cache(1)) + 1; 12:00
hmm 12:01
jnthn removes the + 1
dalek p/pirt: eb0c0a3 | jnthn++ | src/QAST/PIRT.nqp:
Try to make line numbers more correct.
masak accidentally puts a division slash in the right-hand part of a s///e and watches hilarity ensue 12:02
...in p5, where this is an issue, that is.
12:03 drbean joined 12:10 SamuraiJack joined
moritz jnthn: yes, the + 1 ins't necessary anymore. I've fixed line_of some months ago :-) 12:10
12:11 sivoais left
moritz respectesting 12:12
masak .oO( I will respectest you in the morning ) 12:13
moritz 1066MB virtual memory used while building rakudo/pirt 12:17
12:17 Chillance joined
moritz jnthn: seems some of the line number stuff is still a bit off 12:26
jnthn: one of the line number tests in error-reporting.t now fails (the other pass), and the line number from callframe.t is one off
the one that fails in error-reporting.t is sub mysub { \n +Any \n }; \n mysub() 12:27
it reports a warning from line 1, not line 2
weird, it says .annotate 'line', 2 right before the call to &prefix:<+> 12:30
spectest summary: t/spec/S06-advanced/callframe.rakudo Failed test: 2, t/spec/integration/error-reporting.rakudo integration/error-reporting.rakudo (Wstat: 0 Tests: 13 Failed: 1) Failed 12:33
test: 11
if I replace the warning with a die(), it still gets the wrong line number 12:34
the difference in the generated PIR between nom and pirt is that pirt emits two annotations in a row 12:36
.annotate 'line', 1
.annotate 'line', 2
then the call
nom just emits the second (correct) annotation line
I would have expected the second to take precedence
12:38 lue left 12:39 Chillance left 12:40 Chillance joined
felher little question: do methods handle named parameters differently from sub? Because this doesn't die: 12:40
r: class A { method a() { }; }; A.new.a(:unexpected);
p6eval rakudo a25127: ( no output )
felher while this does:
r: sub a() { }; a(:unexpected);
p6eval rakudo a25127: OUTPUT«Unexpected named parameter 'unexpected' passed␤ in sub a at /tmp/Bv6HvYbNt7:1␤ in block at /tmp/Bv6HvYbNt7:1␤␤»
moritz felher: yes, that's by design 12:41
felher moritz: oh, okay. Thanks :) 12:43
moritz: but sub a() is still a matching candidate as far as the dispatcher is concerned, but then fails on binding? 12:45
(if that makes any sense, at all... :) ) 12:46
moritz felher: yes 12:48
felher: multi dispatch is done primarily based on the nominal type of positional parameters
felher: and then where-clauses and named params are "epsilons" tighter than just the matching positional arguments 12:49
jnthn moritz: Investigating. 12:50
felher moritz: is the dispatcher supposed to handle named parameters and such someday in the future? 12:52
jnthn What's really going on is that methods get a *%_ added implicitly. 12:53
While subs don't.
moritz felher: the spec doesn't mandate it; dunno if it's feasible at some point in the future
jnthn: is it? 12:54
jnthn moritz: In the original example. :)
jnthn was responding kinda late :)
Trying a patch for the annotations. 12:55
moritz jnthn: do we expect the dispatcher to treat named args/params the same way as positional ones? (like, in perl 6.2 or so) 12:56
jnthn moritz: I think there's something in the spec that conjectures that a Perl 6 version beyond 6.0 may take named parameters into account during narrowness analysis.
moritz: Working out the details of that may well be extremely hairy, though. 12:57
12:57 lue joined
moritz jnthn: is that because the narrowness analysis doesn't interact well with an unknown number of params that need to be analyzed? 12:58
jnthn moritz: More that you can have lots of partially overlapping sets of named parameters between candidates. 13:00
felher moritz, jnthn: thanks, folks :) 13:01
masak felher: what's *really* going on is the interaction of named parameters being optional, and Perl 6 OO wanting to preserve some kind of Liskov substitutability for deriving types. 13:03
masak grins
13:06 SamuraiJack_ joined 13:07 SamuraiJack left
felher masak: oh, makes sense. Otherwise methods in derived types will thrown an exception if given the old named arguments, right? 13:08
moritz right 13:09
felher \o/ :)
moritz masak++ # nice explanation
13:13 GlitchMr left
jnthn Grr...first fix failed. :/ 13:15
jnthn tries again
dalek p/pirt: 5202f7d | jnthn++ | src/QAST/PIRT.nqp:
Fix dupe annotation emission.
 13:23
13:24 spider-mario left
jnthn moritz: That seems to nail it. 13:24
masak moritz: oh, it actually made sense? huh. though to be fair, that's just echoing what A12 and S12 already say on the subject. 13:37
13:38 tokuhiro_ left 13:39 tokuhiro_ joined 13:41 mtk left 13:47 lue left 13:50 mucker joined 13:52 GlitchMr joined
dalek p/pirt: dc13ec1 | jnthn++ | src/ops/nqp.ops:
Make type checking op quietly say no, so it can be safely used in a wider context.
 13:56
p/pirt: 388a271 | jnthn++ | src/QAST/ (2 files):
A little optimization. Shaves some more seconds of CORE.setting compile time here, and off the spectest also.
13:58 benabik left, Coleoid left, benabik joined 13:59 Coleoid joined 14:00 jimmy1980 left 14:01 jimmy1980 joined
moritz jnthn: spectests on newest nqp/pirt are fine now 14:26
jnthn moritz: OK. Think it should have an ecosystem test before merging? 14:27
moritz jnthn: will try to do one 14:28
jnthn moritz: OK, thanks :)
14:29 tokuhiro_ left 14:36 brrt joined
jnthn submits a YAPC::Asia talk 14:37
14:38 Bucciarati left, Bucciarati joined
GlitchMr Just wondering, is there method like .first, except returning index? 14:40
brrt nqp: pir::compreg("nqp"); 14:43
p6eval nqp: OUTPUT«error:imcc:The opcode 'compreg_p' (compreg<1>) was not found. Check the type and number of the arguments␤ in file '(file unknown)' line 169411186␤»
14:43 sivoais joined
brrt nqp: pir::compreg__Ps("nqp") 14:43
p6eval nqp: ( no output )
[Coke] gist.github.com/3199314 - just built and installed rakudo (recent vintage) in ./rakudo; checkout panda in ./panda, and try to run the bootstrap, it explodes 14:44
masak GlitchMr: no, but you can do .pairs.first({...}).key 14:45
moritz [Coke]: does panda's offline-bootstrap branch work for you?
[Coke] moritz: same error 14:49
This is perl6 version 2012.07-210-g8435346 built on parrot 4.4.0 revision RELEASE_4_4_0
14:50 SamuraiJack_ left
moritz that's pretty recent 14:51
14:51 SamuraiJack joined
dalek kudo/nom: 8c740aa | moritz++ | docs/ChangeLog:
update ChangeLog
 14:55
14:58 JimmyZ joined 15:00 MayDaniel left
felher Hm, since candidates_matching of Routine only asks the dispatcher, is there a way to check if a bind will also work? 15:00
15:13 brrt left
moritz tadzik: it seems that newest Mojolicious comes without render_file_to_file, which broke the emmentaler renderer 15:14
pmichaud good morning, #perl6 15:20
JimmyZ good am, pm
felher o/ pm 15:21
jnthn hi, pmichaud 15:22
dalek mentaler: 75131c9 | moritz++ | formatter:
[formatter] avoid usage of render_file_to_file, which is gone from new versions of Mojo
 15:24
moritz moritz.faui2k3.org/tmp/pirt-smoke.html is the result of the ecosystem test run on top of rakudo/pirt 15:25
59 modules ok, 47 not ok 15:27
GlitchMr Just wondering, is there any policy about modules which use nqp or parrot (read: not Perl 6, but works on Rakudo Perl 6)? 15:28
jnthn moritz: Matches the last run, it seems: tjs.azalayah.net/index.html
15:30 MayDaniel joined, JimmyZ left
GlitchMr github.com/colomon/Math-ContinuedFractions/ 15:32
Also, Perl 6 module without module files... what?
I mean, tests work, but... 15:33
This module cannot be used in real code because of it
jnthn moritz: So, I think merge time :) 15:34
pmurias what creates blib subdirectories during the rakudo build? 15:35
github.com/pmurias/rakudo supports ./perl6 -Cjs -e 'say "Hello World"' | node ;) 15:38
15:40 hoelzro|away is now known as hoelzro
hoelzro if I define my own BUILD submethod, the default BUILD submethod Mu provides should still be called, right? 15:40
jnthn pmurias: !! :D
GlitchMr At least it's not always output "Hello World", but it's still hacky :P
jnthn Everything starts out hacky. :) 15:41
GlitchMr Well, viv is hack too
I'm too lazy to actually compile this modified rakudo, but would it support escapes? 15:42
Such as "\n"
moritz jnthn: +1
pmurias: I don't think create subdirectories during the build 15:43
jnthn Ah
There's a .gitignore in blib/Perl6
moritz pmurias: we put .gitignore files in dirs we put files into
pmurias moritz: thanks 15:44
15:44 uniejo joined 15:46 spider-mario joined
dalek Heuristic branch merge: pushed 27 commits to nqp by jnthn 15:47
GlitchMr 27 commits? 15:48
Oh, I see, branch merge
jnthn yes, pirt
15:50 uniejo left, uniejo joined
dalek kudo/nom: 99d850b | jnthn++ | src/Perl6/Actions.pm:
Initial update to get Rakudo to build with nqp/pirt. Eliminates the final PAST::Node usages also. Some test failures; seems CORE.setting memory is a bit lower with this, which was the aim.
 15:50
kudo/nom: 2b57323 | jnthn++ | src/Perl6/Grammar.pm:
Fix user defined ops, which probably deals with many of the test regressions.
kudo/nom: eea3e60 | jnthn++ | src/Perl6/Actions.pm:
Remove a hack we no longer need, fixing one of the other regressions.
kudo/nom: e5fe863 | jnthn++ | / (5 files):
Merge branch 'nom' into pirt
kudo/nom: 4e3d0ff | jnthn++ | src/Perl6/ (2 files):
Merge branch 'pirt' into nom
kudo/nom: c1bfbbb | jnthn++ | tools/build/NQP_REVISION:
Bump to an NQP revision with PIRT.
cognominal pirt?
jnthn cognominal: More memory-efficient replacement for POST.
cognominal nice 15:53
jnthn Seems that the next Rakudo release ships with a ~ 35% memory saving during CORE.setting compilation over the previous one. :)
moritz guesses he can now leave firefox open while compiling rakudo :-) 15:55
pmichaud rakudo no longer does inlining of native stuff? 15:57
e.g., my int $i = 5000; $i = $i - 1; isn't inlining the subtract :-/
jnthn pmichaud: No, I didn't re-do the inliner yet 15:58
pmichaud: It's pretty much next on my todo list.
(That is, we'll have it again well before the next release.) 15:59
pmichaud okay. we're probably paying a penalty for that with list stuff, then, as I switched much of the native nqp:: stuff to rely on the inliner.
(I can switch back to avoid that reliance if we wish.)
mhasch when will the next release be due? 16:00
moritz pmichaud: since jnthn will re-do the inliner soon anyway, I don't think it's worth switching back
16:00 am0c joined
jnthn pmichaud: I'd leave it; I suspect the inliner is back within days. 16:00
pmichaud mhasch: next release is scheduled for 08/23
jnthn So, well over 3 weeks away :) 16:02
mhasch very fine, thanks pmichaud.
16:02 soooga left
pmichaud 3 weeks away, yes, but we lose _some_ time due to yapc::eu :) 16:02
anyway, if you say the inliner will be back within days, I'm fine with that :) 16:03
moritz ponders writing a bot which reports release dates
mhasch the time on yapc::eu will be well spent, I am sure :-)
pmichaud I'll just have to keep it in mind when doing timings and tests on list code until then :)
jnthn pmichaud: Yeah, it didn't feel worth holding back getting more testing/feedback on all the rest of QAST for that. 16:04
pmichaud jnthn: sure, no problem. I just wasn't quite expecting loops to slow down like they did. :)
ooc, did we agree that the 2012.08 release would be "Frankfurt"?
jnthn pmichaud: At least we know the optimizer actually does something worthwhile now :P 16:05
16:05 xinming_ left
moritz I don't know if we agreed on that, but that was my plan when I snatched away "Frankfurt" as a release name two months ago :-) 16:05
pmichaud ...or do we want to wait and see if there are actually any Perl 6 talks at YAPC::EU? ;-)
jnthn
.oO( only if there are...dang, Pm beat me :P )
 16:06
japhb Anyone here have any skills hinting GraphViz layouts? My type graph visualization iteration is hitting a wall. :-/
16:06 xinming joined
pmichaud oh, I see masak++ got a email reply 16:07
jnthn pmichaud: When you do the <ws> tweaks, may be worth seeing if it can be sped up any along the way.
16:08 uniejo left
cognominal jnthn, switch to QAST is still on the ROADMAP. I thought that was completed even stuff like optimizations is not back 16:08
au japhb: there's a really good, although proprietary, GraphViz .dot hinter application called OmniGraffle.app
moritz cognominal: nqp hasn't switched to QAST yet
jnthn pmichaud: Looking at CORE.setting profile output, over 5% of time (inclusive) is spent in ws. Compared to 2.4% or so for the entire optimizer phase. :) 16:09
cognominal thx, moritz
pmichaud jnthn: sure... but much of that is just because .ws is called sooooooo often
16:09 GlitchMr left
jnthn pmichaud: Aye, the call counts say as much. 16:09
225,000 calls while parsing CORE.setting. 16:10
moritz might be a worthy target for the inliner :-)
pmichaud <.ws> doesn't inline easily.
jnthn Aye.
moritz cheers on jnthn to get to the method inliner :-)
japhb au, Thank you. Unfortunately, I'm actually looking for hints I can generate through relatively simple algorithmic means, starting from e.g. a human-friendly text file
Along the lines of "I want Pod::Config near to Pod::Block, but otherwise run normal layout" 16:11
pmichaud jnthn: are there any subrules of <.ws> that show up high on the list? 16:12
jnthn pmichaud: MARKED
pmichaud number of calls, time spent, or both?
jnthn pmichaud: 262k calls, 2.45% 16:13
pmichaud 2.45% of the CORE.setting time?
jnthn Yes.
pmichaud okay, that might be where we can get some speedup.
mhasch just realized there is a significant lack of perl6 talks on the Franfurt talk list. What happened?
pmichaud blogs.perl.org/users/moritz/2012/07...-2012.html 16:14
jnthn pmichaud: To put it in context, !cursor_pass gets 757128 calls.
pmichaud: And it is at 2.41%
pmichaud well, every call to <.ws> also generates a call to MARKED
but yes, MARKED is obviously on the high side.
moritz mhasch: ine ssence, the organizers did a majority vote to accept talks, and nobody had any overview of talks per topic or so 16:15
pmichaud I'll definitely look at it.
jnthn OK. I'm optimistic that once we get NQP on QAST, and I write an NQP::Optimizer, we can get some more wins. 16:17
(In general.)
back later & 16:19
hoelzro did anyone see my question about custom BUILDs? 16:21
moritz where? 16:27
moritz did not 16:28
are niecza releases on the last Monday of the month?
pmichaud moritz: yes, that's the pattern I remember. 16:29
moritz say Date.today.truncated-to(:week).day-of-week 16:34
r: say Date.today.truncated-to(:week).day-of-week
p6eval rakudo c1bfbb: OUTPUT«1␤» 16:35
moritz r: say Date.today.truncated-to(:week)
p6eval rakudo c1bfbb: OUTPUT«2012-07-23␤»
hoelzro if I define my own BUILD submethod, the default BUILD submethod Mu provides should still be called, right? 16:36
it was a good hour ago or so.
moritz Mu doesn't provide a BUILD submethod for you
submethods are *not* inherited
hoelzro right
moritz so, Mu # aka, the question doesn't make sense
hoelzro but if I don't provide my own BUILD submethod, the 'default' fills in my public attributes from the arguments to bless, doesn't it? 16:37
moritz that's correct
hoelzro ok
so, what I should be asking is this:
moritz r: class A { has $.x; submethod BUILD() {} }; say A.new(x => 3).x 16:38
p6eval rakudo c1bfbb: OUTPUT«Any()␤»
moritz r: class A { has $.x; submethod BUILD(:$!x) {} }; say A.new(x => 3).x
p6eval rakudo c1bfbb: OUTPUT«3␤»
hoelzro how do I provide my own custom BUILD submethod, but still populate the attributes I didn't initialize in my own BUILD?
ex: gist.github.com/3199546
16:38 uniejo joined
moritz hoelzro: add :$!foo to the signature of your BUILD 16:39
16:39 uniejo left
hoelzro moritz: so if I have a good deal of attributes in my class, and I want to have a custom BUILD, I need to do that by hand? 16:40
moritz hoelzro: yes
hoelzro =(
moritz I know, it kinda sucks 16:41
though in your case, a default value works
has $.bar = YourClass.new
hoelzro I was thinking nextsame; would do it, but it's not valid because it's a submethod, right?
moritz right
hoelzro moritz: default values are evaluated per instance, right?
moritz r: gist.github.com/3200120
hoelzro: right
p6eval rakudo c1bfbb: OUTPUT«parrot: 2012-08-21␤rakudo: 2012-08-23␤niecza: 2012-07-30␤»
hoelzro moritz: ok, thanks! 16:42
au japhb: PodConfig -> PodBlock [weight=999]; 16:43
hoelzro moritz: also, we were talking about interpolating strings into regexes last night; where is that defined in the spec? 16:44
I'd like to add a test to roast for it
au or, programatically (P5), $graphviz->add_edge($node_a, $node_b, weight => 999);
moritz S05
hoelzro seciton? 16:45
*section
moritz I don't know that by heart :-
)
hoelzro heh
TimToady S05:1165 for one 16:46
hoelzro I looked in S05 a bit earlier, but I couldn't find anything definitive
moritz there are some tests in S05-interpolation/ already, more welcome
16:46 benabik left
TimToady well, S05:1148 16:46
hoelzro oh, ok
TimToady: thanks
moritz hoelzro: there's a whole section, perlcabal.org/syn/S05.html#Variable...erpolation
16:46 benabik joined
TimToady and S05:1535 16:47
"leading character" there means first char inside <...>
nr: my $x = '.'; say 'foo.bar' ~~ / $($x) / 16:48
p6eval niecza v19-15-g051783d: OUTPUT«#<match from(3) to(4) text(.) pos([].list) named({}.hash)>␤»
..rakudo c1bfbb: OUTPUT«#<failed match>␤»
TimToady rakudobug
nr: my $x = '.'; say 'foo.bar' ~~ / $x /
p6eval niecza v19-15-g051783d: OUTPUT«#<match from(3) to(4) text(.) pos([].list) named({}.hash)>␤»
..rakudo c1bfbb: OUTPUT«q[.]␤␤»
moritz I think $(...) in regex is simply NYI 16:49
hoelzro "$x" is as well, iirc
TimToady nr: my $x = '.'; say 'foo.bar' ~~ / "$x" /
moritz correct
p6eval niecza v19-15-g051783d: OUTPUT«#<match from(3) to(4) text(.) pos([].list) named({}.hash)>␤»
..rakudo c1bfbb: OUTPUT«#<failed match>␤»
hoelzro I'm just being a good reporter and making sure there's a test in roast for it =)
moritz hoelzro++
TimToady nr: my $x = '.'; say 'foo.bar' ~~ / <$x> / 16:50
p6eval niecza v19-15-g051783d: OUTPUT«#<match from(0) to(1) text(f) pos([].list) named({}.hash)>␤»
..rakudo c1bfbb: OUTPUT«q[f]␤␤»
pmichaud getting "$x" to work shouldn't be too difficult. 16:54
japhb au, Hmmm. That sortof worked, once I made it style=invis, but it still thinks Pod::Config should be at a rank below Pod::Block. And if I set constraint=false, it doesn't just turn off the rank constraint, it turns off the nearness constraint as well. 16:59
japhb tries adding rank=same to the link
Harumph, didn't work.
Ah, that only works for a subgraph 17:00
TimToady r: sub foo (&infix:<op>, $a, $b) { say $a op $b }; foo &[/], 42, 7;
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤Unable to parse blockoid, couldn't find final '}' at line 2, near "say $a op "␤»
TimToady doesn't seem as though user-defined ops are quite fixed yet
pmichaud I suspect rakudo doesn't recognize &infix:<op> in a signature as needing to modify the parser. 17:01
beyond that, what is the precedence of <op> in that case? same as &infix:<+> ? 17:03
TimToady that would be the default
pmichaud okay.
moritz Timbus: ping 17:04
TimToady
.oO(also &infix:<op> is equiv(&[X]))
moritz Timbus: I'm trying your Net::IRC. The README says the way to reply to an event is $e.reply($msg), but I get " such method 'reply' for invocant of type 'Net::IRC::Event'"
pmichaud but that example likely implies a re-think/refactor of how Rakudo handles user-defined operators.
I think Rakudo might currently do it in the sub declaration.
moritz does my &infix:<op> also modify the grammar?
TimToady well, it probably means there are two codepaths where there could be one 17:05
japhb moritz, the docs for Net::IRC lie. I played with that a bit, so lemme check my notes ...
moritz n: my &infix:<op> = -> $a, $b { $a + $b }; say 1 op 3
p6eval niecza v19-15-g051783d: OUTPUT«4␤»
moritz japhb: seems I need to use $e.msg 17:06
TimToady n: sub foo (&infix:<op>, $a, $b) { say $a op $b }; foo &[/], 42, 7;
p6eval niecza v19-15-g051783d: OUTPUT«6␤»
TimToady that one works in niecza too
japhb moritz, correct, yes
but there was something else I wanted to pass on as well.
pmichaud yeah, the grammar needs to be modified whenever &infix:<op> is encountered in a declarational context.
japhb looking for that
pmichaud and not just when the name of a routine happens to be &infix:...
17:07 mikemol left
pmichaud it probably also implies something about how 'is equiv' and 'is tighter' are handled. 17:09
since they attach to the declared name and not to the routine being declared. 17:10
moritz which brings us to the topic of trait handling
IMHO we should pass the name of the traitee to the trait 17:11
that way constant + is export can work together more easily
japhb moritz, gist.github.com/0e3dd578cdad1d93c4b6 17:13
moritz japhb: thanks 17:14
japhb np
diakopter TimToady: what is .classify in radix_sort 17:20
TimToady is still wondering if 'has $.foo = stuff;' can attach stuff lazily to FETCH such that in BUILD a FETCH before a STORE gets the default, but a STORE before a FETCH clobbers it 17:21
japhb chuckles ... in attempting to clean all the gunk out of that gist, I accidentally left out a semicolon on line 5.
17:21 lue joined
TimToady diakopter: it, er, classifies things 17:22
it's defined in S32/Containers:158 or so 17:23
17:23 spider-mario left
TimToady note that neither implementation actually implements it correctly 17:23
it's supposed to return a Hash, not a list of Pairs
pmichaud didn't moritz++ make a fix for this recently? 17:24
TimToady nr: say (1,2,3,4,5).classify(* % 3).WHAT
p6eval rakudo c1bfbb: OUTPUT«Hash()␤»
..niecza v19-15-g051783d: OUTPUT«Unhandled exception: Unable to resolve method classify in type Parcel␤ at /tmp/nc7IxXi1hN line 1 (mainline @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3929 (ANON @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3930 (module-CORE @ 564) ␤ at …
TimToady ohh, a Hash
nr: say (\classify(* % 3, 1..10).WHAT 17:25
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤Unable to parse postcircumfix:sym<( )>, couldn't find final ')' at line 2, near ""␤»
..niecza v19-15-g051783d: OUTPUT«===SORRY!===␤␤Unable to parse parenthesized expression at /tmp/w3RrJAFK6X line 1:␤------> say (⏏\classify(* % 3, 1..10).WHAT␤Couldn't find final ')'; gave up at /tmp/w3RrJAFK6X line 1 (EOF):␤------> say (\classify(…
TimToady nr: say classify(* % 3, 1..10).WHAT
p6eval rakudo c1bfbb: OUTPUT«Hash()␤»
..niecza v19-15-g051783d: OUTPUT«List()␤»
TimToady yeah, looks fixed in rakudo now
pmichaud moritz++ 17:26
TimToady nr: say classify(* % 3, 1..10).{0,1,2}
p6eval rakudo c1bfbb: OUTPUT«3 6 9 1 4 7 10 2 5 8␤»
..niecza v19-15-g051783d: OUTPUT«Unhandled exception: Cannot use hash access on an object of type List␤ at <unknown> line 0 (ExitRunloop @ 0) ␤ at /home/p6eval/niecza/lib/CORE.setting line 354 (Any.at_key @ 9) ␤ at <unknown> line 0 (ExitRunloop @ 0) ␤ at /tmp/vyCvcC6jHl line 1 (mainli…
17:26 lue left
TimToady it only makes sense, since it has already constructed a Hash, to just return it, rather than forcing the caller to rebuild the Hash if they want one 17:27
nr: say classify(*.substr(*-1,1), (-100..100).roll(20))< - 0 1 2 3 4 5 6 7 8 9 > 17:29
p6eval niecza v19-15-g051783d: OUTPUT«Unhandled exception: Cannot use hash access on an object of type List␤ at <unknown> line 0 (ExitRunloop @ 0) ␤ at /home/p6eval/niecza/lib/CORE.setting line 354 (Any.at_key @ 9) ␤ at <unknown> line 0 (ExitRunloop @ 0) ␤ at /tmp/bXHQM1nqM8 line 1 (mainli…
..rakudo c1bfbb: OUTPUT«Unable to coerce value for '$start' from WhateverCode to Int; no coercion method defined␤ in method substr at src/gen/CORE.setting:1919␤ in block at /tmp/M9btQUCyZW:1␤␤»
17:31 kaare_ left, adu joined
skids r: (-1 +> 16).say; (-1 +> 8).say; (-1 «+>« (16,8))».say 17:31
p6eval rakudo c1bfbb: OUTPUT«-1␤-1␤0␤0␤»
TimToady r: say classify(*.substr(0,1), (-100..100).roll(20)).WHAT 17:32
p6eval rakudo c1bfbb: OUTPUT«Hash()␤»
TimToady r: say classify(*.substr(0,1), (-100..100).roll(20)).{'-',0..9}
p6eval rakudo c1bfbb: OUTPUT«-97 -89 -100 -22 -96 -12 -100 -33 -14 -7 -64 -25 -36 Any() 13 17 22 25 Any() Any() 57 Any() Any() 82 95␤» 17:33
TimToady hmm, missing entries should return (), not Any
pmichaud it's a Hash.
missing entries of Hash are Any. 17:34
17:34 am0c left
TimToady r: say classify(*.substr(0,1), (-100..100).roll(20)).{'-',0..9}:p 17:36
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤Confused␤at /tmp/kTOLPaqIh0:1␤»
TimToady we had something in there about skipping missing entries when slicing somewhere 17:37
don't remember if it was with or without the :p
pmichaud probably needs to be with, otherwise %h<a b c> = .... doesn't dwim
17:38 brrt joined, GlitchMr joined
hoelzro should rx { "$!instance-var" } function the same as my $var = $!instance-var; rx { "$var" }; ? 17:39
moritz yes 17:42
hoelzro ok, so that needs to go into the test as well, then
17:46 lucas_ joined
tadzik moritz++ # fixing smoked cheese 17:47
jnthn Being able to smoke the cheese is terribly useful. 17:49
pmichaud jnthn: should I open github issues in perl6/nqp for qast-related things? 17:50
dalek ast: 7ad5915 | (Rob Hoelz)++ | S05-interpolation/regex-in-variable.t:
Add test for double quote regex interpolation
 17:51
ast: f0f43c4 | (Rob Hoelz)++ | S05-interpolation/regex-in-variable.t:
Add test for interpolating instance members
pmichaud hoelzro: is there a ticket for "$x" in RT somewhere? 17:52
hoelzro pmichaud: yes (rt.perl.org/rt3/Public/Bug/Display...id=113958)
pmichaud thanks 17:53
pmichaud takes the ticket, will likely work on it tonight.
hoelzro I think I found a bug in the Rakudo REPL; where should I put the test for that? 17:54
moritz hoelzro: we don't have a good way to test the REPL yet :(
hoelzro ah 17:55
moritz and since the REPL isn't specced, they can't go into roast
pmichaud ...what's the bug?
hoelzro well, let's say I stick a role named Animal into Animal.pm6
17:56 p6rd joined
hoelzro if I do "use Animal;\nmy $dog = Any.new;\n$dog does Animal", the REPL complains that it doesn't know about Animal 17:56
although this could be my limited experience with Perl6 talking here.
moritz p6rd: help
p6rd Perl 6 release dates. Usage: p6rd: [ 'next' | 'next month' | YYYY-MM | YYYY-MM-DD ]
moritz p6rd: next month
p6rd rakudo: 2012-08-23, parrot: 2012-08-21, niecza: 2012-08-27
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤Confused␤at /tmp/B5eA4REbJr:1␤» 17:57
moritz oh, it shouldn't talk to p6eval :-)
17:57 p6rd left, p6rd joined
moritz p6rd: next 17:58
p6rd rakudo 2012-08-23, parrot 2012-08-21, niecza 2012-07-30
pmichaud ...should it sort by date there? ;-) 17:59
i.e., niecza 2012-07-30, parrot 2012-08-21, rakudo 2012-08-23
18:00 MayDaniel left
moritz pmichaud: good idea 18:00
hoelzro so is that behavior I saw in the REPL a bug?
or a defiency in my understanding?
pmichaud hoelzro: my initial guess is that it's a bug. It probably has to do with the fact that each line of input to the REPL is a separate compilation unit. 18:01
18:01 whiteknight joined
hoelzro ok 18:02
18:02 p6rd left
pmichaud and so when it gets down to $dog does Animal the compiler doesn't have "Animal" in a symbol table somewhere that it needs. 18:02
18:02 whiteknight is now known as Guest60380
pmichaud jnthn: nqp::while(cond, body) is now always casting cond to a PMC... any way to avoid that? 18:02
or perhaps it's that nqp::isge_i($i, 0) is always casting to a PMC, although that seems less likely. 18:03
jnthn pmichaud: Hmm...I suspect it's the "while".
pmichaud: I need to review a few of those once I get the inliner in place.
pmichaud: Does "if" do the same, ooc? 18:04
pmichaud checking.
18:04 p6rd joined
moritz p6rd: next 18:05
18:05 p6rd left
pmichaud jnthn: doesn't look as though "if" is doing the same. 18:05
let me re-check 'while'
18:06 p6rd joined
pmichaud well... hrm. 18:06
18:06 spider-mario joined
pmichaud nqp::while boxes the result of the isge, but never actually uses the boxed result. 18:06
jnthn pmichaud: oh...
r: say do while False { } 18:07
p6eval rakudo c1bfbb: OUTPUT«False␤»
jnthn r: say do while 0 { }
p6eval rakudo c1bfbb: OUTPUT«0␤»
jnthn pmichaud: I suspect it's the "'cus we want it for the return value" thing...
Fixable of course.
pmichaud ...except I'm not sure that's the p6 semantic anymore :)
moritz p6rd: next 18:08
p6rd niecza 2012-07-30, parrot 2012-08-21, rakudo 2012-08-23
moritz now sorted by date, pmichaud++
skids But did he use a radix sort? :-) 18:09
pmichaud jnthn: the while loop isn't context aware -- it always generates the PMC even if the caller is requesting something else? 18:10
18:10 MayDaniel joined
felher moritz++ #p6rd 18:13
moritz github.com/moritz/p6-release-dates # source code
jnthn pmichaud: I think it should be; I can believe there's something up with that.
18:13 sergot left
jnthn pmichaud: Will investigate in a bit. 18:14
dinner, bbs
pmichaud okay. I need lunch myself... and perhaps I'll go work on something else for a while.
japhb Whee, finally starting to figure out dot hinting for type graph visualization! 18:15
au, the trick was to do: subgraph "cluster: foo" { rank=same; style=invis; "node1"; "node2"; } 18:16
That pulls the two (or more) nodes together into a same-rank packed area of the output 18:17
skids Hrm. RE: irclog.perlgeek.de/perl6/2012-07-27#i_5850800 -- Now that I have July Star locally, still I get a "value is not known at compile time" running that line locally without the BEGIN -- but evalbot is happy with it... weird. 18:18
18:19 hoelzro is now known as hoelzro|away
moritz skids: the fix came in after the release 18:20
skids Ah.
moritz skids: well, after the compiler release, and star uses the compiler releas
skids OK, then not some difference in the way evalbot is built or run. Good.
moritz++ 18:21
moritz speaking of which, I should simplify the rakudo build 18:22
on p6eval
skids BTW, what is the current story with the --parrot-option Configure, and should that go into the README? 18:23
moritz it should go into INSTALL 18:24
README is too long already
skids Ah... didn't notice INSTALL.txt as it's tucked in a subdir in the star distro tree. 18:27
18:29 sergot joined 18:31 adu left
jnthn pmichaud: Yeah, I'm gonna need to spend a bit more time on this area. Will take care of if when working on inlining. It is trying to take care of type stuff, but apparently not getting things quite right there. 18:34
18:39 brrt left
moritz Timbus: and it seems that the IRC bot only joins the first channel that I pass to :channels 18:40
18:41 brrt joined 18:44 nebuchadnezzar left
felher moritz: Why "$date - $date.truncated-to(:month)" instead of "$date.day-of-month - 1"? Matter of taste or something i missed? Since i'm a big fan of types i would have used "Event $e" in "said" instead of "$e" - but this certainly is a matter of taste. Apart from that i find it to be quite cute :) 18:48
masak r: sub foo(&x) { my $value = "inside foo"; return { say $value; say "[&x()]" } }; my $value = "outside foo"; foo { $value }; # moritz: actually, this is not altogether different from how macros work 18:51
p6eval rakudo c1bfbb: ( no output )
moritz felher: I didn't think of day-of-month, it's certainly easier with it
masak hm.
r: sub foo(&x) { my $value = "inside foo"; return { say $value; say "[&x()]" } }; my $value = "outside foo"; (foo { $value })();
p6eval rakudo c1bfbb: OUTPUT«inside foo␤[outside foo]␤»
masak there we go.
18:56 birdwindupbird joined 19:01 p6rd left, p6rd joined, Coleoid left 19:02 adu joined
moritz p6rd: upcoming 19:02
p6rd niecza 2012-07-30, parrot 2012-08-21, rakudo 2012-08-23
jnthn p6rd: 6.0.0 19:03
.oO( dang, it didn't say christmas )
moritz it now runs on a 24/7 server, so we'll see how stable it runs
and with a ulimit, just to be sure 19:05
GlitchMr perl6: my $var = 42; print "${var}"; 19:13
p6eval rakudo c1bfbb: OUTPUT«===SORRY!===␤Non-variable $ must be backslashed at line 2, near "${var}\";"␤»
..niecza v19-15-g051783d: OUTPUT«===SORRY!===␤␤Unsupported use of ${var}; in Perl 6 please use {$var} at /tmp/dWC3YtJE68 line 1:␤------> my $var = 42; print "${var}⏏";␤␤Parse failed␤␤»
GlitchMr :)
I prefer Niecza error message
19:17 brrt left 19:20 brrt joined 19:21 crab2313 joined, brrt left 19:23 nebuchadnezzar joined, mikemol joined 19:24 GlitchMr left 19:25 SamuraiJack left, SamuraiJack joined 19:26 nwc10 joined
pmurias jnthn: how does QAST::Var.decl work? 19:34
what does it mean?
jnthn pmurias: Contains 'var' or 'param'
pmurias: Or it's empty to mean "this is just a normal variable access" 19:35
pmurias: Note that Rakudo uses its own signature binding stuff, so you'll only have to contend with decl.
(For Rakudo.)
masak github.com/moritz/p6-release-dates...e-dates.pl # niiiiice 19:37
moritz++
19:39 lucas_ left 19:40 birdwindupbird left, birdwindupbird joined 19:41 lue joined
arnsholt Is there a difference in the kind of thing returned from return $thing rather than falling off the end of a block ending with $thing? 19:41
lue hello world o/
masak lue! \o/
r: sub foo { my $thing = 42; return $thing }; sub bar { my $thing = 42; $thing }; say foo.^name; say bar.^name 19:42
p6eval rakudo c1bfbb: OUTPUT«Int␤Int␤»
masak arnsholt: doesn't look it.
r: sub foo { my $thing = 42; return $thing, $thing }; sub bar { my $thing = 42; $thing, $thing }; say foo.^name; say bar.^name
p6eval rakudo c1bfbb: OUTPUT«Parcel␤Parcel␤»
jnthn I'd not expect one. 19:43
arnsholt Cause that's how I fixed my problem in Zavolaj earlier today
pmurias jnthn: for every my variable the is a QAST::Var with a decl equal to 'var'?
arnsholt With return, breakage. No return works fine
jnthn pmurias: Correct.
arnsholt Mind, the value returned form the sub is bound with := not assigned
jnthn arnsholt: What sort of value is it? 19:44
arnsholt nqp::list
pmurias jnthn: and what does the WVal with an Array contain?
19:45 bbkr joined
jnthn arnsholt: That explains it somewhat. 19:46
pmichaud once you start working with nqp::list, then Rakudo's perl6izer often comes into play.
jnthn arnsholt: Those are not proper Perl 6 types so you can get into some fun with them.
arnsholt Right, right. That's probably it then
jnthn pmurias: I don't get the "with an Array" part
pmurias: WVal = World Value = a reference to an object created at compile time. 19:47
skids IIRC the returned form is specced to be an "abnormal control" exception... but implementation status of that... ??
pmurias a mysterious WVal containing an array appears in the past for -e '1'
jnthn Hm 19:48
jnthn looks
oh, I wonder if it's Pod related...
pmurias: Yeah. Grammar.pm:514 19:49
pmurias jnthn: and I get a segfault when calling pir::typeof__SP on it's first element 19:50
jnthn: thanks
jnthn typeof is a Parrot-y thing rather than a 6model-y thing. You might want to do $obj.HOW.name($obj) or so 19:51
If you're after the name.
pmichaud jnthn: I think I'm wanting a nqp::stmt or nqp::stmts opcode.
pmurias jnthn: how do you display the past for a code snippet? 19:52
pmichaud jnthn: any suggestions about how to add one?
jnthn $ops.add_hll_op('perl6', 'stmts', -> $qastcomp, $op { $qastcomp.as_post(QAST::Stmts.new( |@($op) ))
});
pmurias s/past/qast/
jnthn Add something like that in src/Perl6/Ops.pm 19:53
You can promote it to NQP later if you like
pmichaud I think I'll want to add it to nqp, but I .... right.
jnthn It's easy, just s/hll/core/ and toss the 'perl6' arg
pmichaud I'm looking at nqp::list, and it appears to create a PIRT::Ops node, though.
some of the nqp ops can create PIRT directly?
jnthn No, only QAST::Compiler/QAST::Operations generates PIRT 19:54
If you do nqp::list in Perl 6 source, it makes a QAST::Op.new( :op('list'), ... )
That is, nqp::foo is just a way to make QAST::Op nodes 19:55
The translation happens later
Not like with nqp:: before with PAST, which could translate directly to a PAST node with the pirop in.
pmichaud okay, just a sec. 19:56
(this makes sense at a high level; I'm wanting to understand a lower level)
jnthn pmurias: In theory, --target=past. In practice, it explodes...
pmichaud ...because --target=past still uses Parrot's Data::Dumper, and not everything (especially meta/knowhow stuff) knows how to dump itself. 19:57
jnthn oh, ouch!
I thought it'd been updated. D'oh.
pmichaud it was updated for one type, but I think there are still some missing. 19:58
github.com/perl6/nqp/blob/master/s...s.nqp#L174 <-- can you describe what this is doing?
jnthn OK. It's on my todo list, after the inliner.
Looking
pmichaud: QAST::Operations contains a registry. 19:59
pmichaud yes, I know that part.
jnthn OK, to clarify - what bit do you want to know about?
pmichaud specifically, unlike add_hll_op above, it appears to have a closure that returns a PIRT instead of a QAST
oh oh oh
jnthn No
Both return PIRT
pmichaud right
I see the .as_post now
jnthn Yeah 20:00
pmichaud so, when you define a new "op", you pass it a closure that knows how to construct the PIRT
jnthn Correct.
pmichaud I think I like that, but I'm not sure yet. :-)
jnthn And then there's an auto-mapper for ops that go directly to PIR.
Note that I don't encourage *widespread* use of the trick I showed (that is, making QAST to immediately de-sugar). 20:01
However, it is very convenient in situations like the one you just asked for :)
pmichaud so, in the stmts example above, I could just as easily have created something like 'list' does instead of going through QAST::Stmts
jnthn pmichaud: Well, actually if you just have a list of statements to compile, you could do exactly what QAST::Stmts does. 20:02
multi method as_post(QAST::Stmts $node) {
self.compile_all_the_stmts($node.list, $node.resultchild, :node($node.node))
}
Stmt does similar but with an extra bit of diddling to handle register allocation stuff.
pmichaud right, Stmt imposes a temporary register sequence point. 20:03
jnthn So you could write it as $qastcomp.compile_all_the_stmts($op.list)
And avoid the temporary QAST::Stmts node.
pmichaud okay, that tells me what I need to know for now.
The advantage of going through the QAST node of course is that you gain any future modifications (i.e., DRY) 20:04
jnthn Yeah
The disadvantage is making throwaway objects, but I doubt this is a hot path.
pmichaud correct, it's likely not a hot path. and I think that DRY trumps throwaway concerns until demonstrated otherwise.
the more I look at QAST, the more I like it. I'm really glad you did the work and not me, I think we end up with a much better product (my views would've likely been too colored by the PAST implementations) 20:06
masak jnthn++ 20:07
moritz jnthn++ indeed
I haven't looked much at the design, but I like the results :-)
pmichaud I'll probably be doing some refactors and shifting a few things around over time... but it looks really solid.
masak if you read anything, let it be github.com/perl6/nqp/blob/master/t/qast/qast.t 20:08
it's a wonderful test file.
pmichaud I'll look at it also :)
pmichaud looks now. 20:09
masak oh, actually github.com/perl6/nqp/blob/master/s...mpiler.nqp is a good read too. it made me realize a few things about compilers.
(primarily that OO can be applied to them in nice ways)
pmichaud wow, Compiler.nqp is only 1463 loc 20:10
and that's including the regex stuff
20:11 tokuhiro_ joined
pmichaud so, waitaminnit. rakudo is using PIRT now, yes? 20:11
moritz yes
pmichaud so are we using _any_ of PCT anymore? 20:12
(for Rakudo, I know NQP has yet to be converted.)
jnthn pmichaud: No
pmichaud whoa.
masak \o/
jnthn++
pmichaud that's.... so incredibly cool.
jnthn pmichaud: Unless today's HLL::Compiler still inherits from PCT::Compiler
moritz my $compiled := PAST::Compiler.compile($parse);
from Perl6::Grammar 20:13
pmichaud jnthn: it does not.
moritz isn't that a PCT thingy?
jnthn Yeah.
moritz (something regey, iirc)
jnthn Because of NQP :)
moritz: Yeah, it's dynamically generating an action method.
To go into Perl6::Actions 20:14
(Which is written in NQP.) 20:15
I think that gen_op may be the only mention of PAST.
It also must be re-done.
(In a bigger sense that "re-write it to use QAST" 20:16
I think that code is the great survivor from the ng days or soemthing.
masak rn: my @a = <a b c d e>; say $_ ~~ /^ f @a oo $/ for "faoo", "fcoo", "fzoo"
p6eval rakudo c1bfbb: OUTPUT«#<failed match>␤#<failed match>␤#<failed match>␤»
pmichaud if you're just installing an action method, NQP doesn't care if it was generated using PAST or QAST, though.
p6eval ..niecza v19-15-g051783d: OUTPUT«#<match from(0) to(4) text(faoo) pos([].list) named({}.hash)>␤#<match from(0) to(4) text(fcoo) pos([].list) named({}.hash)>␤Match()␤»
jnthn pmichaud: Yeah, it probably could be updated.
masak rn: my @a = <a b c d e>; say so $_ ~~ /^ f @a oo $/ for "faoo", "fcoo", "fzoo"
p6eval niecza v19-15-g051783d: OUTPUT«True␤True␤False␤»
..rakudo c1bfbb: OUTPUT«False␤False␤False␤»
pmichaud at least, it shouldn't care :-)
jnthn pmichaud: As I said though, gen_op is horribly broken in other ways. 20:17
pmichaud jnthn: correct, it is.
and there was another example of that a few hours ago; it needs to be called on the declaration of the name
not from routine_def
jnthn Yeah.
That ain't so bad to fix.
But it's what it does that needs a good going over.
pmichaud sure, I agree. 20:18
jnthn We really need to do the is assoc/is loser etc traits too.
*losser
argh..
looser
pmichaud fwiw, those apply to the name declaration and not the routine also
jnthn pmichaud: Probably need to be carried on both. 20:19
pmichaud could be that, too.
jnthn I'm thinking exporting them.
We need to associate the traits with the name on import.
pmichaud the easiest way to start is to have a is PREC('c=') trait which forces a given precedence.
once that works, getting looser/tighter/equiv to work is a smop 20:20
(where 's' really is 'simple')
jnthn Good point.
masak heh -- I found rt.perl.org/rt3/Ticket/Display.html?id=95504 last summer when I implemented crypt the first time. now I hit upon it again. :)
moritz yep, NYI 20:21
pmichaud but shouldn't be too hard to YI
moritz btw, is the interpolation syntactic? 20:22
I mean, if I do my $s = [1, 2, 3]; /$s/
is that same as if I did it with a @ sigil?
pmichaud my current guess is that $s is treated as a string there.
but I could be wrong about that, since if $s is a Regex it is treated as a regex. 20:23
moritz TimToady: ^^ any comment?
I guess @($s) is still available to do @ interpolation semantics 20:24
pmichaud see also github.com/perl6/mu/blob/master/misc/pm.txt#L7 :-)
TimToady I'd guess it should be string
20:24 SamuraiJack left
pmichaud well, "$s" is also available to force string semantics :-) 20:24
TimToady std: / @$_ /
p6eval std e52e3ca: OUTPUT«ok 00:00 41m␤»
TimToady I think people will expect string coercion there 20:25
(without @)
moritz since Array is Cool, we should use syntax to determine the operation, not type
so, +1 to $ forcing Str (unless it's a Regex)
masak pmichaud: well the odd thing is, as the RT ticket explains, that there appears to be support for array interpolation.
and ISTR someone actually implementing it. 20:26
pmichaud masak: there was support in alpha and/or beijing, iirc
masak ah.
moritz by the way, can we have Parcel to be Cool too, pleaes?
nr: say Parcel ~~ Cool
p6eval rakudo c1bfbb, niecza v19-15-g051783d: OUTPUT«True␤»
pmichaud it's Cool already, afaik
moritz oh, it already is
never mind then
nr: say Parcel.^mro
p6eval rakudo c1bfbb: OUTPUT«Parcel() Cool() Any() Mu()␤» 20:27
..niecza v19-15-g051783d: OUTPUT«Unhandled exception: Unable to resolve method mro in type ClassHOW␤ at /tmp/vx4q5_ZycW line 1 (mainline @ 4) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3929 (ANON @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 3930 (module-CORE @ 564) ␤ at /ho…
pmichaud I'm likely to have Parcel be Iterable, too.
[Coke] blah. getting more pugs failures in daily spec run via cron that don't show up in a standard run.
moritz is Iterable a role or a class? 20:28
pmichaud it's currently a class.
moritz the name just sounds role-y, IMHO
pmichaud agreed
the name may also want a change, too.
moritz though looking at Iterator and Iterable in Rakudo right now, they look as if they could very well be roles 20:29
pmichaud it wouldn't be too difficult (now) to make it into a role if we decide we need to do that. At the time I was implementing Iterable, role support in nom was still not completely there.
bbkr example from spec does not work anymore: github.com/perl6/specs/issues/20 20:30
pmichaud bbkr++ # good catch 20:31
pmichaud can update pm.txt :-)
masak what determines if something should be Cool? why should Parcel be Cool? 20:34
(not questioning the decision, just curious)
pmichaud for many of the same reasons that Array and List are Cool (more)
if I have sub foo(Cool $x) { say +$x );
then I'm basically saying that I expect most of the standard types to be passed in as $x and that they'll coerce between num/str/list as appropriate. Parcel does that. 20:35
(did that answer the question satisfactorily?) 20:37
masak how do I say "what's been matched by the current rule" inside a <?{ ... }> block?
pmichaud masak: $/ ought to work
masak oh, good. 20:38
pmichaud r: say 'abc' ~~ / .* <?{ $/.say }> /; # checking
p6eval rakudo c1bfbb: OUTPUT«q[abc]␤␤q[abc]␤␤»
pmichaud r: 'abc' ~~ / . <?{ $/.say }> /; # checking
p6eval rakudo c1bfbb: OUTPUT«q[a]␤␤»
masak \o/ 20:40
moritz masak: in general, stuff is Cool if it's a type that occurs often (and not just in very specific situations), and that has a sensible string and number interpretation 20:42
pmichaud r: say 'abc' ~~ / . <?{ $/.say }> .. /; # checking 20:43
p6eval rakudo c1bfbb: OUTPUT«q[a]␤␤q[abc]␤␤»
moritz for example Parameter is not cool, because it doesn't have a sensible numeric interretation, and you don't just run accross Paramter objects in normal p6 code 20:44
s/cool/Cool/
masak lue: you'll be pleased; I'm writing a small grammar for command parsing :) 20:45
moritz: ok, fair enough.
lue I figured, judging by your recent questions :)
masak heh :) 20:46
lue++
tadzik masak: as in: "make me spaghetti?"
moritz wants "sudo extinguish fire" as a command in crypt 20:47
:-)
masak :P 20:49
jnthn has blaught: 6guts.wordpress.com/2012/07/29/raku...eductions/
moritz jnthn++ 20:52
am I right in assuming that the QAST tree isn't GC'ed after the PAST -> POST stage? 20:53
lue jnthn++
tadzik jnthn++ indeed
jnthn moritz: s/POST/PIRT/ :) 20:54
bbkr Is is possible to get class method using "&" operator? For example I can do "&substr" to get Callable object instead of executing it. But if I have class A {method a {}} neither &A.a or &A::a works.
tadzik Parrot IR tree?
masak "Happily, though, we’ve actually got a little bit faster at compilation." -- wow!
jnthn tadzik: PIR Tree
masak tadzik: the tree is infrared, so it glows a dull red. 20:55
tadzik kewl
jnthn bbkr: No, you canny do that
masak tadzik: and there are clogs hanging in it.
moritz bbkr: $obj.can('methodname')[0]
jnthn bbkr: Easiest way are is $obj.^find_method('methodname')
20:55 pjcj left
masak bbkr: the methods are in a namespace not accessible through the & sigil. we discovered this as we hammered out 'our' and 'has' scopes around 2009. 20:56
jnthn moritz: I think it should be collectable beyond the PIRT stage
moritz: Oh, wait...
moritz: That contains a link to $/
bbkr thanks for explanation
jnthn moritz: Which in turn references the QAST tree through .ast
So basically, the whole lot goes away togehter.
*together
moritz jnthn: watching the memory growing during setting compilation, it seems it's not collected 20:57
20:57 sergot left
moritz jnthn: huh, I thought we didn't reference $/, just extract its .orig and .from? 20:57
at least that's what PAST did
20:57 sergot joined
masak "In other exciting news, the arrival on QAST means masak++ is digging into taking macros some steps further. Maybe there’ll be something nice there for the next release too. :-)" -- yes, I hope so too. and it doesn't feel that far-fetched. 20:57
jnthn That...wasn't my understanding of what PAST did
masak though $family is going to require my attention in the next few days. 20:58
jnthn I thought it was at the POST stage that we pulled out the orig/from
masak (nothing serious, just spending time)
21:02 birdwindupbird left
jnthn pmichaud: For inlining, I'm going to tag with of the ops (as in QAST::Op) are free of side-effects that'd mean you can't inline them. I'm pondering :pure(1) being specified at the point you add them. Sound reasonable? 21:02
er, I messed upt he sentence :) 21:03
...autopun :/
I meant free of side-effects meaning you *can* safely inline them...
.oO( Clearly it's time for a beer... )
 21:04
nwc10 *a* beer?
I think you've earned more than one 21:05
although, yes, one at a time :-)
jnthn I'm not sure beer drinking parallelizes well on a single node. :)
masak beer serialization context.
nwc10 gets even worse if you try to paralellise drinking (say) beer and cider 21:06
tadzik slurp()
21:06 Guest60380 left 21:07 whiteknight joined
jnthn I'd never seen snakebite as an example of parallelism... 21:07
21:07 whiteknight is now known as Guest55706
masak learns about "shandy" -- eeew 21:10
21:10 bbkr left
pmichaud jnthn: (:pure) thinking 21:11
geekosaur try not to turn into a monad. or a comonad, for that matter 21:14
pmichaud jnthn: the approach sounds fine; I don't remember if "inlinable" is exactly what :pure was meant to indicate. 21:16
jnthn pmichaud: Well I could make the name inlinable 21:17
It felt a bit...unweidly :)
pmichaud :inline(1) would be fine :-)
jnthn *unwieldy
pmichaud except that .inline means something else in P6 (on callframes)
jnthn Hmm, that...right.
Well, and it feels instructive rather than stative.
It means "something can feel free to inline code containing this op elsewhere, 'cus this op runs on risk of screwing it up" 21:18
*no
pmichaud anyway, +1 from me.
we can change the names easily enough later if needed.
jnthn True.
pmichaud reads blog post.
jnthn Was mostly fishing for a better name. 21:19
21:19 pmurias left
pmichaud :pure works for now; I don't have a better name at the moment. I'm hoping it's not a conflict with :pure in STD.pm6 21:20
blog post (and the work it describes) is excellent. jnthn++
jnthn pmichaud: It's in the same semantic niche at least. 21:22
pmichaud jnthn: should I update docs/nqp-opcode.txt to reflect the QAST opcodes?
21:22 spider-mario_ joined
jnthn pmichaud: Yes, feel free. 21:22
pmichaud that would give me a good opportunity to review them, too :)
ooc, is there still a flag equivalent to :flat on nodes? 21:24
21:25 spider-mario left
jnthn You can :flat(1) on any of the nodes. 21:26
21:26 spider-mario_ is now known as spider-mario
pmichaud excellent. 21:26
I may use that sometime in nqp::list and nqp::hash
i.e., nqp::list(1, 2, |@a, 3) should inline the elements of @a
jnthn oh...
I didn't mean it quite like that.
pmichaud (I know that doesn't currently work; just wanted to know that we had the flag.) 21:27
jnthn I meant "call and callmethod pay attention to it, and it can go on any node"
Oh, in that case, fine :)
pmichaud nqp::list(1, 2, @a, 3) should end up with a list of four elements, nqp::list(1, 2, @a, 3) will generate code that at runtime will splice the contents of @a into the list that is being built.
oops, |@a on the second nqp::list 21:28
and it can do that by looking at the :flat attribute of the child node containing @a
21:28 nwc10 left
pmichaud but that's a future improvement. :-) 21:28
I had been planning to do something similar in PAST but never got around to it. :) 21:29
21:29 tokuhiro_ left 21:33 nwc10 joined
nwc10 OK, on the Raspberry Pi with pirt 21:33
Stage parse : 5220.855
QAST was
Stage parse : 23840.491
before all of this was something like 52535.669
pmichaud nwc10: those are seconds?
nwc10 jnthn++ # use less 'swap' :-)
yes
note, it hasn't yet managed to build without crashing 21:34
either SEGV
pmichaud so, we're now down to 1.5 hrs instead of 15 hrs? ;-)
nwc10 or the whole thing locks up
yes. 1.5 hours for the parse stage
masak wow.
pmichaud well, we're getting closer then :)
masak Perl 6 on Raspberry Pi! \o/
nwc10 it is rather wow compared to the previous
masak: I can install Rakudo with apt-get
(just not a current one)
but I suspect that debian's buildbots are beefier machines 21:35
so far:
Stage start : 0.000
Stage parse : 5220.855
Stage syntaxcheck: 0.048
Stage past : 0.015
so I note that the formatting is a tad optimistic - assumng that nothing takes longer than 999 seconds :-)
21:36 crab2313 left
jnthn chuckles 21:36
pmichaud hey, we have to be optimistic *somewhere*! :)
nwc10 note, `vmstat 5` now looks typically like this:
procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu---- r b swpd free buff cache si so bi bo in cs us sy id wa 1 1 149464 8748 44 2020 1050 356 1050 356 8823 1607 0 22 0 78 0 2 149980 8328 44 2020 1195 473 1195 473 8836 1733 1 23 0 77
ug
21:37 Guest55706 left
nwc10 oh, actually, it's not that pretty any more 21:37
it's now in 99% swap hell
but for the parse stage it seems to be only about 80% swap hell
dalek p: 06b7631 | jnthn++ | src/QAST/Operations.nqp:
Mark up the ops that we should be able to inline. (Not using the data for anything yet.)
 21:38
pmichaud I think that must be optimizer swap hell, then.
nwc10 I'm going to go to bed. And see what the morning brings.
jnthn The optimizer walks the whole tree.
I can see how that could be swap-ouchy.
nwc10 poor little USB disk.
skids r: class A { multi method Int () { 3 } }; class B is A { multi method Int() { 2 } }; my B $b .= new(); $b.Int.say; 21:39
p6eval rakudo c1bfbb: OUTPUT«2␤»
pmichaud jnthn: what's the idea of putting the 'defor' definition immediately after the 'for' definition? It took me a few minutes to figure out that it's "def-or" and not "de-for" :-P
skids r: role A { multi method Int () { 3 } }; class B does A { multi method Int() { 2 } }; my B $b .= new(); $b.Int.say;
p6eval rakudo c1bfbb: OUTPUT«Ambiguous call to 'Int'; these signatures all match:␤:(B , Mu *%_)␤:(B , Mu *%_)␤␤ in method Int at src/gen/CORE.setting:325␤ in block at /tmp/UTU5w7zrEB:1␤␤»
jnthn pmichaud: To make people really read the code to see it's not actually a for loop in reverse. :P 21:40
nwc10 oh well, the build on the 32 bit FreeBSD system has failed thanks to (I think) -Werror=strict-prototypes tripping up over a header in /usr/local/include which isn't perfectly strict 21:41
really not convinced that parrot setting all those -Werror options is the smartest thing to be doing. 21:42
pmichaud jnthn: (06b7631 commit).... ummm... what does :pure really indicate, ooc?
jnthn pmichaud: Routines are assessed for whether they could possibly be inlined. 21:43
21:44 adu left
jnthn pmichaud: If they can be, inlining info is stashed in the Routine object. 21:44
pmichaud so, :pure(1) on a 'list' node means that the list node itself is inlinable, but doesn't say anything about the pureness of its children?
jnthn pmichaud: Correct
pmichaud and the inliner would have to check that? 21:45
jnthn pmichaud: The algorithm is basically, walk the AST of the routine until we find something that says "NO!"
pmichaud okay.
wfm.
jnthn That tends to happen pretty quickly
For non-simple routines.
Which are the ones we want to inline.
Goes without saying you can spend weeks tweaking such things... 21:46
21:46 UncleFester6 left, UncleFester6 joined 21:50 nwc10 left
dalek p: 3f4d70f | pmichaud++ | docs/nqp-opcode.txt:
Update nqp-opcode.txt for QAST a bit.
 21:55
21:57 adu joined
jnthn Time for some rest 21:58
'night
22:04 thelazydeveloper left
pmichaud how does one run the tests in t/qast ? 22:05
...and I'm getting some fails.
masak just ./nqp it?
pmichaud rebuilding nqp w/o my patch to get a baseline 22:06
I'm guessing qast should be part of the standard 'make test' for nqp now.
masak I would've guessed so too. 22:07
but it isn't.
pmichaud fixing.
masak pmichaud++
pmichaud the fails are all
not ok 30 - Compilation failure in test_qast_result: Contextual $*QAST_BLOCK_NO_CLOSE not found
okay, got it fixed. 22:12
dalek p: 93d6777 | pmichaud++ | t/qast/qast.t:
Force QAST_BLOCK_NO_CLOSE in compile_qast() in qast tests.
 22:13
22:13 pjcj joined
dalek p: 87b5356 | pmichaud++ | tools/build/Makefile.in:
Add qast and pirt tests to standard 'make test' target.
 22:17
22:21 spider-mario left
masak lol, I blog! \o/ strangelyconsistent.org/blog/july-2...b-synonyms 22:25
lue masak++ 22:29
22:30 jimmy1980 left 22:31 MayDaniel left
masak lue: the game should be much more playable now, with both abbreviated directions and verbs. 22:31
lue: do let me know if you find something unexpected. 22:32
lue goes to break^W play crypt again.
masak is sorely tempted to add 'extinguish fire' and 'sudo extinguish fire' (with appropriately acerbic responses) for moritz
lue my pull of crypt (just now) still says "There is a car here.", by the way. 22:34
masak fixed. 22:39
it was because of a hard-coded description at the start of the game.
a bit of a wart that that's needed, but I haven't devised a fix yet. 22:40
22:40 adu left
dalek p: 2a467a6 | pmichaud++ | / (3 files):
add nqp::stmts() op.
 22:42
lue er, how do I put the tiny disk back on hanoi? I'm sure I'm just not using the right sentence. 22:43
22:44 Chillance left
masak that's currently not possible. 22:44
a wrapping fix similar to the 'take' one is needed there.
actually, the whole disks handling needs to be re-done a bit. 22:45
lue well, then, uh-oh :) [I thought I'd make the hanoi game a bit easier by taking the tiny disk out of play]
masak :P
lue s/play/the hanoi game/
masak yes, that is a cute shortcut.
[Coke]++ discovered it too.
feel free to patch the game.
you currently can't examine any of the disks either. I have a better way to factor that, but I forgot to do it today.
22:46 MayDaniel joined 23:00 Chillance joined 23:01 Chillance left
dalek p: 71ffe4f | pmichaud++ | src/QAST/Operations.nqp:
Fix typo in pure-ification of nqp::isgt_n .
 23:02
lue (found bug in crypt, putting on gist...)
masak: gist.github.com/3179112 (the 29 July entry) 23:04
23:04 whiteknight joined, whiteknight is now known as Guest13642
masak hm. 23:08
that should trigger failing tests, too.
testing.
indeed.
lue++ 23:09
tests++
lue: fixed. pushed. 23:11
lue karma C 23:12
just a thought :) . Anyway, I have to go, bye for now! o/
masak \o
'night, #perl6
23:16 MayDaniel left 23:18 Guest13642 left 23:20 spider-mario joined 23:21 adu joined 23:29 spider-mario left 23:42 Moukeddar joined, Moukeddar left

Powered by Cro and the Raku® Programming Language.

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

Thank you!