svn switch --relocate svn.openfoundry.org/pugs svn.pugscode.org/pugs/ | run.pugscode.org | spec.pugscode.org | paste: sial.org/pbot/perl6 | pugs.blogs.com
Set by avar on 16 November 2006.
00:16 marmic joined 00:18 Khisanth joined 00:35 Khisanth joined 00:56 neonse left 01:07 riffraff joined
riffraff hi 01:07
Tene hi 01:08
Patterner hi 01:09
01:10 riffraff left, riffraff joined 01:19 lyokato joined 01:40 buubot joined 01:56 QwestDawG joined 02:07 hexmode left 02:39 riffraff joined
riffraff hi 02:39
02:44 bucky joined
riffraff help please: what is the perl6ish way of putting unit tests in the same file of the code? I was going to use something like "if $?FILE eq $*PROGRAM { use Test; plan 4; ok...}" but it seems to much work 02:45
02:49 drbean joined
avar riffraff: Pod probably, nobody has implemented that so far afaik 03:10
riffraff I see 03:15
thak you
03:40 audreyt_ joined 03:46 diakopter joined
diakopter wb audreyt 03:46
TimToady oohie! 03:48
diakopter spoke too soon, I guess. 03:51
TimToady maybe we scared her away... 03:53
more likely the last remaining fiber to Taiwan melted...
diakopter irssi auto-reconnect perhaps 03:54
TimToady historically she usually attaches from feather
so I'm assuming she's trying to get out on a business-prioritized line
or is that her new DSL line? 03:55
hinet.net is definitely a chinese site... 03:56
diakopter TimToady: sorry for brashly bantering with Titus and 1Co the other day. 03:59
TimToady he's been dead for 1900 years, I don't think Paul is the least offended. :) 04:00
allbery_b suppose that depends on your theology :)
TimToady I don't think he's offended wherever he is... :) 04:02
hinet.net is definitely audreyt's ISP, I think. 04:07
04:08 LeeD joined
diakopter in Titus he's referring to believers vs. unbelievers; in 1Co he's referring to confident believers among less... liberal ... believers. A misconflation. i'll be back in 13 hours. 04:10
allbery_b is not really the right person to be discussing Christian theology anyway :> 04:15
04:17 audreyt joined 04:20 b00t joined
bucky what does pugscc do? perlcabal.org/~audreyt/darcs/pugs/script/pugscc 04:21
TimToady never tried it. it has a pod section of docs embedded though. 04:23
bucky hmm... 04:24
The 'pugscc' script allows you to create an exectuable image from a Perl6 script 04:25
ic 04:26
TimToady not heavily used, so likely subject to bitrot 04:28
bucky i just found it in Perl6-Pugs-6.2.13 04:30
TimToady my dir shows it at Oct 18, so not much changed since then. 04:32
bucky this is interesting: Perl6-Pugs-6.2.13/util/livecd/mklivecd.pl
TimToady most of us here keep an up-to-date copy from svn/svk
bucky i like this
TimToady if you'd like to hack on it, I can give you a commit bit 04:36
bucky thank you.. but i'm a beginner 04:37
perl6 looks like something i would want to use from now on though 04:38
04:38 demy joined
bucky i just didn't realize that pugscc came with the distribution 04:40
05:01 luqui joined 05:12 Eidolos joined 05:31 drbean joined 05:41 audreyt joined
Khisanth luqui: you could use $ENV{'REMOTE_ADDR'} if you are using CGI, if using mod_perl then you can get it from the request object 06:13
luqui Khisanth, thanks.
06:15 BooK joined 06:22 LeeD is now known as leed, spacebat joined 06:29 audreyt joined
beppu_ ?eval class President { } 07:16
evalbot_r14996 undef
beppu_ ?eval class President { }; my $president = President.new(); 07:17
evalbot_r14996 \President.new()
beppu_ ?eval class President { }; my $president = President.new() 07:18
evalbot_r14996 \President.new()
QwestDawG ?eval print "hello, world!\n"; 07:22
hahaha
evalbot_r14996 OUTPUT[hello, world!ā¤] Bool::True
QwestDawG was worth a shot
woah
beppu_ ?eval [ 0 .. 31 ] >>~<< [ "text" xx 32 ] 07:28
evalbot_r14996 ("0text", "1text", "2text", "3text", "4text", "5text", "6text", "7text", "8text", "9text", "10text", "11text", "12text", "13text", "14text", "15text", "16text", "17text", "18text", "19text", "20text", "21text", "22text", "23text", "24text", "25text", "26text", "27text", "28text", "29text", "30text", "31text")
Tene [ 0 .. 31 ] >>~>> "text" 07:29
?eval [ 0 .. 31 ] >>~>> "text"
evalbot_r14996 Error: ā¤Unexpected ">>~>>"ā¤expecting operator
beppu_ There must be some way to not have to repeat "text" 32 times. 07:30
?eval [ 0 .. 31 ] >>~<< "text"
evalbot_r14996 ("0text", "1text", "2text", "3text", "4text", "5text", "6text", "7text", "8text", "9text", "10text", "11text", "12text", "13text", "14text", "15text", "16text", "17text", "18text", "19text", "20text", "21text", "22text", "23text", "24text", "25text", "26text", "27text", "28text", "29text", "30text", "31text")
luqui beppu_, you are correct
it's just not implemented
(and that behavior is incorrect)
beppu_ really... hmmm..
what's the correct way gonna look like? 07:31
luqui [0..31] >>~>> "text"
beppu_ I've heard that perl 6 would allow people to define operators in unicode. Is anyone here familiar with how that is going to work? 07:33
luqui beppu_, yeah 07:34
what do you want to know?
beppu_ the basic syntax for defining an operator.
Tene sub prefix:<operator> { do some stuff } 07:36
beppu_ Oh, that's simple enough.
Tene ?eval sub prefix:<double> -> $a { $a * 2}; double 2 07:37
evalbot_r14996 Error: ā¤Unexpected "->"ā¤expecting "handles", "is", bare trait, subroutine parameters, trait or block
Tene ?eval sub prefix:<double> { $_ * 2}; double 2
evalbot_r14996 4
beppu_ cool
Tene same for infix and postfix 07:38
beppu_ is a similar mechanism used for operator overloading?
like if you had a date and timerange object and wanted to add them together...
luqui ?eval multi *infix:<*> (Str $x, Str $y) { "$x--$y" } [3*4, "x"*"y"]
evalbot_r14996 [12, "x--y"]
luqui (short answer, yes) 07:39
beppu_ I see. you declare types and rely on multimethod dispatch, right?
luqui ya
beppu_ very interesting. thanks for teaching me.
?eval sub prefix:<ā˜ > { kill(9, $_) }; ā˜  9999999; 07:45
evalbot_r14996 Error: Unsafe function 'kill' called under safe mode
beppu_ hehe
Someone's going to have to come up with a good unicode input mechanism.
Tene ?eval sub prefix:<ā˜¢> { say "$_ curies" }; ā˜¢ 1234 07:47
evalbot_r14996 OUTPUT[1234 curiesā¤] Bool::True
beppu_ this will be an obfuscators wonderland ;-) 07:48
07:51 DHGE joined 07:59 ayrnieu joined
masak beppu_: true, but I must confess that I'm even more looking forward to what one will be able to do with p6 when one's intents are to be as clear as possible :) 08:02
beppu_ of course. :) 08:03
we're gonna have to be creative
there will be a perl renaissance
08:05 frankg joined 08:30 devogon joined
masak beppu_: maybe. right now I'm not sure what the status is really of the pugs project 08:36
it's so much easier when audreyt is present 08:37
then the status is "steadily progressing" :)
beppu_ ?eval [+] (1 .. 8, 6) 08:39
evalbot_r14996 42
beppu_ good night 08:40
08:45 iblechbot joined
masak ?eval multi *infix:<*> (Str $x, Str $y) { "$x--$y" } (3, "x") >>*<< (4, "y") 09:04
evalbot_r14996 ("3--4", "x--y")
masak hm
09:04 buetow joined 09:13 kanru joined
QwestDawG ?eval [-] (1..3) 09:22
evalbot_r14996 -4
QwestDawG lisp influenced 09:23
?
TimToady not really 09:28
09:41 drrho joined 09:43 elmex joined
svnbot6 r14997 | audreyt++ | * Pugs.Prim: Fix t/xx-uncategorized/say-crash.t by printing 10:02
r14997 | audreyt++ | in 4096-byte chunks instead of running the strict encodeUTF8
r14997 | audreyt++ | on the entire string. This commit fixed the bug, but the
r14997 | audreyt++ | next one should generalize this fix to make encode/decode
r14997 | audreyt++ | sufficiently chunk-lazy...
r14997 | audreyt++ | Reported by gaal++.
pasteling "evalbot_r14996" at 194.145.200.126 pasted "Pugs build failure" (429 lines, 24.1K) at sial.org/pbot/22131 10:04
gaal there seems to be some binary junk in your src/Pugs/Help.hs ? 10:07
oh, "your"==evalbot. hahah
audreyt spj++ # fix ghc-trunk in time so it builds pugs _and_ so the watch/break/step debugger feature in ghci would work 10:08
gaal audreyt: need 'make prof' for that, right?
gaal considers a 'make professor' alias 10:09
audreyt no, "make ghci" 10:10
it's builtin ghci
gaal wow. that's... outstanding
spj++ indeed
audreyt there's also :sprint
which prints a value without forcing computation
gaal but this is not in 6.6? 10:11
audreyt no
it just landed in usable form last week
Options for ':breakpoint':
list list the current breakpoints
add Module line [col] add a new breakpoint
del (breakpoint# | Module line [col]) delete a breakpoint
stop Stop a computation and return to the top level
it's somebody's SoC
delayed delivery
but damn useful
gaal d/ls bleadghc
audreyt like, real debugger.
10:11 fglock joined
gaal somebody++ 10:11
audreyt ?eval $?PUGS_VERSION 10:12
evalbot_r14997 \"Perl6 User\'s Golfing System, version 6.2.13, October 17, 2006 (r14997)"
audreyt generalized fix done (make encode/decode sufficiently semilazy) 10:13
benchmarking to see if we lost any perf that way
10:15 miyagawa joined
fglock audreyt: happy new year! 10:15
I'm speccing post-mp6 10:16
please review if you have time 10:17
audreyt happy new year 10:19
cool
pathname?
fglock v6/KindaPerl6 :) 10:20
v6/v6-KindaPerl6 10:21
audreyt heh ok :)
fglock the idea is to have a compiler workflow framework first,
ad then start the module devel
reuse as much as possible from existing compilers
by writing adaptors
audreyt nod 10:22
fglock so that new impls can be done inide this workflow engine
audreyt that sounds most sane
fglock and start writing the optimiers
optimizers
audreyt seems to help perf, amazingly 10:23
resmoking to make sure it didn't help perf by silently doing something wrong ;) 10:24
MP6 -> KP6 10:25
asciibetically the next one in the sequence should be IPv6
fglock re mp6, the spec is now frozen, with a bootstrapped p5 impl, and half-implemented parrot
ipv6 sounds good - what would it expand to?
audreyt wow. nice!
Internet Protocol Version 6
I think
gaal haha 10:26
fglock the parrot impl compiles tokens already, and has a Match class
it has problems with blocks and class variables
blocks need to be compiled as subroutines, requires some work 10:27
and class vars does not seem to be supported natively
audreyt yeah, /me am aware of those places :)
clkao .kasjo6 10:28
audreyt so in KP6 we'd have something that assumes all the subset of perl6 that's required for writing a reasonable fullp6 impl is there
sure. excellent plan 10:29
fglock thanks :) 10:30
btw, mp6-MO worked, but it's too slow to use as a compiler tool 10:31
but it's fine to make experiments wit 10:32
with
audreyt cool!
which part was slow?
fglock the profiles show a lot of calls to Class::MOP::Class::compute_all_applicable_attributes, Class::MOP::Class::get_attribute - possibly fixable with memoization 10:35
audreyt those two are prime memocands
esp get_a
or inlinecands even 10:36
so it's still using oldland C::MOP
not nothingmuch's new MO?
or has it became C::MOP?
audreyt has lots of catching up to do
fglock I guess MO uses C::MOP
audreyt internally, I see 10:37
fglock anyway, I was planning to write a wrapper to p5 native oo, in order to have a common pluggable API
audreyt right
fglock so we can have .HOW on native p5 objects (maybe reuse v6.pm) 10:38
does p6 allows to use different metamodels in the same program? 10:39
this could be a problem
audreyt gaal: if you go about build trhunkghc yourself you need this in mk/build.mk 10:43
GhciWithDebugger = YES
fglock: it allows different HOW
but the protocol is same 10:44
fglock ok
gaal audreyt: thanks, no, i used a binary 10:46
svnbot6 r14998 | audreyt++ | * Pugs.Internals.String: encodeUTF8 and decodeUTF8 is now
r14998 | audreyt++ | lazy in 4k-chunk units, instead of completely strict.
r14998 | audreyt++ | This allows reading/writing of huge files.
audreyt gaal: ok, then you don't get debugger cmds
I think. may be wrong
gaal oh! that's... too bad 10:47
audreyt try ":debugger" in ghci and see if it groks it
but! ghc now compiles itself much faster
:)
gaal doesn't, but :? does show breakpoint options
svnbot6 r14999 | audreyt++ | * Pugs.Prim.List: Remove an unneccessary toInteger.
audreyt yeah, but that's meaningless
gaal why's it off by default? :-(( 10:48
ghc will take up the rest of the weekend to build here...
audreyt still being worked on I guess 10:49
oy. sorry then :/
is your ghc bincompat with feather perchance?
gaal dunno, how do i check?
i just finished building '97 i think 10:50
(x86/32bit here)
(debian unstable) 10:51
audreyt dunno either
you can cp from feather and try...
masak welcome back, audreyt!
audreyt thank you :)
gaal yeah, will. but need to figure out lunch first
audreyt still in backlog mode
masak apparently a backlog mode that does commits along the way, however 10:52
10:52 chris2 joined
audreyt yeah, well, that's a bonus 10:53
svnbot6 r15000 | audreyt++ | * Pugs.Prim: Now encodeUTF8 autochunks, remove hardcoding in op2Print.
fglock is 'class Dog {...}' the same as 'our class Dog {...}' ? 10:54
svnbot6 r15001 | audreyt++ | * Pugs.Types: Compare TMVar with their addresses, not bind Eqs.
r15001 | audreyt++ | This is important for well-orderedness of Thread objects.
gaal does bsl expose that 4069b constant somewhere? 10:55
gaal needs more ram, keeps getting evacuate errors 10:56
src/Pugs/AST/Internals.hs:637:16: 10:57
Not in scope: type constructor or class `VThread'
forgotten svk add?
audreyt no, a sec 10:58
10:58 ruoso joined
svnbot6 r15002 | audreyt++ | * Pugs.AST.Internals - Change "VThread Val" to "VThread" 11:00
r15002 | audreyt++ | since Val is all one thread can conceivably return.
audreyt try again?
(r15003)
gaal pulling
audreyt gaal: no it's not bsl. it's P.Internals.String 11:01
we're not using bsl yet
bsl means all ops on it are 4klz
not only en/de
svnbot6 r15003 | audreyt++ | * Regen instance.
gaal ok looks better on the compilation moose, but ghci is hanging when i do "main", probably just a system problem on my machine 11:05
gotta go now though
see you later :)
(why does ghci not cache the results of compilation somewhere? eg every time i enter make ghci I need to recompile lots of code.) 11:06
nothingmuch gaal: do you have trail map #1?
audreyt bytecode is not yet saved
nothingmuch (hermon)
gaal (yes i realize it's not a compiler mode -- but it could have serialized asts or whatever)
nothingmuch: maybe one from ~10 years ago, but i won't be able to moose it till next weekend 11:07
anyway, i gotta go. <3s to y'all
*waves* & 11:08
y after linking all the libs it just sits there
fglock I've been in hermon - it makes 20yrs :) 11:10
they had 1m snow 11:11
audreyt ?eval 42 11:17
evalbot_r14997 42
gaal ooh i got a pugs shell eventually, just took a while 11:25
fglock: you must have hit a cold year! :)
fglock I still have the dubon :) 11:27
11:27 ozo joined 11:28 mj41 joined
fglock is there a drawing somewhere showing when to execute BEGIN blocks, 'use' statements, etc 11:28
avar where's the code that does perl5 embedding in pugs?
fglock: aside from the synopsis, not a drawing though 11:29
?
audreyt avar: src/perl5/ and src/Pugs/*/Perl5 11:30
fglock I'm trying to sketch a detailed compilation workflow - I wonder if there is something done I could reuse
audreyt fglock: the closest is docs/perl6_compilation 11:32
nothingmuch @tell gaal thanks anymoose, i'm going to buy one now 11:44
lambdabot Consider it noted.
12:05 Grrrr joined 12:21 marmic joined 12:22 buetow joined
fglock if I have separate parse, analyze and emit, when is it the right time to run BEGIN blocks? 12:22
is the answer "you can't split it like that" 12:24
12:43 SCalimlim joined 13:21 ruoso_ joined 13:23 luqui joined 13:27 marmic joined 14:01 gnuvince joined 14:08 fglock joined 14:13 [particle] joined 14:23 baest joined 14:35 neonse joined 14:45 nipra joined 14:59 ozo left 15:02 hexmode joined 15:05 vel joined
fglock is there some kind of OUTER relation between modules like ::A and ::A::B which are declared separately? 15:19
to what extent is ::A::B 'inside' ::A ? 15:20
nothingmuch fglock: that slowness is due to MO objects being constructed 15:25
Class::MOP::compute_all_applicable_attributes is used in Moose constructors
and all the MO meta objects use Moose now
my next step is to bootstrap MO using itself
and then I can finish functionality 15:26
15:35 spoop joined
fglock nothingmuch: ok 15:38
luqui fglock, was your question about perl 6? 15:39
fglock luqui: yes
luqui there is hardly any semantic relationship 15:40
fglock just like p5
luqui right
I'm not actually that well-versed in p5 territory here 15:41
fglock what should I read to know more about 'BEGIN' and 'use' during compilation?
in p6
luqui as far as I know, it's pretty much the same 15:42
modulo new stuff like "is export"
fglock what should I read to know more about it in p5 :) 15:43
is 'use' executed at BEGIN time?
15:43 turrepurre joined
luqui yeah 15:43
fglock so it needs to be compiled recursively 15:44
luqui in p5, use X is equivalent to BEGIN { require X; X->import }
fglock ok
luqui you're asking about circular dependencies?
or...
fglock I'm wondering where to execute BEGIN 15:45
luqui ?
fglock probably inside the return-blocks in the parser
luqui oh. right after } is parsed...
fglock like this:
luqui (rather, statement ::= 'BEGIN' <block> *here*) I imagine... 15:46
fglock token BEGIN { BEGIN <block> { execute_begin; return ... } }
luqui that's what I'd guess, too
fglock another point is, how much of the program is actually supposed to 'exist' when BEGIN is executed 15:48
my and our were not executed yet, right?
15:48 thepler joined
luqui I know we talked about this a lot some months ago 15:49
fglock only 'sub' definitions
luqui but I think it would be pretty hard to find the conversation
anyway, like perl 5
so my's *do* exist
and that's screwed up, and we understand that
let me see if I can find the p6l thread 15:50
15:51 lumi_ joined
luqui okay, 2006-08-14 is the timestamp of a thread asking about this sort of thing 15:54
gaal it's X->import if UNIVERSALL::can('X', 'import'), which nobody seems to remember -- plus the parser hint for indirect object syntax, which nobody seems to forget :)
lambdabot gaal: You have 1 new message. '/msg lambdabot @messages' to read it.
luqui looking at IRC logs nearby
gaal @moosages
lambdabot nothingmuch said 4h 9m 52s ago: thanks anymoose, i'm going to buy one now
gaal s/ALL/AL/
the "readmore" on the latest entry in the pugs blog is broken 15:59
fglock gaal: what's the 'parser hint for indirect object syntax' 16:00
gaal fglock: man perlobj/"Indirect Object Syntax"/'For example, a call to a method "new" in indirect notation' 16:02
luqui fglock, if you are ready for your head to explode thinking about scoping and compile phases together, look at colabti.de/irclogger/irclogger_log/...=771#l1375 16:03
lambdabot Title: #perl6 2006-08-14,Mon, tinyurl.com/ubbbl
fglock luqui: do I have a choice :)
thanks1 16:04
thanks!
luqui oh
I pointed you to the end of that conversation
I guess it never got resolved
we just asked all the questions :-)
fglock heh
luqui can't remember if TimToady was on IRC back then 16:05
fglock gaal: I guess my misunderstanding is - do you mean a hint to the parser, or a hint to everybody 16:06
luqui probably: import { X } LIST 16:07
so that a definition of import in that package wouldn't interfere with parsing... 16:08
er, in the calling package
gaal oh, there's not problem in import -- not THAT problem anyway -- because it really is invoked correctly 16:09
the problem is that C<meth $obj> might mean __PACKAGE__::meth($obj) or $obj->meth 16:10
this problem does not exist in Perl 6
avar gaal: there's more to it than that, see the mess that is toke.c 16:12
iirc at least, there were all sorts of nasty exceptions:)
it's the part of the parser which perldoc calls "not strictly predictable":)
16:15 jrockway_ joined
avar well of course it means one or the other, but the parsing rules are explained in S_intuit_method(pTHX_ char *start, GV *gv, CV *cv) 16:18
luqui ?eval sub foo (&infix:<+>) { 2 + 3 } foo &infix:<*> 16:19
evalbot_r15003 6
luqui ?eval sub foo (&infix:<@@>) { 2 @@ 3 } foo &infix:<*> 16:20
evalbot_r15003 Error: ā¤Unexpected "@@ 3"ā¤expecting operator or "}"
luqui aww
?eval sub infix:<!!> ($x, $y) { $x + $y } 3 ?? 4 !! 5 !! 6; 16:21
evalbot_r15003 Error: ā¤Unexpected ";"ā¤expecting "_", fraction, exponent, term postfix, operator or "!!"
luqui ?eval sub infix:<!!> ($x, $y) { $x + $y } 5 !! 6;
evalbot_r15003 11
luqui ?eval sub infix:<!!> ($x, $y) { $x + $y } 3 ?? 5 !! 6;
evalbot_r15003 Error: ā¤Unexpected ";"ā¤expecting "_", fraction, exponent, term postfix, operator or "!!"
luqui ?eval 3 ?? 5 !! 6;
evalbot_r15003 5
luqui ?eval sub infix:<!!> ($x, $y) { $x + $y } 3 ?? (5 !! 6) !! 7; 16:22
evalbot_r15003 Error: ā¤Unexpected ";"ā¤expecting "_", fraction, exponent, term postfix, operator or "!!"
luqui ?eval my $x; BEGIN { $x = '@@' }; sub infix:{$x} ($a,$b) { $a + $b }; 3 @@ 4; 16:23
evalbot_r15003 7
luqui perl is going to be far too powerful at compile-time 16:24
:-)
?eval my &infix:<@@> ::= &infix:<+>; 3 @@ 4 16:25
evalbot_r15003 Error: ā¤Unexpected "@@ 4"ā¤expecting operator
luqui ?eval &infix:<@@> := &infix:<+>; 3 @@ 4
evalbot_r15003 Error: ā¤Unexpected "@@ 4"ā¤expecting operator
luqui ?eval &infix:<@@> = &infix:<+>; 3 @@ 4
evalbot_r15003 Error: ā¤Unexpected "@@ 4"ā¤expecting operator
luqui ?eval &infix:<@@> = &infix:<+>;
evalbot_r15003 Error: Can't modify constant item: VStr "MkCode {isMulti = True, subName = \"&\", subType = SubPrim, subEnv = Nothing, subAssoc = ANil, subParams = [], subBindings = [], subSlurpLimit = [], subReturns = (mkType \"Any\"), subLValue = False, subBody = Prim ([Pugs.AST.Internals.Val] -> () Pugs.AST.Internals.Val), subCont = Nothing, subPreBlocks = [], subPostBlocks = [], subFirstBlocks = [], subLastBlocks = [], subNextBlocks = [], subKeepBlocks = [],
luqui ?eval &infix:<@@> := &infix:<+>; 16:26
evalbot_r15003 Error: Bind to undeclared variable: "&infix:@@"
16:26 kanru joined
luqui ?eval { my sub infix:<@@> ($x, $y) { 3 @@ 4 } } 16:26
16:26 silug joined
evalbot_r15003 Error: ā¤Unexpected "@@ 4"ā¤expecting operator or "}" 16:27
luqui ?eval { my sub infix:<@@> ($x, $y) { $x + $y } 3 @@ 4 }
evalbot_r15003 7
luqui ?eval { my sub infix:<@@> ($x, $y) { $x + $y } } 3 @@
evalbot_r15003 Error: ā¤Unexpected "3"ā¤expecting operator
luqui ?eval { my sub infix:<@@> ($x, $y) { $x + $y } } 3 @@ 4
evalbot_r15003 Error: ā¤Unexpected "3"ā¤expecting operator
luqui okay, so remaining problems:
?eval sub infix:<@@> ($x, $y) { $y @@ $x } # no parse 16:28
evalbot_r15003 Error: ā¤Unexpected "@@ $x"ā¤expecting operator or "}"
luqui sub foo(&infix:<@@>) { 3 @@ 4 } # no parse
?eval sub foo(&infix:<@@>) { 3 @@ 4 } # no parse
evalbot_r15003 Error: ā¤Unexpected "@@ 4"ā¤expecting operator or "}"
fglock try adding ; after eval
because it doesn't have a line break
luqui wait, where?
fglock oh sorry - I was misreading ?eval 16:29
luqui: the reason I'm asking is because I'm adding 'BEGIN' to the mp6 implementation, and I'm trying to define the compiler workflow at a higher level 16:31
luqui cool. 16:32
I think the strange interplay between compile-time and run-time variables is going to prevent the maximally elegant design 16:33
but do ask if you have any questions, and I'll *try* to answerthem
fglock I'll need some clean way to have the compile-time env, too 16:34
luqui you mean after a program compiles, you need to save the state 16:35
?
so that it can be run (perhaps later, multiple times) 16:36
fglock I mean, where does it store the pad structure 16:37
the state would be preserved as AST nodes, I presume 16:38
luqui I recall coming up with a nice design for this at the hackathon
ask audreyt when she awakes
(hopefully she will remember better than I) 16:39
but can't it just store it alongside everything else 16:40
storing references to pads (so that more than one thing can refer to the same pad)
16:40 jrockway_ is now known as jrockway
luqui i.e., all lexical scopes exist at compile time 16:40
and they are cloned at runtime as needed
to create pads...
it seems a little asymmetrical, but it makes sense 16:41
I am just dumping words as they come into my head
fglock let's see - mp6 pads are native, while compile-time pads are object or hash 16:42
luqui oh, ouch
fglock after parsing, the compile-time pads are not needed, because all info is stored in the AST 16:43
luqui hm... food for thought: my $x; BEGIN { $x = 42; } say $x;
where does 42 go?
fglock dumping the env actually means this is compiled to: my $x; INIT { $x = 42; } say $x; 16:46
?
luqui that is conceivable 16:47
luqui tries to think of a case where that doesn't work... 16:48
fglock it would work for 'use', I think
luqui you mean to call import? 16:49
of course this: BEGIN { say "Hello, World!" } doesn't turn into INIT { say "Hello, World!" }
but that's obvious
fglock maybe it would turn into INIT { 1 } because that's what BEGIN returned 16:50
s/returned/dumped/
luqui my $x; BEGIN { $x = { $x } } # just brainstorming 16:51
fglock is import compile-time only?
luqui yeah
fglock it could just be implemented as a run-time no-op 16:52
luqui import is just a sub call 16:53
there is nothing special about it 16:54
fglock you could tell 'use' to only call 'import' at compile-time 16:56
luqui supposing that use X; is shorthand for BEGIN { require X; X.import if X.can('import') }, wouldn't that happen anyway? 16:58
I think the contents of the {...} in the BEGIN case and in the INIT case being the same was only a coincidence
(above)
you have your compile-time pads 16:59
evaluate everything
then during code generation, output INIT blocks to bring the runtime-pads to what they were at compile-time
it's very strange, because in the following: 17:00
my $x = { my $y; BEGIN { $y = 42 } }
the pad containing $y exists *before* the closure does! 17:01
indeed, in the following, I'm not even sure what the behavior is: 17:03
fglock I guess that from the env point of view, it's not a closure
luqui ?eval for 1,2 { my $x = { my $y; BEGIN { $y = 42 }; say $y }; $x() }
evalbot_r15003 OUTPUT[ā¤ā¤] undef
luqui perhaps not
I would like to have a nice, pretty, consistent picture of all of this in my head 17:04
fglock yes - I was reading about this case right now - aspn.activestate.com/ASPN/Mail/Mess...rs/2132427
lambdabot Title: Re: perl6 style BEGIN {} anyone? :: ASPN Mail Archive :: perl5-porters, tinyurl.com/yh2anv
luqui except this is much more subtle than that problem 17:05
it's assigning to a variable which certainly doesn't exist until runtime at compile time
constant folding is just replacing an ast node 17:06
hmm
I think there is room for specification here
and I am inclined to outlaw the above case
so, how about this: it is illegal to assign to non-existant varaibles at compile-time 17:07
we just need to define 'exists' appropriately
a variable exists if... it is in a closure that cannot be cloned? 17:08
that is, if $x can only ever refer to one thing, then it exists at compile-time
I like that...
so crap like this: for 1,2 { my $x; BEGIN { $x = 10 }; say $x } 17:09
is illegal
but: { my $x; BEGIN { $x = 10 }; sub foo() { $x } }
is not
(assuming that closure was in void context...)
then, in the pad structure, you can store existent variables and reasonably initialize them at the beginning of runtime 17:10
inside their scope... 17:11
17:11 nipra joined
luqui mulls over this a bit more before trying to write it up for p6l 17:12
fglock I'm ok with that - it seems to avoid the problem 17:13
how is INIT {} different from {} in this context? 17:20
luqui not terribly different... 17:21
I'm not actually sure in which context... 17:22
:-)
fglock I mean, can BEGIN blocks be compiled to plain blocks 17:25
containing the local env dump
luqui you mean INIT? 17:26
oh
no
you do mean BEGIN :-)
no, I think init is necessary 17:27
my $x; say $x; BEGIN { $x = 42 }
fglock I think my confusion is, how do you run INIT blocks in a closure
luqui that is what the proposal solves 17:28
this closure is unique, so you can create it at the beginning of the program if you have to
then just change the environment to it
fglock that would be compiled to { $x=42; my $x; say $x } 17:29
luqui (I am making many assumptions about your architecture)
ahh
I think that would work
fglock or { my $x; { $x = 42; say $x } }
luqui now I see what you're saying
given the new restrictions, I think that would work out
the new, hypothetical, restrictions
let's hope they make canon :-) 17:30
fglock it could work like this - each 'my' decl creates a new pad, at compile-time 17:31
and BEGIN blocks are moved to the start of the pad code
luqui where by "moved to " you mean "executed and then the pad's environment is dumped at" ? 17:33
fglock yes
luqui it is a little scary because it seems too simple 17:34
but I do think that it works... 17:35
I am in no position to prove it though
fglock if you 'execute' the begin block repeatedly, it's not a big problem 17:37
luqui ? 17:38
17:38 _bernhard joined
fglock a BEGIN block inside a for-statement is not a big problem 17:40
it will only execute once
luqui true
but if it refers to variables whose scope is the for block, then you get into trouble
fglock example? 17:42
luqui for 1,2 { my $x; BEGIN { $x = 3 } } 17:43
which $x is BEGIN referring to?
[particle] for 1..3 -> $x { BEGIN { $x = 3; } say $x }
heh, very similar examples :)
17:43 kanru_ joined
[particle] how many times does the loop run? 17:44
luqui oh, careful with that
for isn't shorthand for while...
I'd say it runs three times no matter what 17:45
but that is not the point at the moment :-)
[particle] good.
fglock this would compile the same as if without BEGIN
luqui so the $x = 3 is completely ignored
?
fglock because $x=3 dumps to $x=3
luqui oh
I think that is special to this simplistic example 17:46
fglock I guess it dumps an env diff
luqui for 1..3 { my $x; BEGIN { $x++ } }
fglock at the end of BEGIN, $x=1
so it would compile to 17:47
for 1..3 { my $x=1 }
luqui tries to come up with an example that drives home that there are three different $x's
17:48 justatheory joined
luqui my $y; for 1..3 { my $x; BEGIN { $y = { $x++ } } } $y() 17:48
17:49 justatheory joined
[particle] better example 17:49
but why {} around $x++ ?
just for scoping?
luqui it is a closure, closing over the $x inside the for
except there are three different $x's to close over... which one is it? 17:50
[particle] ah, yes. very good example
luqui I suppose the semantics are unambiguous there. To make them ambiguous: 17:51
my $y; for 1..3 { my $x = $_; BEGIN { $y = { $x } } } $y() 17:52
fglock let's see - that de-BEGINs to: my $y; for 1..3 { my $x; { $y = sub{ $x++ } } } $y()
luqui there, that is sufficiently f'ed up
fglock, so you assign to $y three times?
what if $y is tied?
fglock you can use a FIRST block, I guess 17:53
luqui so I guess my point is that in at least this situation, it isn't meaningful to say { $x } 17:56
(as a closure creation)
oh, here we go:
my $y; if rand() < 0.5 { my $x = 4; BEGIN { $y = { $x } } } $y() 17:57
throw in a BEGIN { say $y } before $y() if you please
[particle] or after 17:58
18:01 weinig|bbl is now known as weinig 18:02 buetow joined 18:06 kanru joined
fglock does placing a 'ran' flag in the block helps? such that it is only run once 18:15
[particle] so, all closures get a loop counter? 18:16
18:17 ofer0 joined
fglock only for desugared BEGIN-blocks 18:17
BEGIN { $x=1+1 } desugars to { $x = 2 } unless $__global1234++ 18:18
oops - something like that 18:19
luqui see rand() example 18:20
[particle] after BEGIN is executed, it returns a closure with a reset loop counter
luqui: isn't $x undefined at BEGIN time in that example? 18:22
luqui okay...
but what happens if the random number is less that 0.5? 18:23
does $y() return 4?
and if not, $y() returns undef?
then modify to this:
my $y; for 1..rand(10) { my $x = $_; BEGIN { $y = { $x } } } $y() 18:24
now what?
I assume it will be 1 if the body ever gets executed 18:25
and undef otherwise
okay
hmm
so the question is, does such a transformation always work
oh... 18:26
so even if the block is never executed.. $y is not undef
so how do you handle that?
(this was the problem in the rand() case too)
I just distracted myself... :-) 18:27
fglock how does p5 handles this?
luqui so to clarify, $y() can return undef, but regardless $y is a code object, and has to refer to something
fglock is the 'for' pad pre-allocated? 18:28
luqui the first time you execute a block, you don't clone the pad
it's already there
it can get away from that because it doesn't separate compile time and runtime
s/from/with 18:29
[particle] what about C< BEGIN { $x } say $x >: does this complain that $x is undeclared? (no C<my>) 18:30
or is there a pad entry because it's used in begin
TimToady no, the BEGIN would fail with "undeclared". 18:31
in fact, the *compilation* of the BEGIN would fail.
luqui TimToady, did you see my proposal above?
TimToady it seems sane to the first approx 18:32
fglock does this means that the rand example is forbidden?
TimToady whether we can get the "first time through" uses compiler's pad thing to work is the real issue, I expect
[particle] and C<my $x; BEGIN { $x = 3 }> compiles?
TimToady not forbidden, just nonsensical
GIGO 18:33
I don't think we have to go to extraordinary lengths to make non-determinstic code deterministic
at some point you say, we don't penalize the innocent with the guilty
luqui well my proposal would like the rand example to be forbidden
TimToady and it's just erroneous
fglock do you think it's possible to desugar BEGIN blocks out after parsing? 18:34
TimToady no
er, yes
not sure what you're asking exactly 18:35
BEGIN blocks are executed and turned into a value, side effects have to have something at compile time to effect
the question in my mind
is whether a lexical that will be cloned
has an initial value at compile time 18:36
in the absence of an explicit =
so you could make a lexical spring into existence with a value when cloned rather than undef.
but generally,
I think the approved way to do that is not with an assignment in a BEGIN
but to say
my $x = BEGIN {...}
so I don't think we need to go to pains to try to make 18:37
for 1..3 { my $x; BEGIN { $x = ... }}
do anything special
if a decent error message can be produced 18:38
then that's better than merely declaring it erroneous
fglock I'm asking if you can substitute BEGIN {} with some run-time code, possibly derived from the env difference before and after the BEGIN was executed...
TimToady maybe something a little better than "variable will not stay shared"
luqui like: "error, example.pl line 6: you must be out of your mind"
TimToady :) 18:39
[particle] error: BEGIN inside loop construct, RETARDO!!!
luqui fglock, I've been trying to prove that you can't in general, for the last four examples
I may be mistaken, but I think the rand() example gets at the heart of it 18:40
fglock luqui: I mean, considering that you have disallowed some constructs
TimToady I agree with luqui on the general principle--a BEGIN can only influence structures that exist at compile time by side effect. Other than that, all it can do is return a value.
luqui fglock, oh, then I think you can 18:41
TimToady the "extra" question is whether there are extraordinary structures we can have at compile time that do things at run time
and I think the answer is, don't sweat it for now.
18:41 weinig_ joined
TimToady I suppose one could view the value returned by BEGIN as somehow lazy till run time, if the need arises, but I don't really see much need. 18:43
the point of BEGIN is to make something happen Right Now as far as the compiler is concerned.
fglock I'm wondering how to implement separate compile/run in p5, without executing BEGIN blocks at both 18:44
p6-on-p5
TimToady in that sense the BEGIN blocks are gone by run time
all you see is their results
fglock hmm - that's what I was meaning with 'desugaring' 18:45
luqui right. I think you can generate code to initialize all "compile-time existing" variables at the beginning of their scopes, and the semantics will be correct
(though it is indeed a bit tricky if they refer to each other) 18:46
fglock but in the case of the 'random' example, I need to do this with a BEGIN block, right?
in p5 land 18:47
luqui you mean it is impossible to factor that out into a program with no BEGIN?
I think so.
fglock think so == impossible ? 18:48
luqui yeah 18:49
TimToady are you asking if BEGIN is a primitive?
BEGIN is purposeful installation of the halting problem within the compiler. :)
18:49 rindolf joined
rindolf Hi all! 18:49
I saved a perlbug bug report to a file. Now how do I submit it using perlbug?
TimToady around here you just fix the bug yourself and check it into svn. :) 18:50
rindolf TimToady: heh.
fglock man perlbug ..?
rindolf fglock: I searched it for "save" and could not find a single occurence of the word. 18:51
integral you email it to [email@hidden.address] (iirc)
rindolf integral: OK.
18:52 kanru joined
fglock how about this: you execute the BEGIN block, and then generate code that has an INIT block in that place - the INIT block redoes the side-effects 18:53
at run time
[particle] how does it interact with INIT blocks in the source? 18:54
svnbot6 r15004 | luqui++ | Parsing of a sub definition now goes into effect before seeing
r15004 | luqui++ | the definition block. I.e. this parses:
r15004 | luqui++ | sub infix:<@@> ($x, $y) { $y @@ $x }
r15004 | luqui++ | It doesn't work with postfix operators for some reason!
rindolf integral: will /usr/sbin/sendmail [email@hidden.address] < perlbug-segfault-in-compilation.rep work? 18:55
luqui fglock, sure
so long as the side-effects are just the env differences 18:56
but [particle] has a good point
fglock if you wanted other run-time side-effects, you were supposed to put then in an INIT block, right? 18:57
rindolf Hmmm... I cannot compile perl5.
rgs: ping.
luqui my $x; INIT { $x = 1 } BEGIN { $x = 2 }; say $x;
or even: 18:58
fglock that's a problem
luqui n/m
oh dear
I was going to say that you have to move side-effects to the beginning of the scope where the affected variables were 18:59
however:
18:59 koye joined
luqui my $x; ...stuff { my $y; BEGIN { $x = { $y }; $y = { $x } } ...stuff } 18:59
ahh, pathological examples, my favorite
it is impossible to initialize $x at the beginning of its scope 19:00
ok... so, BEGIN blocks can only work on "compile-time existing variables" (this assumption is getting more and more useful)
so you don't scope them 19:01
they are all global
then you can rig them to refer to each other
(are you compiling to perl 5? I guess I don't fully understand what your target is)
but of course, that has GC problems...
fglock I'm compiling to "desugared p6", which compiles either to p5 or parrot 19:02
19:02 theorbtwo joined
[particle] you'll need weak references 19:02
fglock see mp6
luqui okay, that was enough information
hmm, so we need something where the creation of an object can claim a reference instead of creating it 19:05
this is not a standard language feature :-( 19:06
let's see here (brainstorming commencing)
all compile-time existing variables (CEVs) are globals with an extra level of indirection 19:07
fglock what's a non-CEV example? 19:08
luqui so: "my $x; BEGIN { $x = 4 };" turns into "my $__CEV_x1 = \do { my $x }; $$__CEV_x1 = 4;" 19:09
non-CEVs are things like $x in for 1..3 { my $x }
fglock k
luqui then you undef the reference at the end of its scope 19:10
so: "{ my $x; say $x }" turns into "my $__CEV_x1 ... ...; { say $$__CEV_x1; LEAVE { undefine $__CEV_x1 } } 19:11
the initialization of all CEVs happens first thing in the the program 19:12
before any other INIT blocks
I think that actually does solve the circular-dependency pathology
and the GC problem 19:13
it's just... ugly
19:13 awwaiid joined
luqui detecting the circular references could be really tricky 19:14
luqui ponders...
another pathological example
my $x; { my $y; BEGIN { $x = { $y } } }; { my $z; BEGIN { $x = { $x() + $z } } } 19:15
I know how to convert that 19:17
but coming up with an algorithm to do it, which can stand turing-completeness....
luqui wonders if it is possible just to "trace your path" in some sense 19:19
it's easy for numbers and strings and things like that which have an easy lexical interchange 19:20
but maybe you thunk closures and references
and just remember what you did to them 19:21
then do that again at the beginning of the program
I wonder if it's possible to do that without re-implementing the entire core to handle it
fglock you can save $obj.perl and eval it back 19:22
luqui if you could do it, I have to say, it would be a freaking sweet architecture
fglock, you can't do that with closures tho, right?
fglock not now, but you should be able to do it 19:23
luqui it's not a feasible problem in general, because of the inter-referentiality closures can have
oh!
but all the variables we are working with here have to be CEVs
wait... no 19:24
19:24 polettix joined
luqui my $y; my $z; BEGIN { my $x; $y = { $x }; $z = { $x } } 19:24
$x is not CEV, but you have to express that the variable that $y and $z are referencing are the same one
which is the unsolvable problem of serializing closures 19:25
fglock yes, but both belong to env, which is a single struct
you are doing a snapshot
luqui okay... how are you going to convince .perl to do that? 19:26
fglock let's see... 19:27
trying with Data::Dump::Streamer 19:28
this works: perl -MData::Dump::Streamer -e ' my $x = sub { 3 }; my $y = { a => $x, b => $x }; print Dump $y; ' 19:31
luqui the more pertinent example is: my $x; my $y = sub { $x++ }; my $z = sub { $x++ }; my $a = { y => $y, z => $z }; print Dump($a) 19:33
which, incredibly, works
fglock :)
luqui okay, DDS kicks butt
diotalevi dmq++ 19:35
luqui ponders compile-time continuations invoked at runtime
luqui ponders committing suicide, but then stops
I think it would be okay if we didn't support that... ever
fglock parrot might support that, with proper .perl 19:36
luqui uh... I'm not sure that's possible
I mean, if so, that'd be cool 19:37
but how do you start running, and then go back to compile-time?
oh
I see
well you wouldn't be doing separate compilation anymore
[particle] yep
it's all deferred to runtime
TimToady It's lazy turtles all the way down... 19:38
luqui oh, btw however, that it *is* impossible to .perl a continuation
you can serialize 19:39
them
but you can't in pure perl
(btw however, that ...; what the hell does that grammatical construct mean? :-) 19:40
TimToady It's a pessimistic Whatever. :)
hmm, then there's the optimistic sub stub: 19:42
sub foo { * }
[particle] it's good to see the optimistic case is properly huffmanized :) 19:43
TimToady say 1..foo()
luqui real question: will that make foo behave exactly like * in all cases?
TimToady should
luqui so * is not syntactic...
TimToady it's the Whatever object
luqui okay. 19:44
TimToady which we match in sigs with the Whatever type.
we had to do way with prefix:<*> to get it.
Juerd *.foo();
TimToady though you can still say *.() if you can persuade * to respond
19:45 rindolf joined
TimToady "just call something, I don't care..." 19:45
Juerd 5 + * + 5
rindolf Hi all.
TimToady "I feel lucky..."
luqui randomly generate a program and execute it
Juerd kill *, $$;
luqui you could use it to check your solution to the halting problem
TimToady kill -INT, *;
Juerd luqui++
TimToady: In typing, can Whatever be abbrev'ed to *? 19:46
TimToady generally not
Juerd sub foo (*|Int $bar) { ... }
TimToady no, * is slurpy in that context
Juerd Of course
TimToady and Whatever needs dehuffmanization
when you're writing multis 19:47
rindolf rgs: I'm unable to compile bleadperl on Mandriva Linux 2007.
luqui shouldn't | be slurpy?
Juerd So $* variables are whatever-scoped, right? :)
TimToady indeed
wherever you go, there they are.
Juerd TimToady: Can * in void context please say "Hello, World!"? 19:48
TimToady you mean |@rest instead of *@rest?
I think it wants to spit out 99 bottles of beer on the wall... 19:49
19:49 weinig_ is now known as weinig
Juerd TimToady: any(hello, bottles, quine) :P 19:49
luqui Aleph-0 bottles of beer on the wall, aleph-0 bottles of beer. Take one down, pass it around, aleph-0 bottles of beer on the wall!
TimToady a bottle of quininety-nine... 19:50
rindolf luqui: heh.
rgs rindolf: I'm not, what's the problem?
rindolf rgs: you mean, you *are* able to compile bleadperl? 19:51
rgs yes
rindolf rgs: ah OK.
rgs: let me try again.
rgs and tell me the error(s) 19:52
rindolf rgs: OK.
19:52 milan joined
[particle] * in void context is an entire language implementation (hq9+)? wow! 19:52
good april fool's idea 19:53
diotalevi Is there a concept of "stashes" in Perl 6? I tried doing a thing in perl 5 where I took a sub ref, localized the stash it came from, gave it an AUTOLOAD, then called the function to have it trigger the AUTOLOAD except it didn't work because the function didn't notice the localized stash. I want to know if I can make such a thing work in Perl 6.
Juerd [particle]: I don't see why not. 19:54
And why it would be an april fool's thing :)
luqui diotalevi, I'm not reaally sure what you're asking
Juerd If anything should be huffmanized, it's Hello, World :) 19:55
TimToady I think the compiler has to be able to optimize based on what it thinks is the symbol table.
luqui oh that, yeah I tried that too 19:56
so, what if you said at compile time that "this isn't a regular symbol table, don't optimize!"
for some tree of the symbol table
s/tree/subtree/
TimToady all things are possible, but I'd like to get Perl 6 out the door by Christmas... 19:57
diotalevi luqui: I'm asking if I can substitute the symbol table for something else at runtime. You already know it can be modified at runtime but I wanted to do the moral equivalent of local %Foo::. Perl 5 has taken a pointer to it so it loses immediately.
luqui ahh, I wanted to tie it 19:58
TimToady you need a declarative tie, which P6 presumably gives you better
luqui no, I meant in perl 7
allbery_b isn't there some discussion of this in one of the synopses already?
diotalevi Oh yeah, I think schwern said that stashes in p5 can't take tie magic. That'd solve my problem too if it worked.
TimToady not if the tie happens too late for the compiler to notice. 19:59
in p5 tie is run-time only.
you need something compile time, like "is XXX"
fglock decommute &
diotalevi Hmm. ok.
diotalevi figures he can accomplish this in Perl 5 with B::Generate's as yet unwritten B::CV::STASH( cv, new stash ) mutator method. 20:00
TimToady package Foo is Pessimized; 20:02
20:07 rhizo joined 20:24 lisppaste3 joined
rindolf rgs: I got to the test stage now. 20:29
rgs: I guess it was a one-time-bug.
rgs arf 20:31
rindolf rgs: what do you mean by "arf"? 20:34
rgs the usual onomatopea emitted by programmers confronting an heisenbug
rindolf rgs: I see. 20:39
rgs: <<< All tests successful. >>> 20:45
21:52 Aankhen`` joined 21:54 frankg joined 21:56 Schwern joined 22:06 justatheory joined 22:07 justatheory joined 22:20 koye joined 22:33 lisppaste3 joined 22:45 ashleyb joined 22:47 dduncan joined 22:58 gene9 joined 23:18 Psyche^ joined 23:34 Psyche^ is now known as Patterner 23:48 py1hon joined

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!