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.
00:58 spoop joined 01:26 fayland joined 01:27 fayland_ joined 01:31 fayland__ joined 01:48 gnuvince joined 01:53 Khisanth joined 01:57 Khisanth joined 02:00 fayland___ joined, fayland___ is now known as fayland, bonesss joined 02:13 fayland___ joined 02:19 rdice joined 02:25 kisu joined 02:32 EisnersAntonia joined 02:38 EisnersAntonia left 02:42 dmq joined 02:47 fayland____ joined, fayland____ is now known as fayland 03:07 nothingmuch joined 03:16 lyokato joined 03:19 spoop joined 03:38 kanru joined 03:42 nipotaway is now known as nipotan 04:04 spoopithy joined 04:20 nothingmuch joined, spoop joined 04:28 LordBrain joined 04:39 avarab joined 05:05 imperator joined 05:16 justatheory joined 05:26 mok joined 05:28 Aankhen`` joined 05:36 luqui joined 05:40 nipra joined 06:09 mok is now known as imperator 06:13 kanru joined 06:15 BooK joined 06:21 RHainsworth joined 06:22 RHainsworth left 07:54 cognominal joined 07:55 stef_ joined, cognominal joined 08:10 elmex joined 08:23 iblechbot joined
gaal mooses 08:24
gaal notices a moose- (well, reindeer-) related typo in spj's cool chapter on stm and beautiful haskell here: haskell.org/haskellwiki/Talk:SantaClausProblem 08:27
lambdabot Title: Talk:SantaClausProblem - HaskellWiki
08:28 pnu joined 09:11 stef_ joined 09:18 b_jonas joined 09:28 ofer1 joined 09:51 integral_ joined 09:54 larsen_ joined 09:55 integral_ is now known as integral 09:57 chris2 joined 09:59 stef_ joined 10:00 stef_ joined 10:16 nipotan is now known as nipotaway 10:24 Lorn joined 10:44 avarab is now known as avar 11:04 kanru joined 11:10 buetow joined 11:13 nekokak joined 11:20 kanru joined 11:24 BooK_ joined 11:39 turrepurre joined 11:54 iblechbot joined 12:04 jiing_ joined 12:06 autark joined 12:30 AtomicStack joined 12:40 ruoso joined 13:07 ruz joined 13:09 RHainsworth joined 13:10 RHainsworth left 13:14 dduncan left 13:36 prism joined 13:42 devogon joined 13:55 b_jonas joined 14:02 b_jonas joined 14:28 [particle] joined 14:35 vel joined 14:41 marmic joined 14:45 bonesss joined 14:53 penk joined
bonesss is away: takedown 15:33
15:47 meppl joined 15:48 nipra joined 15:53 kanru joined 16:20 kanru joined 16:24 fglock joined
fglock nothingmuch: would you provide an example on how a simple p5 program using MO should look like (a class declaration, a 'new' method, and instantiation) 16:27
the example/ files don't run here
nothingmuch the examples run like normal perl 5
due the native package emitter thingy
the registry takes care of that bit 16:28
for other things, see t/si.t
fglock I'm getting 'Can't locate Generate/PMC/File.pm'
16:28 ruoso joined
nothingmuch nothingmuch.woobling.org/cgi-bin/darcs.cgi 16:29
lambdabot Title: Darcs repositories
fglock thanks 16:30
the t/ files seem to take some time to load - can this be speed up by generating lower level code, or is it not worth the trouble? 16:31
I mean, if mp6 generated "less readable" code, would it possibly be faster? 16:33
nothingmuch the tests do no caching
otoh, if you write a small .t file for one of the example files the example .pm gets a .pmc generated 16:34
this cannot yet deal with complex class definitions or stuff like imports yet
but it's as fast as hand written code in every way
when you load $foo->meta then the original .pm file gets loaded as well, for the meta def 16:35
fglock I wonder if mp6 could generate "pmc-level" mo code
nothingmuch why bother? 16:36
MO does it for you
MO takes the meta class bits and installs them into the package
and then uses data::dump::streamer to serialize the package
and massages the output a bit
16:36 ashleyb joined
fglock one reason would be to make it easier to port MO to native mp6 16:38
because it would require less hacks 16:39
nothingmuch uh?
the data structures describing the meta classes can be serialized as YAML 16:40
and moved from p6 space to p5 space
there is no pmc-level mo stuff
there's MO::Emit::P5 which takes meta classes and flattens them into packages
which is a concept p6 doesn't even have
in the same sense 16:41
TimToady sorta by design... :)
mostly because I'm proxying for everyone who doesn't want to think about meta classes.
nothingmuch what does mp6 want to get out of using MO? 16:42
TimToady I'm glad you guys are thinking about it though.
nothingmuch less code in writing a p5 runtime with OO?
because the runtime bits of MO are opaque by design
TimToady we're trying to bootstrap the real P6 compiler.
nothingmuch well, in that case a port of MO::Compile and some of MO::Run to p6 will be fruitful 16:43
and a port of all of MO::Run + code that replaces MO::Run::Aux on the target platform
if the target platform is perl 5 then it's ready
TimToady that plus a translated v6 gets us most of the way there.
nothingmuch if it's parrot it should be fairly trivial to make a set of responder interface type stuff to talk to parrot's native O 16:44
O
then the compiler must parse classes, build meta objects at compile time, and make them into responder interfaces which are basically an OO specific compilation unit
then it serializes that
and whammo 16:45
even easier wuold be to simply serialize the data structure used to construct the meta class
and then do the whole meta calc in p5 land with my implementation
that design is flawed though
because the whole point is making the meta model separated from the runtime 16:46
makes sense?
TimToady treating P5 as a form of assembly language, yes. :)
you want to express it in something more portable?
nothingmuch =)
well
responder interfaces are in a sense dumbed down meta classes 16:47
i think it's safer to standardize a few of these
i don't like my ones so much
16:47 ashleyb joined
nothingmuch ideally the pipeline is that the compiler loads code, uses MO::Compile::* to make MO::Run::*, serializes MO::Run::* 16:48
TimToady Is there a reasonably compromise between what is possible and what is easy?
s/y/e/
nothingmuch well
the problem is that with p6-on-p5 we're trying to get a bit more goodnbess 16:49
TimToady goodness is good...
nothingmuch si si
when it's all purely MO::Run::* then the runtime simply has to support the targetted responder interfaces (natively or using a library or previously compiled code)
TimToady I mean, how close can we get to the "perfect" interface, and how much do we have to settle for a "perfectible" interface? 16:50
nothingmuch however, to get better results in terms of integration with the host language we want more info in there
the "perfect" way is perfect for systems which are not p5
because perl 5 has things like @ISA in the output 16:51
so MO::Emit::P5 digs into the meta class and then makes that more "native"
we have that
16:51 ashleyb joined
nothingmuch we can also have an extremely slow runtime which is completely "perfect" 16:51
TimToady I just don't want to get stuck with an interface that is warped towards P5 and then can't be upgraded.
nothingmuch we can also have a somewhat ugly runtime that doesn't have native perl 5 stuff
i think the wisest thing is to go in baby steps 16:52
start with MO::Compile only
and push it all to perl 5 for now
16:52 ashleyb joined
nothingmuch later we simply move MO to really live inside the compiler 16:52
TimToady we can Officially Break Anything till 6.0.0 comes out. :)
nothingmuch i doubt we'll need to
TimToady whatever gets us to the bootstrap quickest without unnecessary long-term damage... 16:53
nothingmuch this, i think
the 'damage' in a sense is more work writing other runtime
s
but as long as we stop before that starts
fglock the plan for mp6 is to compile modules to simplified-mp6, and then to native. That's how it supports grammars. I wonder how much slower this would be for MO
TimToady is simplified-mp6 simpler than mp6? 16:54
nothingmuch i doubt MO comes into play there
fglock yes, it doesn't support grammars :)
nothingmuch hmm 16:55
i don't see a problem with it being plugged after simplified-mp6
whereby mp6 with grammars contains e.g. macros for classes etc
and simplified-mp6 has BEGIN { $mo_meta_calls }
then simplified -> native means running the BEGIN stuff, and making sure the resulting data structures simply make it out of there 16:56
as long as native is still p5 that's 0 work
just serialization
if we want to support more runtimes we'll have to refactor some
sounds sane?
TimToady how much refactoring for, say, parrot?
nothingmuch fas long as we have hooks for another pass in the compiler before emitting it should be trivial 16:57
again, the diff is that instead of outputting "raw" MO structures we want to output dumber ones, which are more portable
and the translation from raw to dumbed down is MO itself, plus any custom metaclasses
16:58 ashleyb joined
TimToady basically moving from definitional toward operational, I guess... 16:58
nothingmuch that is a good way to put it
of course, operational is also pluggable
but reaching that stage the person extending operational will almost always want to do a per native-runtime port of the operational stuff 16:59
or it'll be slow
i have to go make salad
i hope i helped instead of confused ;-)
TimToady use a circularity saw on your salad. :)
nothingmuch bought a big chef knife
very sharp
TimToady avoid digits 17:00
fglock making both parrot and perl5 work at the same time would assure the design is sane...
TimToady well, saner. :)
nothingmuch hehe
TimToady I'm quite certain there are ways in which Perl and Parrot are similarly misguided. :)
nothingmuch that's still a bit further off... will you guys be here in about 1 hour to discuss? 17:01
TimToady might be commuting about then, but likely around give or take 10 minutes.
nothingmuch strongly believes that his model will work well in real ASM, his second (or maybe third) love
okies.. I'll be back 17:02
TimToady compiling P6 to straight ASM would be cool.
fglock I'll have a meeting soon, then will be around until 3h from now
I was considering an asm output to mp6, but I guess it's better to start with C first 17:03
and I think audreyt said she would work on the Haskell backend 17:05
pugs.blogs.com/photos/visiolization...strap.html - shows how the grammar implementation plugs back into mp6 17:10
lambdabot Title: Visiolization: Mp6bootstrap
17:20 justatheory joined, wamiks joined 17:33 diotalevi joined
diotalevi So is there syntax in Perl 6 yet for saying "this next bit is in prolog?" 17:39
fglock I think it's like: eval "...", :lang<prolog> 17:44
diotalevi Oh bummer. Bummer. Bummer. bummer.
Well... eval <<'...', :lang<prolog> wouldn't be so bad though. Maybe? 17:45
No wait.... it wouldn't be piles of awesome. I'm back to 'bummer.' 17:46
allbery_b hm, I would have thought there would be something like: grammar Prolog { ... } (...) { use Prolog; ... } (akin to "use perl5;") 17:50
TimToady macro this_next_bit_is_in_prolog ($ast) is parsed(Prolog) {...}
diotalevi Oh. Well happy day then.
TimToady Perl 6 is not a language. Perl 6 is many languages. 17:57
nothingmuch btw, i'd like there to be an "official" micro perl 6
which contains just enough features to bootstrap extensible grammars, and the notion of BEGIN etc 17:58
diotalevi only wanted the Perl 6 that'd let him speak a few languages simultaneously (or at least adjacently)
nothingmuch that is, the thing without all the "just a macro" features
which is no fun to work with, but is well defined and easier to port 17:59
TimToady how is this different from mp6?
allbery_b "official"?
nothingmuch i thought fglock said a bit earlier that the "simpler" form has no extensible grammars? 18:00
that said, I have another answer: the purpose is slightly different
miniperl 6 is mini enough to bootstrap
but is not mini out of minimalism
fglock it is bootstrapped
yes 18:01
nothingmuch we could milk a bit more mini out of it, theoretically
and the reason *I* am interested is that this language interests me
because it's the mother of all meta languages ;-)
TimToady seems like the opposite approach to a circularity saw, but maybe you'd write your circularity saw in this eventually...
nothingmuch it hasn't much to do with it 18:02
it's like a "For later" thing
it would just be a nice spec layer
for the language underlying all the ergonomics that is perl 6
TimToady the problem with designing meta stuff is you never quite know what you'll want to be meta about later...
nothingmuch well, the language *is* in there 18:03
macro processing, BEGIN { }, the notion of subroutines and data types
but on the other hand I don't think that e.g. gather/take is part of that
TimToady and I think the circularity saw is related to that distinction 18:04
18:04 Lorn_ joined
nothingmuch it is related, but it's purpose is different 18:04
because gather/take is very useful for writing a perl 6 compiler 18:05
but gather/take is not very useful when you're extending a core language to make a language that is not anything like perl 6
you'd have many bits to take away first
TimToady I don't see how gather/take would be useful for writing P6 and not other things too. 18:06
fglock did you see miniperl6.spec ?
nothingmuch TimToady: because by design when you make a DSL in perl you are normally "subclassing" the perl 6 grammar 18:07
but for writing non perl 6 related tools maybe you'd rather have a higher base class
fglock: no, I haven't
wow, my sister is really talented 18:12
today she didn't come home from school
so i went looking for her
in the rain
now my shoes were drying in the bathroom
and to thank me she spilled more water into them
*sob*
diotalevi Oof! 18:13
What a nice sister.
TimToady I never had a sister...
18:14 BooK joined 18:16 Lorn joined
fglock nothingmuch: svn.pugscode.org/pugs/v6/v6-MiniPer...6-spec.pod 18:21
18:30 Lorn_ joined
TimToady commuting & 18:34
fglock nothingmuch: so do you have a use for mp6? 18:51
nothingmuch not now
but I will be studying at some point in my future, i'd like a nice putty language to make some weird ones with ;-) 18:52
19:31 b_jonas joined 19:53 koye joined 20:08 Lorn joined 20:09 polettix joined 20:10 DaGo joined 20:32 Lorn joined 20:36 marmic joined 20:37 Dr_Pi joined
Dr_Pi Would I be mistaken if I said pugs is the most important software development project written in Haskell? 20:39
Tene important to who? 20:40
integral
.oO( some important bits of Pugs are written in Perl 6
Dr_Pi Maybe significant might be a better word. 20:41
kolibrie darcs might be considered an important software development project written in Haskell
and is part of the reason pugs uses Haskell (IIRC) 20:42
stevan Dr_Pi: Actually GHC is written in Haskell 20:50
That would probably be the most signifigant, since it compiles Haskell :) 20:51
20:51 Lorn_ joined 21:11 elmex joined 21:13 weinig is now known as weinig_, weinig_ is now known as weinig 21:15 Aankhen`` joined
allbery_b ghc is certainly the largest. most significant? IIRC Deutche Bank hires Haskell programmers --- it occurs to me a financial application for a multinational bank is potentially more significant than ny of the above :) 21:25
21:28 Lorn joined 21:33 weinig is now known as weinig|bbl 21:40 DaGo joined 21:44 Lorn_ joined
stevan allbery_b: that assumes that they use Haskell in a production environment for a financial app 21:46
just cause they are hiring, doesnt mean they are running the bank with it :)
and of course,.. if they use GHC to compile their Haskell, then it seems we might be back full circle
22:10 xpika joined 22:49 stevan joined 22:53 elmex joined 23:05 weinig|bbl is now known as weinig 23:07 frankg joined 23:09 stevan_ joined 23:11 dduncan joined 23:18 Psyche^ joined 23:32 Psyche^ is now known as Patterner

Powered by Cro and the Raku® Programming Language.

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

Thank you!