Run Perl6 Now! & pugscode.org ('Overview', 'Journal') & pugs.kwiki.org & channel log: xrl.us/e98m & < autrijus> damn, my journal keeps getting obsoleted by #perl6.
Set by Juerd_ on 12 March 2005.
nothingmuch new file::Spec fixed it 00:03
GRR! 00:09
r955
966
jabbot pugs - 966 - Create backlinks in synopses 00:17
pugs - 969 - refactor 00:47
pugs - 968 - Kwidi (2html) handles paragraphs, header
pugs - 967 - Document the backlink index
stevan hey nothingmuch, theorbtwo 01:07
theorbtwo is not-so-much here.
stevan theorbtwo: before you fade away,.. is there a pugs equivalent to C<perl -I lib/> 01:09
theorbtwo PERL6LIB=lib pugs ... 01:16
stevan hmm
ok
that should work
thanks :) 01:17
ingy hi stevan 01:20
stevan hey ingy
ingy it is insanely windy here 01:21
stevan no rain?
oh wait
you are in Taiwan?
ingy Seattle 01:22
stevan ah,.. no rain?
ingy not right now
stevan when are you going to Taiwan?
any regexp people out there? 02:09
Alias_ "regexp people"? 02:10
stevan people good at regular expressions
Alias_ I'll do as a start
stevan what does this mean \Z(?!\n) 02:11
I read it as:
match the end of a string which is not followed by a newline
is that right?
Alias_ yes
Although it seems a little nonsensical
stevan yeah it does
its from File::Spec :) 02:12
Alias_ then it's probably "smarter than you"
:)
stevan ?
Alias_ It might mean something in a multistring context
as in not //s
"The end (in multi mode), but not because we hit a newline 02:13
stevan this is the whole one
!/^\.{1,2}\Z(?!\n)/s
Alias_ unusual..
It's just a "ignore the . and .. files" 02:14
stevan yeah from no_upwards
I am thinking of refactoring it into a junction
Alias_ I think it's a very-safe format...
stevan $file ne ('.' & '..')
Alias_ that would probably be fine
stevan cool
Alias_ assuming it isn't slower
stevan we shall see 02:15
:)
Junctions are actually a very cool feature 02:16
Alias_ I'll be much happier when someone demonstrates auto-multi-threading is viable 02:17
stevan I agree, but they are still a nice feature
Alias_ I was relatively anti-perl6 until Damian mentioned that junctions will be parralelisable
It's possibly the first thing in Perl6 that is truly shiny
stevan things like ($file ne '' && $dir ne '') into ('' ne ($file & $dir)) 02:18
Alias_ purl, cpan scoreboard 02:21
ack
chip \z is the real EOS. \Z may match before a newline. So \Z(?!\n) is a synonym for \z
stevan chip, one more question if you have a second 02:27
chip sure
stevan would this regexp !/^\.{1,2}\Z(?!\n)/s
chip that's equivalent to $_ ne '.' && $_ ne '..'
stevan yes
thanks
so refactoring it into a junction would work then 02:28
Alias_ yes
stevan excellent, 2 'yes's is all I need :)
chip $_ ne ('.'&'..') ?
stevan @filenames.grep:{ $_ ne ('.' & '..') } 02:29
Alias_ steven
stevan Alias_
Alias_ steven: You can junction that too...
if ( any( @filesnames ) eq ('.' & '..') ) { 02:30
stevan Alias_: very cool
Alias_ actually...
ingy stevan: in 32 hours
Alias_ I wonder if lists have an .any method that creates a junction for them
if ( @list.any eq (' 02:31
etc.
stevan Alias_: however that doesnt return the ones which didnt match
ingy: cool
Alias_ do you care?
stevan Alias_: yes, that is what the function does
Alias_ oh
stevan File::Spec::no_upwards
it removed the . and ..
Alias_ So it's a filter and not a test... right
stevan yes
Alias_ then grep is indeed what you want 02:32
stevan but that junction would work
although I dont think that lists have .any, .all, etc
at least I have not seen it in my Synopsis reading
jabbot pugs - 970 - begun a perl6 File::Spec hack, it is slo 02:47
ayrnieu Perl6 will still have HERE docs, yes? 02:57
Alias_ yes, but better 03:01
tabbing ignored, from memory
so you can indent them properly
ayrnieu yay! 03:03
ingy which is problematic with tabs 03:06
jabbot pugs - 971 - Split Kwid test data from test code. 03:07
ingy unless tabs are preconverted to 8 byte boundaries, which is also slightly problematic
because what if some of the leading tabs are content and some are indentation 03:08
bd_ IIRC, it senses the indentation of the ending delimiter and strips that off the preceding lines
ingy bd_: which is about the worst way to do it 03:09
because that makes it very fragile and hard to debug
I think it would need to be smarter heuristic than that 03:10
bd_ like what?
ingy I've dealt with this heavily in the YAML spec
which is similar
well you need to detect the end marker, and then expand the tabs in the indentation and then expand just enough tabs in the body lines and then strip off that many spaces from each 03:12
and then what do you do if one of the lines is less indented 03:13
warn, error, or look for another less indented end marker
? 03:14
bd_ There's also the problem of inconsistent usage of spaces and tabs, and differing tab width settings
ingy bd_: that's what I'm talking about
bd_ presumably if it doesn't find enough indentation it should either warn and treat it as an empty line or error
looking for another end marker would be confusing
ingy tabs need to be always considered 8byte expanding
and a lot of M$ tools default to 4 03:15
so there is really no win here
bd_ warn if tabs are used at all? :)
ingy YAML wins by outlawing tabs altogether
bd_: that is by far the best choice
YAML wins by outlawing tabs altogether (for indentation) 03:16
tabs are truly evil in this regard
I spent 3 months discussing these issues
I would simply not allow tabs for indented heredocs 03:17
but many will cry foul
oh well
bd_ maybe warn unless a use tabs width => 8; or similar is present?
ingy damian uses more tabs than the devil herself
bd_ additionally, if the heredoc always uses tabs consistently, there's no need to warn 03:18
it's only if it mixes tabs and spaces in the prefix it needs to warn
ingy yep
jabbot pugs - 972 - A failing test for lists 03:47
autrijus stevan++ 03:50
nothingmuch++
ingy++
gaal++ 03:51
ingy :P 03:52
autrijus more than a dozen commits whilst I was asleep :)
28 more to go till r1000 03:53
ingy pushing for 1000
jinx
autrijus jinx
ingy autrijus: it's harder now that MANIFEST isn't in there ;)
MANIFEST was probably 10% of checkins :D 03:54
autrijus actually less than 50 03:56
so 5% is more like it
(but if you count the checkins that contains MANIFEST with other real changes too, then it's 92) 03:57
lucs Does 'perl6doc' or 'pugsdoc' exist yet? 05:23
ayrnieu lucs - no, but see 'perldoc Perl6::Pugs' in the latest release. 05:27
s/no/NAFAIK/
lucs , thanks.
Er, OK, thanks.
ayrnieu: Er, what am I supposed to see in Perl6::Pugs related to the doc format? 05:30
mugwump I'm sitting in the POD café, working on that right now in fact :) 05:31
ayrnieu ... that I didn't realise that you wanted to know about a perldoc for kwid.
lucs Ah :)
cafAc? 05:32
mugwump cafe, encoding problems I guess
lucs Oh, makes sense.
revdiablo I remember seeing someone in here paste the url to a list of perl6 operators generated by a script a week or so ago. anyone know what I'm talking about? 05:33
mugwump the periodic table of perl 6 operators, perhaps? that was a lot older though 05:34
revdiablo no, it was an HTML table
it's driving me crazy, because I remember looking at it, but not much else about it. :-/ 05:35
lucs revdiablo: I remember that too. Maybe you can check the channel logs?
(See topic)
revdiablo I looked through the logs... couldn't think of what to search for
found it: desert-island.dynodns.net/perl/prims.html 05:39
nothingmuch work & 06:23
dvergin 'Nite all. See you again soon. 06:34
nothingmuch morning 06:54
lightstep so partialApply or curry? 08:59
nothingmuch ... continued discussion in public chat, so it'd be logged
about the implementation of currying
curry's [VAL]
lightstep it's the partial parameter list
but the return value should be a Val, not a VSub 09:00
so it'd be a thunk
nothingmuch where does the parameter list get into doApply? brb, going to read this a second
because doApply already has it's args split up 09:01
to invocants and args
line 449 of Eval.hs
lightstep i think we should use apply, not doApply 09:03
nothingmuch sorry, keep getting distracted 09:04
let [subExp, Syn "invs" invs, Syn "args" args] = exp
how does this know to split it up?
lightstep magically
it sees that the list is of length 2, and that the 2nd and 3rd elements have the required form, and vinds 09:05
nothingmuch hah
lightstep if exp isn't perfect, the program crashes
i don't know how to create a "wildcard" sub, that would catch everything
nothingmuch ok, i think we don't want that anyway 09:06
because that would make curries of curries unverified
at .assuming time
i think we construct a new VSub whose params are those that have not yet been bound
lightstep oh
nothingmuch that way implementation is more, err, "correct" 09:07
i think
perhaps more work, but i'm not sure if harder
lightstep yeah
we'lll take out those the sindSomeParams took?
nothingmuch uhuh
i think we might want to make the state a bit smarter
make it a pair
([ <binding> ], [remaining]]) 09:08
wow, i'm so consistent ;-)
lightstep it doesn't have to
nothingmuch ([(Param, Exp)],[Param])
im an oop head, so that seems simple to me =)
lightstep we can make pugs ineffcient but working, without touching bindSomeParams
nothingmuch i think that it doesn' 09:09
lightstep filtering them out ourselves
nothingmuch t really matter
bindSomeParams is very sterile
it's new, and it's not used yet
ca. svn up an hour ago
lightstep ok 09:10
nothingmuch so i think it's better we get it right
because we can't be cheapish either, with takeWhile etc
the process needs to be more like a merge sort, since optional params may be litterred in there
one minute, test run at work needs prodding 09:11
lightstep lists are not a good data structure for this
nothingmuch actually, why not? 09:13
they are parted in O(N) time
lightstep yes
Control.Monad.Writer seems cool 09:14
but i din't know if it would be the mose useful here
nothingmuch it looks scary for me 09:15
i think O(N) shouldn't worry us now
but now anything but slurpy should be OK
since lists are lists
in the past passing sub(@biglist) would have been heavy in this case
perhaps the bindThing functions should be made more closed boxy
and they take a Binding type, whatever that is 09:16
and create the bindings themselves
but i think this optimisation is a bit big for now ;-)
so anyway =)
lightstep since bindArray and bindHash don't return the non-bound parameters, i don't see how bindSomeParams can do it prettily 09:17
btw, x <- return y is the same as let x = y 09:18
nothingmuch i think for now we just shove a filter into bindSomeParams 09:21
and take care of it later
lightstep brb
nothingmuch but we shove the complexity down one layer
arguably a good thing, since it is not our concern
elmex hi 09:22
nothingmuch hi!
lambdaheads: are there any set operations in haskell, somewhere, like prelude? 09:24
i'd like to intersect two lists
malaire have you checked Data.List ? 09:26
there's also Data.Set (in GHC 6.4 at least) 09:27
nothingmuch i'll look at those 09:28
malaire there is Data.Set.intersection, and Data.Set.fromList to create Sets from lists 09:29
nothingmuch yeah, it looks good 09:30
malaire++
malaire toList $ intersection (fromList [1,2,3]) (fromList [2,3,4]) -- and finally toList to get list back 09:32
lightstep ack
also List.intersect
nothingmuch Data.List intersect seems good 09:33
let binding = concat [prebound, boundInv, boundNamed, boundReq, boundOpt, boundHash, boundArray, boundScalar]
let remaining = (map fst inding) `intersect` params
does that look sane?
return $ (binding, remaining)
s/inding/binding/; 09:34
malaire Haskell has far too many libraries... - One needs to learn quite many of them to use Haskell efficiently (Quite like Java :)
lightstep no sane 09:35
you want map fst binding \\ params
nothingmuch malaire: it's nicer than having a set of square wheels which you wrote yourself, handy for every situation ;-)
lightstep difference, not intersection
err, the other way around
params \\ map fst binding
nothingmuch oh, right 09:36
what we already have is the intersection
*blush*
lightstep so SubRoutine or SubBlock? 09:37
nothingmuch the curried return?
i think it's a subRoutine
lightstep and the name? "<anon>"? "oldSubName-curried"? 09:38
nothingmuch anon
you can have a bajillion curries of curries of the same sub 09:39
or maybe oldSubName, period
lightstep ick, multisubs 09:40
nothingmuch indeed ick
i don't think we want to deal with type disambiguation yet
lightstep howcome they have a parameter list?
nothingmuch what do you mean? 09:41
lightstep do all have the same number of parameters?
nothingmuch all multisub?
s
they have several invocants, and n parameters
lightstep do all multisubs with the same name have the same number of parameters and the same return context?
nothingmuch no
all multisubs have a distinct signature 09:42
which is arguably the name
lightstep so Sub is not good
nothingmuch and is the parameters, the invocants (with type constraints), the parameters, and the return context
isn't Sub capable of that?
or do you mean not good in the 'is wrong' sense, instead of the 'won't work here' sense 09:43
lightstep or do we have separate Sub objects for the different subs in a multisub?
nothingmuch i have no clue
lightstep perlbot: nopaste 09:44
perlbot Paste your code here and #<channel> will be able to view it: sial.org/pbot/<channel>
pasteling "lightstep" at 212.235.51.245 pasted "some code to facilitate the discussion" (12 lines, 288B) at sial.org/pbot/8432
nothingmuch i think it would look like, err 09:45
copying isMulti, and copying subAssoc, and subReturns
etc etc 09:46
maybe even subName
lightstep ok, cool
nothingmuch override only subType, perhaps, i don't know what it means, but that's where we're sticking things into, right?
and override subParams with the snd (bindSomeParams ....)
pasteling "lightstep" at 212.235.51.245 pasted "some code to facilitate the discussion" (7 lines, 153B) at sial.org/pbot/8433
nothingmuch and make it's subFun a (\x -> bindSomeParams ... ) 09:47
with the original params and the partial binding
lightstep oh, yes, i forgot the subFun
nothingmuch isn't it supposed to return a Val?
which is actually, err, just VSub in another type? 09:48
lightstep i thought about it
and i was mistaken: we create a new sub from an old sub, in a pure operation
nothingmuch i'd like an explanation of each one's pro/con, i don't think i understand them yet 09:49
lightstep the thunk is in subFun, not the result of currySub
nothingmuch what is a thunk?
lightstep we get a sub
a thunk is a value that holds a computation to be run later
(the computation is only thunk when creating the value, not performed) 09:50
nothingmuch ah
a sort of self-evaluating expr?
ah
hehe
lightstep almost. more like an expr-containing value
nothingmuch this is to facilitate the concept of imperative execution in a lazy language?
lightstep yes 09:51
nothingmuch ok,
so why is our currying not a sub?
lightstep and also useful for p6 infinite lists
nothingmuch thunk
lightstep the result of currySub is a regular sub. being a sub delays computation by itself (like \x -> sib 6 + 8) delays the computation 09:52
nothingmuch so if we make it a thunk it will delay the ...
no op that would be returning the sub?
lightstep returning a thunk will cause the delay of the (pure) creation of the Sub object 09:53
yes
nothingmuch ok
lightstep VThunk (MkThunk (return x)) == x
nothingmuch what exactly is subType? 09:55
lightstep SubBlock or SubRoutine (also 2 more, in AST.hs)
nothingmuch what's the difference in meaning? 09:56
lightstep dunno yet
nothingmuch bbiab
food for thought: 09:57
doApply :: Env -> VSub -> [Exp] -> [Exp] -> Eval Val
apply :: VSub -> [Exp] -> [Exp] -> Eval Val
how do these types related to currySub?
obviously env is delayed
but invs, and args should be split up
*poof*
ok, back 10:04
lightstep so how is the splitting done? 10:05
nothingmuch i have no clue
but it's done somewhere in reduce
line 449 of Eval.hs 10:06
i have a speculation
that they are already split before that line
brb, someone needs help compiling 10:07
lightstep Parser.hs must be it. it's the one that creates Exps
processFormaps, line 714 of Parser.hs 10:09
pasteling "lightstep" at 212.235.51.245 pasted "another version" (39 lines, 1.5K) at sial.org/pbot/8434 10:25
nothingmuch by haking doApply, you mean breaking it to two pieces? 10:30
i'll do that if you want
damn, brb agian
lightstep yes 10:33
but in the paste, there is already the change to doApply
except my signature of doApplyBinding is broken (it should be [(Param, Exp)] -> Env -> VSub -> Eval Val 10:34
and probably there's an unuse parameter there too
but currySub needs the new, improved, futuristic bindSomeParams 10:35
in addition to a name change
and splitParams should be done too, ut this is a change in Parse.hs 10:38
Parser.hs, rather
nothingmuch ug 10:41
i have the new improved bindSomeParams implementeds
what is splitParams?
the invs<->args partition?
lightstep yes 10:44
it's really funny, Parser.hs works with 2-element lists instead of pair
bad, bad autrijus
so you'll commint, and i'll do the Parser.hs change? 10:47
nothingmuch i think 10:50
i've just become busy
the story:
so i'll do it when i'm done
our OSF box died several months ago
now i have to port all the changes to the product from the last 3 months
and I just hit a big change
lightstep ok
should all the test pass? 11:01
elmex yaho 11:47
elmex dances a little bit
jabbot pugs - 975 - r1424@speights: samv | 2005-03-21 22:4 11:49
pugs - 974 - * Refresh AUTHORS and META.yml; bump ver
pugs - 973 - * Module::Install now groks vstrings, so
elmex begins to cry and hit his head agains the table
jabbot++ 11:50
nothingmuch you seem err, a bit, bipolar
elmex?
elmex !?
i love pugs 11:51
i wonder when the string semantics will be discussed and finished 11:56
jabbot pugs - 976 - fix 2-element lists to pairs 12:09
nothingmuch lightstep: i'm having some issues 12:10
perlbot nopaste 12:11
perlbot Paste your code here and #<channel> will be able to view it: sial.org/pbot/<channel>
pasteling "nothingmuch" at 212.143.92.226 pasted "currySub" (63 lines, 2.7K) at sial.org/pbot/8437
nothingmuch this is as far as I got
binding' <- verifyBindings was done
but it needs to use the params of the original underlying sub
that is, perhaps we need a curriedSub value type, that contains the subParams of the bottomest one 12:12
alternatively, we can do, as you proposed, the list difference at the curry level
always keep the subParams list as the original
and then bindSomeParams binding (params // (map fst binding)) 12:13
in the closure itself
as for splitParams params - i don't know what it does
i think it's not our responsibility
we should delegate to whoever does it for apply 12:14
lightstep yeah 12:15
i don't think we need a curried sub type
after all, it's indistinguishable from a block
nothingmuch ok, so i'm moving the param difference into the curry apply
lightstep the distinguishing is based on "appearence" 12:17
the line 711 in Parser.hs
there, it parses a list of lists of names
like a1, a2, a3 : b1, b2, b3 : c1, c2, c3 12:18
processFormals ensures there are only 2 12:19
nothingmuch that's for separating invs from args, yes?
lightstep yes 12:20
nothingmuch btw, why not currySub :: VSub -> [Exp] -> [Exp] -> MaybeError VSub, but [Val] -> [Val]? 12:22
shyte
when currying curried subs we need access either to the already bound list 12:23
theorbtwo Mornin, all.
nothingmuch or to the remaining params
hola theorbtwo
currySub sub invs args = do
binding <- bindSomeParams [] (subParams sub) (map Val invs) (map Val args)
lightstep nothingmuch, not really
nothingmuch this needs to keep binding on top of the partial binding
if wrapping a wrapped sub
if currying a curried sub 12:24
because there's the default value issue
lightstep you don't care how the sub you're currying was constructed
you only need its signature
doesn't bindSomeParams handle default values? 12:26
nothingmuch err, wait 12:27
i mean things like foo($x, +$y = $x)
this becomes very tricky
lightstep why?
jabbot lightstep: why is the make command 'ghc --make -o pugs src/Main.hs pcre.o'
nothingmuch tricksy nasssty larry
lightstep: err 12:28
lightstep jabbot: forget why
jabbot lightstep: ok
nothingmuch lightstep: /ignore jabbot
=)
lightstep i though that calling bindSomeParams with few exprs, then with some more, then some more, etc, is equivalent to calling it with the concatenated list 12:29
in which case, the doubly-curried sub can be calledwith the first assumed arg list, then the second, then the real one that the user specified 12:32
nothingmuch err, i'mnot sure 12:36
let me think about it a while
nothingmuch is hating work today 12:41
lightstep: can you figure out where ($x, +$y = $x) is done? 12:48
lightstep can you clarify? 12:50
is it a parameter list?
nothingmuch yes
sub foo ($x, +$y = $x)
means that $x is mandatory 12:51
and $y is optional
and it's default value is $x
autrijus yo!
I just experimented with unoptimized and optimized compiler backends
nothingmuch autrijus, you can answer this
autrijus with mandel.p6
nothingmuch sounds interesting 12:52
lightstep ruleSubParam (Parser.hs, line 193) calls ruleFormalParam, in which `+' is parsed
is mandel.p6 fast?
nothingmuch i don't mean the parsing, i mean the binding
autrijus Eval.hs line 677 12:53
Parens exp -> local fixEnv $ enterLex pad $ expToVal prm exp
that does the defaulting.
ok. some numbers (identical output)
[not|autrijus]~/work/pugs/out$ time ./interpreted.out
69.928u 0.233s 1:17.29 90.7% 2370+477k 0+0io 0pf+0w
[not|autrijus]~/work/pugs/out$ time ./unoptimized.out
0.685u 0.007s 0:01.35 50.3% 2392+476k 0+0io 0pf+0w
[not|autrijus]~/work/pugs/out$ time ./optimized.out
0.170u 0.008s 0:00.39 43.5% 2339+455k 0+0io 0pf+0w
nothingmuch ah
so that's done by apply, and we don't care
autrijus yup. 12:54
I'm pretty psyched by the numbers.
nothingmuch lightstep++ # was right
my god
lightstep autrijus++
lumi Hi 12:55
nothingmuch hola lumi
lumi And wow
nothingmuch this is the haskell backend?
autrijus yes.
I can't do parrot backend without help. 12:56
nothingmuch could you explain how it stores the AST?
autrijus and it's all still in the Eval monad.
nothingmuch lightstep: why env <- get in the currySub's returned closure? 12:57
autrijus (meaning, could be faster)
nothingmuch how does this compare to p5?
autrijus anyone want to do mandel.p5? 12:58
ozone autrijus: pwhoar
autrijus nvm I'll do it
ozone autrijus: you're outputting haskell, and compiling with GHC?
masak what do the numbers mean?
something unoptimized is 1000x faster than something interpreted
autrijus ozone: yeah 12:59
masak but what?
jabbot masak: but what is replacing it is really just Pod ;)
lightstep nothingmuch, 'caue doApplyBinding has an Env as a parameter. we can't call apply which'll do it for us, since is always applies everything
nothingmuch masak: pugscc outputs
autrijus [not|autrijus]~/work/pugs/examples$ time perl p5
0.061u 0.000s 0:00.14 42.8% 9+362k 0+0io 0pf+0w
nothingmuch ah, ok
autrijus so, 3x slower than p5.
lightstep nothingmuch, i thought that this was the best point to break the sequence (since apply and doApply are already used)
autrijus (but that's because we're still in the slowish Eval monad)
ozone autrijus: are you using any GHC exts, like unboxed ints etc? 13:00
autrijus (and still using Val)
ozone: no, not at all, and all the vals are still dynamically casted
ozone that's mighty impressive, even though it's sort of expected :)
autrijus i.e. no type annotations
nothingmuch that's pretty good
autrijus autrijus.org/mandel.ast
ozone autrijus: sheesh 13:01
autrijus: you might have a very good parrot vs ghc fight on your hands later :)
autrijus heh heh :)
I'm not really using primitive haskell :)
I'm still using the "op2" etc in Prim.hs
which means the semantics are guaranteed to be right 13:02
but also means it's slow.
also newsym generation needs work
can't just replace sigils with _
ozone autrijus: right 13:03
if you want speed, talk to dons
he'll be very happy to help
autrijus ok.
ozone he's done far too much evil stuff in GHC with unboxed types and unsafeCoerce#
autrijus but considering this is a two-hour hack or so
I'm pretty happy already
ozone autrijus: very impressive 13:04
then again, it is in haskell i suppose :) 13:05
(not to take away any credit from you; you know what i mean)
autrijus yup :)
autrijus hugs [| |] 13:06
ozone what are you using QQs for?
autrijus TH is really da bomb.
uh, code generation.
ozone ah!
autrijus haskell code generation.
ozone ah ah
of course
autrijus of course!
ozone it all makes sense now
autrijus compile (Val (VRat v)) = [| VRat $v |]
autrijus grins
ozone autrijus: you really should submit a paper to haskell workshop
if they don't accept it, i'll beat them personally 13:07
autrijus ok. I will learn how to write ACM format papers :)
lightstep to TMR 13:08
ozone autrijus: eh, if you can write a perl 6 compiler, you can write an ACM format paper ;)
autrijus those require different capabilities :) 13:09
but this is not really a perl 6 compiler. it's more of a static perl 6 compiler.
we don't do BEGIN{} or parser reshaping yet.
that will be terribly difficult.
would need more caffeine.
ozone suspects autrijus is dons's long lost twin
autrijus heh heh :) 13:10
autrijus reads backlogs 13:12
nothingmuch bah 13:14
shapr I feel a disturbance in the force... 13:15
what happened?
nothingmuch hello shapr
i think i have a solution
shapr hiya nothingmuch
nothingmuch pugscc haskell output is nearly as fast as p5
when optimized
shapr I have a problem, you have a solution, let's have lunch!
nothingmuch i have a solution to many of my own problems ;-)
err, that didn't come out right 13:16
i have one solution, many problems
anyway, lightstep, autrijus
lightstep yep?
nothingmuch currySub returns a sub { subType = subCurried, subFun = Curry fun
applyExp bound (Curry f) = f bound 13:17
then we don't have to resplit bound params
it becomes a simple iteration
now, even funnier is to do it at apply on a curried sub 13:18
lightstep but then you lose all parameter-checking
nothingmuch and let there be curry folding without creating new pads
no i don't
f bound does that
and eventually it calls doApply with the real sub
and just before it calls verifyBindings
i'll try to draft this up 13:19
lightstep bbiab
nothingmuch how do i do type matching on a sub's subType at the function level? 13:20
ciao
anyone? 13:22
shapr alexe: are you really irc'ing as root? 13:23
alexe ? 13:25
autrijus 21:22 -!- alexe [~root@221.0.31.74] has joined #perl6
shapr It's safer to irc as a plain old user. 13:26
alexe thanks
but which irc client has security problem? 13:27
nothingmuch alexe: it's conceivable, and even more probable 13:28
since people won't bother securing IRC clients as much as servers
Limbic_Region autrijus - www.perlmonks.org/index.pl?node_id=441199 # just a press release from me on your press release 13:32
autrijus thank you. 13:33
stevan morning all
nothingmuch hola stevan
stevan hey nadda-mucho
autrijus stevan: check out the numbers! :) 13:34
stevan the speed?
I just saw that
stevan reads the backlogs with coffee every morning :)
autrijus =)
stevan very very very vool
autrijus++ # as if you need any more :P
gaal i get a 25% speed improvement on 'make test' after 'make optimized'. 13:35
autrijus that actually means pugs is ok for production use, speed-wise, for my clients :)
shapr w00
shapr waves the lambda and camel pom-poms
nothingmuch ponders a camel posing as a lambda as the p6 logo
castaway lurks
theorbtwo Allo, love.
stevan autrijus: all I need now is mod_pugs and I will be happy to use it in prod too :) 13:36
nothingmuch it could put it's head in a funny way
shapr shapes
castaway hugs theorbtwo.
Just wondering where you're active :)
autrijus stevan: there is a mod_haskell :)
so... there ya go :)
stevan autrijus: yummy
theorbtwo Head looking up, leaning on it's front legs, with rear legs placed inline with the body.
...would look like a lambda with a hump. 13:37
theorbtwo is active nowhere at present.
shapr theorbtwo: can you sketch that?
oh, I think I get it.
nothingmuch theorbtwo: uhuh, sort of
theorbtwo Think of a human leaning on their arms and stretching their back, then translate.
nothingmuch or nicer
shapr yeah, I think I got it now.
Not sure I can draw it though. 13:38
theorbtwo can't draw.
shapr I've always been able to draw, dunno why. I've always been able to sing too. Programming has always been hard for me, that's why it's more fun.
stevan autrijus: I started working on oo tests *hint* *hint* 13:39
jabbot pugs - 978 - * convert more things from Prim to from
pugs - 977 - goto tests parsefail again
castaway different brain, shapr
autrijus stevan: sure :)
theorbtwo Oh, whoever was looking at the prims.html on d-i, it's produced with utils/grokprims.pl
shapr castaway: yeah, probably has something to do with the ritalin prescription as well :-) 13:40
Programming is the best of everything, it's like executable poetry.
stevan autrijus: I am also talking/working with Darren Duncan to write more
autrijus eggcelent! 13:41
on 23th mugwump and ingy will arrive
shapr Programming is the magic executable fridge poetry, it is machines made of thought, fueled by ideas.
stevan my plan is to have a fully loaded OO test suite before you even begin work on that part
autrijus and we'll work on pugs object metamodel :)
shapr Or maybe made of ideas, fueled by thought? whatever... fueled by hard work, which I should start doing now. 13:42
kungfuftr mugwump's going to tapei... yeah?
hcchien yes
autrijus yup 13:43
kungfuftr ah, give him a nice big slap from me... the goit
stevan autrijus: on a totally different note, I have found an issue while port/hacking the FileSpec stuff 13:44
pugs -e 'sub foo (*@f) { return pop(@f) }; say foo(1, 2, 3);'
gives me this error: Fail: cannot modify a constant item: VArray (MkArray [VInt 1,VInt 2,VInt 3])
autrijus yup
I really need to get IType done.
stevan ok 13:45
also, should I un-TODO the tests again?
I have some time to kill before $work actually starts
autrijus yup, please do so
stevan ok
autrijus and audit todo_s that you feel are 1)wrong 2)important a bit
stevan ok 13:46
lightstep wtf? SubType's order really matters? 13:47
does it mean anything?
autrijus lightstep: well, yeah 13:48
lightstep "more named"?
autrijus there's bares, pointys, subs
and each has different semantics
the first two doesn't even havescope
nothingmuch perlbot nopaste 13:58
perlbot Paste your code here and #<channel> will be able to view it: sial.org/pbot/<channel>
lightstep yay, code from nothingmuch!
pasteling "nothingmuch" at 212.143.92.226 pasted "hypothetical curry unwrapping" (67 lines, 2.8K) at sial.org/pbot/8439
nothingmuch is doApplyBindings sane? 13:59
stevan autrijus: since I am going through each test anyway I am building you a nice TODO list as well
autrijus danke!
stevan yah wilkommen 14:00
nothingmuch i'm concerned about this: | subType = SubCurried = (subFun sub) env bindings
autrijus er
| subType == SubCurried
my typo. sorry.
nothingmuch ok
autrijus otherwise sane.
nothingmuch tries to compile 14:01
lightstep it looks like a type error: fun has 2 params and should have only one 14:04
nothingmuch where do i defines subCurried?
lightstep but perhaps the definition of Exp is not cached right in my head
nothingmuch errm
so how do i get env <- ask to happen at the right time?
autrijus line 381 ASt.hs
lightstep line 381 in AST.hs
autrijus is how you define subcurried
er.
brb :)
lightstep i'm always the second
nothingmuch well, i've got to go 14:07
i'll try to make progress at home
actually, why is env <- ask done out of doApply?
shouldn't our new doApplyBindings take care of it?
lightstep it should 14:08
but as it goes, doApply is called by itself in some smelly, complicated piece of code
and doApply and doApplyBinding have almost the same code 14:09
nothingmuch so how do i pass env through the curry closures?
lightstep so it made sense to split like that
you can extract the env before creating the new sub, and enclose it in fun
nothingmuch ride is here, ciao! 14:10
evening 15:05
lightstep hi 15:06
nothingmuch compiling is sooo sloooow 15:16
castaway get a faster machine! 15:17
nothingmuch i'm etherwaking one up just now
masak good thing at least that computers are becoming faster to accomodate for our increasingly complex programming languages :) 15:18
castaway :)
hmm, could set one up that does nothing but compile the thing 15:19
masak all the time? :)
castaway yup
masak wouldn't once be enough? 15:20
castaway no cos it keeps changing
masak faster than it can compile? oh my
castaway loop { grab newest from cvs; compile; save somewhere; redo }
(or some such)
nothingmuch i had a box at work do that 15:21
castaway why past tense?
nothingmuch while true; do if [ `svn up | wc -l` -gt 1 ]; then perl Makefile.PL && make && make test; fi ; sleep 600; done 15:22
because it's at wor,
gaal nm: better throw in a make clean too
nothingmuch i have to ssh 3 times just to bypass the fw
it's not fun
castaway I see
nothingmuch err, make that make optimized
gaal: why make clean?
gaal paranoia?
PerlJam waits for someone to setup a pugs smoke-testing sever
nothingmuch heh
PerlJam: i can set a box up, if you like 15:23
we are thinking of synching it with the backlink format
i think i'll try to use vimcolors for the output
and color each line red or green as reported by fail or ok
for the most recent rev
gaal fibonaci.babylonia.flatirons.org/perl6.vim ? 15:24
ah
nothingmuch gaal: exactly
gaal nm: i hope to have time tomorrow for some hacking. is the yaml ouput format still in vogue? 15:25
nothingmuch i think so
i can just hook up with Test::Harness::Straps directly for that, methinks
it'll update the index and the test status for each revision 15:26
gaal feel free to beat me to it :)
nothingmuch ok =) 15:28
first i want to get currying done
the haskell type system is, err, teaching me itself
read: giving me a hard time
gaal look on the bright side, you'll find OCaml easy. 15:30
nothingmuch hah 15:32
wooow
the tests pass!
tres cool 15:33
autrijus: how do i implement .assuming?
autrijus just look at "goto"
&goto
in Eval.hs
and go from there
I'll abstract away these two Bloody Hacks 15:34
nothingmuch ok
gaal um, i broke the goto test again earlier
i mean i didn't break it, but added some tests that parsefail.
nothingmuch what is subExp?
an Exp that will result in a VSub? 15:35
what is shiftT? 15:36
lightstep subExp is the expression that evaluates to the sub 15:37
(the sub which is gone to) 15:38
nothingmuch in my case the sum that will be assumed
perlbot nopaste
perlbot Paste your code here and #<channel> will be able to view it: sial.org/pbot/<channel>
pasteling "nothingmuch" at 212.143.91.217 pasted "attempt at .assuming" (5 lines, 171B) at sial.org/pbot/8444
nothingmuch sub <- fromVal vsub is causing type errors 15:39
lightstep fromVar is a pure function
nothingmuch oh shit, i mixed tabs & spaces
lightstep so in a monad, you use it as let sub = fromVal vsub 15:40
nothingmuch the goto hack does <-
the first two lines are unchanged
lightstep but you really want fromVal'
nothingmuch what's that?
lightstep oh, i was mistaken
they are the same, except fromVal is better 15:41
nothingmuch what do i do with retVal VSub? 15:43
do i make a Val out of the VSub? you said no, because it's pointless
so just return VSub?
because reduce :: ... -> Eval Val 15:44
lightstep actually, i wanted currySub to do as least as possible
i didn't want it to bother with casting the VSub into Val, or Exp, or Symbol, or whatever 15:45
castV is your friend
nothingmuch hah, it seems to actually compile 15:46
ok, now i have to fix stuff
camn 15:49
damn
what's wrong with this: data SubFun = Exp | (Env -> [(Param, Exp)] -> Eval Val) 15:53
lumi Don't you need constructors? 15:59
nothingmuch oh, right
*blush*
damnit 16:01
damnit damnit damnit 16:02
i'm just not capable enough
castaway then learn :) 16:03
nothingmuch trying
chip but are you fully functional?
chip wonders whether that joke is as old on #haskell as it probably is
nothingmuch =)
autrijus hi chip! 16:04
chip: have you seen the pugscc numbers? :)
chip autrijus: jeepers, no
autrijus mandelbrot:
Pugs Interpreted: 77.29 seconds 477k memory 16:05
PugsCC Unoptimized: 01.35 seconds 476k memory
PugsCC Optimized: 00.39 seconds 455k memory
Perl 5: 00.14 seconds 362k memory
and it's still doubly boxed and with slow op.
nothingmuch autrijus.... help... me... *cough&
autrijus ok. ok. 16:06
chip Yowie.
"doubly boxed" - Perl 6 box inside Haskell object box?
autrijus chip: so. got any time for parrot things? I'd like to talk a bit about parrot ast
yeah.
exactly.
slow op means it's using the same op as evaluator
(so, guarantees the same behaviour, but not fast) 16:07
nothingmuch: do you want to SEE?
nothingmuch in theory isn't p6 supposed to "compile" operators even when it's just "interpreting"
chip Surely I would like to help you, but at the moment I suspect I will be learning from your questions
Please ask away in either case
nothingmuch that is, dispatch should be precalculated if possible?
autrijus: sure
autrijus chip: ok. as I gathered, Leo did a bit work on parrot/ast/ 16:08
but nothing is really using it
what is sam ruby using then?
nothingmuch autrijus: would you like to ssh into pasta?
that would let you access the code
chip See, I told you I'd be learning. What's Sam Ruby doing?
autrijus nothingmuch: but SEE won't? 16:09
chip: pirate, that's python on parrot
nothingmuch see will
but it's several files
chip Ah.
autrijus sure.
see://woobling.org/?
nothingmuch yes, one second 16:10
i have to close all the gvims and open instead
autrijus chip: anyway. I have this idea of modelling Parrot AST with haskell
because haskell is great for modelling trees
nothingmuch ok, see is open
autrijus and then write the parrot ast -> pasm emitter in haskell too 16:11
then I switch Pugs to use that AST
then port pugs to perl 6. voila
clkao hiya autrijus
autrijus hi clkao .
chip OK, if you're emitting pasm anyway, is a conversion to the Perl 6 AST necessary? What would pugs gain by it? 16:12
autrijus the ability to implement macros
because the perl6 src -> perl6 ast -> parrot ast -> pasm picture is Wrong 16:13
it assumes that perl6 is a static language
nothingmuch autrijus: is that you who joined?
autrijus yes
that view presume that perl that has separate parsing, compiling and runtime
which perl does not.
chip Ooh, this is true.
autrijus see BEGIN{} and eval"" and "is parsed".
nothingmuch anyone else is also free to join, ofcourse 16:14
chip Pirate is generating pasm directly
self.append("typeof $S0, %(what)s" % locals())
#self.append("print $S0")
self.append("unless $S0 == 'PerlUndef' goto %(endCheck)s" % locals())
autrijus ok. aww.
that is so not reusable
chip sadly
nothingmuch autrijus: basically my problem right now is fitting both the curry unwinder and the Exp that is the sub into subFun
autrijus my problem here is I don't know any parrot at all. 16:15
let alone parrot ast.
so I'd like to sync up with someone that does
preferably with IRC or SEE or videoconf
and take a real pugs ast
and try to work with that person on converting it to parrot
nothingmuch parrot ast is going to be a bit more difficult though, no? 16:16
since it's register based and all
autrijus uh no.
ast abstracts that away.
that's the whole point of ast.
chip Now I see
nothingmuch ah
so ast is a layer over parrot asm? 16:17
autrijus over parrot pir :)
but yes.
chip Ever work with gcc?
autrijus parrot pir is a layer over parrot asm that let you pretent you have infinite registers.
chip: no. :(
chip the gcc TREE is the source representation, or at least models the source language's concepts. The RTL (register transfer language) models the machine execution, albeit with infinite psedo-registers in the early stages 16:18
TREE=AST RTL=PIR approx. from the sound of it
autrijus yes. I grok the theory
chip ok
autrijus so is parrot AST going to be very closely based on gcc TREE? 16:19
chip I have a bit of a learning motorcycle jump ramp starting at me :-,
jabbot pugs - 979 - fixed bug where undefined values in $got
autrijus does parrot has a eval_bytecode?
I see load_bytecode
chip autrijus: I doubt it. For all my background & interests, I'm just learning many vital basics about Parrot
That was my attempt to verify my understanding 16:20
(active listening by analogy)
autrijus k.
so I think a static compiler
for the mandel.p6
is a good start.
chip "static compiler", i.e. no ability to do macros? 16:21
autrijus no ability to do BEGIN. 16:22
and eval""
please don't think about macros just now
chip heh
autrijus so, the thing I have here is a perl6 parser that can parser to some sort of ast 16:30
and a compile that can transform that ast into some other sort of ast.
is there a pasm madelbrot? 16:31
ah yes. ./examples/assembly/mandel.pasm
ouch.
that hurts
jabbot pugs - 980 - more Test.pm error reporting improvement 16:49
obra so. could it parse perl6 and spit out perl5? 16:51
pretend I didn't ask ;) 16:52
autrijus ok. that is interesting. 16:53
it means we can write perl6 and have it work with RT ;)
nothingmuch autrijus: all of Bind.hs is filled with [(Param, Exp)] 17:11
should i just s/that/Bindings?
autrijus uh. 17:12
we actually want Exp
nothingmuch hah
autrijus because it's prebound
maybe we can stuff them with thunks.
nothingmuch so i am replacing that
autrijus that may work.
so becomes Val.
nothingmuch and changing type Bindings = [(Param, Exp)] for now
someone smarter will sanity check it 17:13
maybe i'll bug lightstep a bit, he knows a lot
stevan autrijus: I have un-TODO-ed all the tests, and built a TODO list for you
where would you like the TODO list? perl6-compiler?
nothingmuch too bad ghc doesn't do -jx 17:14
autrijus stevan: yes!
stevan ok
nothingmuch autrijus: what's this? subs slurpLen subSyms = 17:17
i don't know which line anymore
i think 596
oh wait, i don't think it matters 17:19
autrijus it does not.
nothingmuch make test
wow! stuff seems to work!
ok, we have refactored vsub with embedded bindings
who has latest test failure report? 17:20
i want to compare
stevan nothingmuch: wait a second
i am commiting all the un-TODO-ed failing tests
ok r981 has about 78 broken tests in it
nothingmuch why are we untodoing, btw?
stevan 78/2444 subtests failed, 17:21
nothingmuch or are we un un todoing from .12?
stevan it should only be TODOed if its an unimplemented feature
but for releases autrijus likes to TODO everything for clean make test runs
lumi Can't there be some more exact datum than TODO? 17:22
nothingmuch lumi - TAP says "# TODO why todo"
stevan I also built a TODO list of all the un-TODO-ed things as well
nothingmuch but we don't implement it in Test.pm
it's mostly useless
lumi More sort of "#TODO - milestone"
stevan see perl6-compiler for that list
nothingmuch ok
shyte 17:23
stevan lumi: thats an idea, but I am not sure the milestones are that "set in stone"
nothingmuch t/magicals/block........................
hangs
hehe
stevan nothingmuch: what did you break now ;)
PerlJam stevan: seems like there should be some programmatic way to say "this is a release, so execute a todo test here"
nothingmuch subroutine bindings
=)
stevan PerlJam: hmmm, I will look into Test.pm, we might be able to do that with an %ENV variable
lumi Milestone is a bit harsh maybe 17:24
nothingmuch if you laugh at me i'll commit it =D
stevan PerlJam: that would make my life much easier :)
autrijus stevan: nothingmuch is working on .assuming()
stevan nothingmuch: *cough* svn revert *cough*
autrijus which is very important :)
stevan yum, curry :)
nothingmuch durnit
autrijus (because that let us use VThunk instead of Exp for bound vars)
nothingmuch now my linux box has a too old Test::Harness 17:25
autrijus (so one more step closer to serializable continuations and other good ideas)
lumi But if you tag it "#TODO - r999" then it automagically can be TODO/unTODOed
stevan autrijus: what does VThunk do that Exp doesnt?
lumi I mean scriptfully TODOed and makefilefully unTODOed
nothingmuch lighstep explained it nicely today, stevan
lumi Or so
nothingmuch VThunk is something you thunk of doing, but didn't actually do
lumi Anyway off or sth
Heh
autrijus stevan: VThunk captures the original context; Exp is just some code to be run in any context. 17:26
stevan autrijus: ahhh, very nice
autrijus VThunk is an ordinary value that can be passed around for lazy evaluation, too.
nothingmuch: but come to think about it, I think 17:27
&foo.assuming($blah)
should eval $blah strictly
stevan really needs to learn haskell
autrijus that is, reduce it to real Val instead ofVThunk
instead of going back to eval it when the curried thing is called
i.e. unlike what haskell do
nothingmuch ah
good point
well, we'll get to that at some point 17:28
it's still Exp now
anyway, i've got to go pick up my mom from the train station
autrijus cool
stevan nothingmuch: NO, you must VThunk!!!!
nothingmuch++ 17:29
autrijus++
jabbot pugs - 981 - un-TODO-ed all the tests and went throug
nothingmuch stevan: soon, soon...
autrijus =)
elmex stevan++
nothingmuch i've been working on it since yday
and today i've had a waiting-for-things-to-go-bad day at work
everyone++
ciao!
autrijus nothingmuch++
ciao!
elmex this is... well err..
stevan has to do some $work work now
nothingmuch: tell mom we said 'hi' 17:30
autrijus yeah :)
stevan nothingmuch's mom++ # for giving birth to nothingmuch :) 17:31
autrijus :D 17:32
stevan autrijus: did I hear you say you were going to YAPC::MA
s/MA/NA/
autrijus yes and do a pugs hackathon there 17:33
stevan nice
I am planning on going too
what does the pugs hackathon entail? 17:34
autrijus woot!
it means that I'll be staying around for 2~3 more days
and find a place with good network
and hack.
stevan hmmm
autrijus it's in June
stevan yes
autrijus so by that time we'll probably be hacking the bootstrap.
or macros.
stevan you mean it wont be finshed by then? ;) 17:36
autrijus perl is never finished :) 17:37
stevan so true :)
we should start a fund to buy nothingmuch a plane ticket to YAPC::NA 17:38
rjbs nothingmuch's location?
theorbtwo Wow! 17:39
stevan Isreal
Khisanth autrijus: "feature complete"? :)
theorbtwo That sounds great; I want to go to YAPC::NA now.
stevan theorbtwo as well
theorbtwo Nonono; I'm useless.
autrijus Khisanth: it's not specced :)
stevan theorbtwo++ # for humility 17:40
autrijus I should also push for cabal to show up there :)
as well as more parrot people.
stevan and get merlyn to throw a really really big party :P
stevan _really_ has to do $work now 17:41
autrijus have fun :)
Khisanth stevan: wouldn't that require merlyn to be at YAPC::NA? 17:42
theorbtwo Canada won't let Merlyn in, apparently.
chip Hm. I wonder what the default prototype of C< method foo > is. (MyClass $_, *@_)? (MyClass $_)? 17:45
stevan Khisanth: not nessecarily :)
Khisanth telepartying... 17:46
Khisanth invents some words :P
chip Perl6::Subs is getting better. 17:53
C< method foo { print $self } > becomes C< sub foo : method { my ($self) = @_; print $self } >
oh, and C< die if @_ != 1 > also (but with message) 17:54
autrijus have a guessimate at when we can use the code? :)
chip This week, I suspect. I think this is far enough from $dayjob interests they'll let me share it
autrijus guesstimate, even
woooooot.
chip++
PerlJam chip: $self magically appears in methods?
autrijus I'll not write it on my journal yet :)
but, wow. 17:55
chip PerlJam: that's a $dayjob standard for referent, so yes. But you can override with C< method foo ( $me: ) > which uses $me instead of $self
autrijus clever.
PerlJam bueno
chip You can have optional parameters, named parameters, *or* slurpy parameters. Perl 5 doesn't really let you mix them
autrijus unless you also source filter the caller. 17:56
which you probably don't want to go.
chip autrijus: .... ooh.
rgs that's spiffy.
autrijus gives chip bad ideas.
that's Spiffy.
rgs :)
chip Y'know, if the source filter's output is itself a source filter.. 17:57
autrijus so, my talk is on 27th March.
chip No, it's just too evil.
autrijus so if Perl6::Subs appears on CPAN before that, I will definitely use it in my demos :)
chip Given a call like C< $x->foo >, how am I supposed to know the type of $x in a source filter?
autrijus you pass the source to PPI
chip autrijus: Excellent
autrijus you alteratively rewrite it in a wrapped form. 17:58
chip autrijus: hm, wait. If the syntax of the call is somehow distinct I can translate it without knowing the target.
autrijus yes.
that also works.
chip Well, even PPI can't help me if the previous statement is C< $x = foo() > and foo() has no declared return type. Halting problem, etc. 17:59
autrijus you need runtime recompilation for that. 18:00
definetely not going to go there. :)
chip 'fraid so 18:02
er, not. whatever
:)
autrijus o/~ yeah / whatever / nevermind o/~ 18:04
journal up. 18:08
will sleep soon. thanks for all the fish!
chip wonders about the engraved writing on his fishbowl 18:11
Hm. 18:34
And given C< sub foo (Array $a) >, is foo(undef) legal?
grr 18:35
This is ... disturbing
theorbtwo I thought the plan was that undef is a member of all types, except the lowercase ones. 18:36
chip Well, that's what I thought. I imagine, though, that it'd be nice to be able to make sure the values weren't undef in a less verbose way than C<where {$_}> 18:38
though that's not really _too_ verbose, now that I look at it
nothingmuch damnit 18:56
stevan hola naddamucho
nothingmuch hola, señor
load load load!
nothingmuch sighs loudly 18:57
stevan whats the problem?
nothingmuch gentoo insists on Test-Harness-30, which doesn't have HARNESS_PERL honoring 18:58
stevan gentoo--
nothingmuch hah! that'll show you
gentoo++ most of the time, actually
i've been very happy with it so far
i just typically rely on installing myself
(perl modules)
but this is the first time i needed something odd on gentoo 18:59
i should learn to use their little tool to make packages
but then again, i do most of my devel on my laptop with is OSX
stevan so I was discussing with autrijus about starting a nothingmuch-to-YAPC::NA fundraiser
nothingmuch oooh
that would be nice
stevan we just need something to sell :P 19:00
nothingmuch hehe
i am willing to sell my body
but in parts
castaway g-cpan, nm?
nothingmuch toe nail clippings, hair, etc
stevan how much for the brain?
or rather, just the part with the Haskell fu :P
nothingmuch err, that's too difficult to package 19:01
haskell fu?
i'm asking sarcasticly-seriously
i really have no fu at all
castaway nm has haskell fu?
nothingmuch maybe foo
or something equally lame
but i'm really having trouble
stevan nothingmuch: you have more than me :)
nothingmuch what i need, IMHO is practice more than anything 19:02
i can figure out a good deal on my own
but slooowly
and sometimes wrongly
and i am having trouble thinking ahead
ingy hola
castaway thinking--
nothingmuch hola ingy 19:03
ingy dreaming++
nothingmuch castaway?
jabbot nothingmuch: castaway is my girlfriend.
nothingmuch damnit, jabbot!
stevan hey ingy
castaway yes, nm?
nothingmuch castaway is theorbtwo's girlfriend
why is thinking--?
castaway cos its hard 19:04
ingy stevan: aye
nothingmuch =P
jabbot stupidity:
why?
castaway thinking ahead is even worse..
nothingmuch it's supposed to say 'why is the blah blah blah?'
hola gaal
gaal heya
stevan ingy: nothingmuch is auctioning off his body to pay for a trip to YAPC::NA, you need anything?
nothingmuch parts of my body 19:05
castaway slaps jabbot
stevan yes,.. only in peices
theorbtwo nm, got a couple of spare thumbs?
castaway good idea 19:06
gaal 's $work occasionally sends people to TO. man, hacking their schedule would be great.
castaway thumbs would be useful
nothingmuch TO?
my thumbs are necessary
for the bass
gaal Toronto.
nothingmuch i don't type with them too much, except for the right hand thumb, for the space bar
ah, toronto is cute
stevan nothingmuch: slap & pop funk bass?
nothingmuch very calm
err, no, double bass
you need for the arco 19:07
gaal it also hosts yapc, no?
nothingmuch and the left hand side feels around the neck
stevan was just picturing nothingmuch ala Bootsy Collins
nothingmuch otherwise you lose the sense of where you are
i can probably give up a pinky or something from my feet
that would be worth a YAPC
stevan gaal: yes YAPC::NA this year is in Toronto
gaal gotta assign myself back on that project then :) 19:08
nothinmuch, i can see if they are hiring. you wouln't mind, oh, coding java for a while? :-p
castaway nm, but I have no thumbs.. :(
stevan gaal: your in israel too right? you could bring nothingmuch in your carry-on luggage?
nothingmuch castaway: phsically? 19:09
gaal hmm, he might in fact fit
castaway correct 19:10
stevan gaal: especially after we sell all his extra parts
gaal and if he gives away his fingernails then the security folks won't take issue with his being there
nothingmuch wonders how you get along without reversable thumbs
gaal so, nm, shall i hack at test::harness? 19:12
castaway my right index finger has kinda evolved to be a thumb
theorbtwo Most people's thumbs aren't reversable -- just opposeable. 19:13
castaway good point
nothingmuch that's what i meant
=)
gaal: yes, today is all haskell
theorbtwo The answer to how she gets along without thumbs is "remarkably well". 19:14
castaway types faster than lotsa people
stevan darwinism++ 19:15
nothingmuch ook, i've ruined $?BLOCK and $?SUB
gaal "ah, yes, today is finishing the specs, adding all the leftover bits of the language, changing over the ast to ponie, speeding up the code 2^10 times. no time for *perl*"
nothingmuch heh 19:16
i meant for me =)
ook 19:22
nothingmuch should add autocmd for *pugs*.hs to set expandtab
keeps pissing off ghc 19:23
and i never understand why
stevan; can i have the faillist again? 19:25
78, right? 19:26
yay! 982 19:36
nothingmuch everyone please svn up and test, this change could give us trouble and I want to find out sooner rather than later 19:39
jabbot pugs - 982 - VSub now encapsulates it's own bindings
gaal my msys environment b0rk. anyone ever see this error? fork_copy: user/cygwin data pass 3 failed 19:40
google mentions it, btu i'm not yet sure if it's a system thing or a make thing -- it happens on perl Makefile.PL
nothingmuch how deep is fork emulated on windows? 19:41
could it be a resource problem?
gaal unlikely as i have 450MB RAM free. 19:42
then again, this is windows and it must have been a month since i last rebooted. what was i thinking? 19:43
how do i even check my uptime in windows? 19:45
nothingmuch cygwin should probably have uptime, no? 19:46
gaal not on my box. and i seem to recall that it gave wrong results when i tried once on a machine at work.
darn, a reboot didn't help. testing if an older revision also has this problem. 19:53
pasteling "gaal" at 192.115.25.249 pasted "make test summary on the linux box" (25 lines, 1.7K) at sial.org/pbot/8455 19:54
gaal what should i prove -v? 19:55
nothinmuch, on win32 mingw, r970 builds. tracking down exact breakage location... 20:00
nothingmuch 78 tests are expected to fail 20:01
stevan untodoed them earlier
oh, but way after 970 20:02
979, i think
gaal are you talking about the win32 *build* fail, or my linux *test* fails?
nothingmuch i thought test fail at first 20:03
but then i understood that you meant build failed
gaal i have reason to believe my earlier fork-related build fail was not a system problem, or at least not only. -- it isn't triggered on a slightly older r
nothingmuch do you have strace? 20:04
gaal no.. brb, phone 20:05
mj hello, can somebody add 'next unless $dir;' before line 19 to Modules\Install\Can.pm. It will fix win32 configure error which occurs if PATH contain two semicolons subsequently. 20:07
set PATH=C:\perl\bin;;c:\ghc\ghc-6.4\bin
perl Makefile.PL
\ghc not found
*** Cannot find a runnable 'ghc' from path.
Thanks.
nothingmuch mj: do you want to do it yourself?
mj I can try. TortoiseSVN - SVN Commit... ? 20:09
nothingmuch something like that 20:10
i'll need an email address to grant commit access
mj [email@hidden.address] 20:11
gaal beh, now r982 builds fine all of a sudden. so maybe it was a transient mingw thing after all :( 20:12
nothingmuch =P
mj: invitation sent 20:14
welcome aboard, mj 20:23
mj done, thanks
stevan Darren_Duncan: are you around? 20:27
Darren_Duncan stevan, right here
stepped away for a few minutes
jabbot pugs - 983 - fix win32 configure error which occurs i 20:29
mj jabbot: something wrong about my first commit? 20:31
jabbot mj: Does that suggest anything else which belongs to you?
qmole :) 20:34
nothingmuch mj: jabbot is a bot, and a stupid one at that 20:41
jabbot?
jabbot nothingmuch: jabbot is a bot, and a stupid one at that
nothingmuch haha!
ook
stevan nice :P
jabbot__
jabbot++ 20:42
(damn off by one)
nothingmuch in two senses ;-)
well, now, off by one and off by two
oh my
this is harder than I thought
binding is a bit overzealous 20:43
mj :-) muhehe
nothingmuch oh my, i feel so... dirty 20:46
well, i've had enough 20:59
but we may have &foo.assuming tomorrow
it currently sort of works but in a bad way
but the commit to parse it does not yet work 21:00
sorry, was not yet made
i'll leave the files open in see://woobling.org if anybody wants to play with them while I'm asleep 21:01
with some todo
ciao!
stevan adios nothingmuch 21:03
mj 983: nmake test failed ... still "cmd too long" (2268 characters :) 21:23
MSYS-MinGW-MSYS-DTK build: 79/2446 subtests failed, 96.77% okay
good night
stevan good night mj 21:24
elmex autrijus++ 22:20
metaperl wilx. wow a real Haskell expert in our midst 22:54
wilx lol 22:55
Am I?
metaperl well, better than most of us Perl hackers
we need to get all the procedural cruft out of minds
it takes some time to trust the Haskell way of just being definitional about things 22:56
wilx I assure you think to highly about me :)
too*
metaperl no, I saw your code the other day
wilx Hm.
metaperl remember when I asked you for an English version of your website
wilx Oh. 22:57
Right.
elmex perl 6! 23:28
ayrnieu elmex - where?! 23:29
elmex i have no clue
theorbtwo pugscode.org/ 23:32
elmex ?
chip Good news, boss thinks I'll be able to CPAN Perl6::Subs 23:51
About a week, so don't start polling the CHIPS directory just yet
ninereasons autrijus, I can't get your mandel.hs to compile with ghc 23:58