Pugs t-shirts www.cafepress.com/pugscode | Pugs 6.2.9 released | pugscode.org | pugs.kwiki.org | paste: sial.org/pbot/perl6 | www.geeksunite.net
Set by stevan on 15 August 2005.
elmex ]a 02:17
and when speed doesn't matter, types get in your way
rantanplan_ elmex, yeah, and you have to write more code while using typeing. ;). 02:37
Khisanth rantanplan_: haskell seems to be a nice compromise 03:24
rantanplan_ Khisanth, yeah, i had to do some Haskell like 3 years ago. But i never got very deep into it. 03:28
Khisanth landover: search.cpan.org/~rcaputo/POE-0.3202...POE/NFA.pm seems pretty close to what you were describing 03:30
obra autrijus: how was the talk? 03:54
04:02 khisanth_ is now known as Khisant, Khisant is now known as Khisanth
buu ?eval "A" ~~ "b" 04:09
er
Evalbot went away =[
fraxtal buu 04:42
buu What? 05:01
fraxtal hi
buu HAI FRAXTAL 05:04
dduncan hai yu 05:08
QtPlatypus asks "Default value for an attribute? C<has $.server = "localhost"> " 05:18
QtPlatypus shakes his head "No thats wrong. (searches for perl6 initlizor examples (I feel so much like a cargo cult programer)) 05:32
QtPlatypus looks confused "I guess pugs doens't implement that, though its the correct code" 05:40
autrijus greetings live from CUFP! 06:04
perlcabal.org/~autrijus/cufp2005.pdf
a "very informal" gathering; just started
geoffb hi! 06:05
autrijus yo geoffb 06:06
dduncan and greetings from where I am 06:08
geoffb Ah, the trek officially begins 2006-02-01. I wonder what we can get done before then? 06:10
Must sleep. Best of luck at CUFP, autrijus 06:11
autrijus whew done 06:49
lots of laughters despite me being nervous
I'm sold on Takahashi 07:00
it takes 3x time to prepare but is worth it
gaal morning! 07:17
glad it went well :)
autrijus :D
gaal oh wait, *went* well or is still going on well? (hi! cufp!) 07:20
autrijus it's still going on 08:02
david roundy is going to talk about darcs now
the mldonkey talk was cool 08:03
lamenting that OCaml doesn't have a CPAN (this is a recurring theme)
svnbot6 r7120 | gaal++ | add ignore property to /src/System/FilePath.hi 08:04
gaal they just need to interface pugs then. 08:05
ah btw i don't know if you backlogged, but it turns out putter++ fixed the win32 wrapper, which is quite nice because he didn't afaik actually have a win32 box to hack on. :) 08:07
autrijus oooh good
gaal regardless, pugs-the-binary also understands -B now :)
autrijus that's wonderful... 08:08
so why are we keeping the .bat around?
gaal timidity :) 08:09
autrijus I see :)
gaal is there a "proper" way to revert something that took several svn checkins?
autrijus svk reverse merge
svnbot6 r7121 | gaal++ | add tags to svn:ignore on / 08:10
autrijus but svn merge works too
gaal but i want a little more than merge -rX:Y
because there were further changes other than X
should i just perform a few svn merge -rX:X-1, for all Xs I want to revert? 08:11
autrijus nod, that's the best you can do for svn
but you can ignore that and just commit a change that reverts them
we don't really need the properness :)
gaal okay, i'll go for that then :) 08:12
works; waiting for smoke to run. 08:47
autrijus mm "FP stands for finger-pointing" (talking about scheme48 for financial engineering) 08:50
(as in, concise enough to be reasoned with and accountable) 08:51
avoidance of "unsound methods"
gaal autrijus: hee. (but did you mean to say that on #haskell?) 08:58
Base64 is still hanging. i'm committing anyway.
autrijus no, I'm just typing randomly 09:01
dduncan good night folks 09:03
svnbot6 r7122 | gaal++ | revert the wrapping `pugs` script; pugs now spawns backends
r7122 | gaal++ | itself from haskell.
autrijus night
dduncan fyi, I've been doing a whole variety of code reformatting in my modules, following Best Practices suggestions 09:04
autrijus woot!
dduncan the latest ones involved using postfix-if for flow control, such as "return if condition" or "next if condition" vs if-then or cond-and or cond-or etc that I used before 09:05
hopefully this process will wind down and be on CPAN this weekend
actually, might as well commit the corresponding perl 6 updates ... 09:06
autrijus cool 09:07
dduncan I hope that Pugs currently supports q{} and qq{} etc, as well as labels, since I added use for those, but haven't done a test compile yet 09:08
once I've put the perl 5 on cpan, I'll try to compile the perl 6 updates and see if perl 6 does the same things in different ways 09:09
svnbot6 r7123 | Darren_Duncan++ | /ext/Locale-KeyedText : resync with perl 5 version in progress
gaal d'oh! the hangs in the test suite seem to be caused by parrot 09:27
QtPlatypus: you may be interested in that fact too. haven't looked at the code for Base64, but don't use rules unless you must. 09:30
QtPlatypus The only time I use rules is in the tests. 09:42
QtPlatypus is wondering if he can rewrite them so that the tests down't use rules. 09:43
QtPlatypus will have to check that in base64 though.
wolverian Juerd, right.
svnbot6 r7124 | gaal++ | don't install an empty lexical pragma set (this also broke all BEGIN{} 10:27
r7124 | gaal++ | blocks, oops)
gaal mswin32 r7121 smoke up on the smoke server, linux coming up soon too. 10:28
so, parrot simply hangs when called from pugs on win32. anyone know how to fix it? 10:49
if parrot.exe is killed, pugs stops blocking and continues to work.
svnbot6 r7125 | gaal++ | more precise failure message for unimplemented variants of executeFile 10:51
rafl Hello 10:53
gaal hey rafl
wolverian damn, I've confused myself with fetchmail/postfix/procmail 11:03
this happens too often. :/
Aankhen`` o_O 11:16
Identity crisis? :-S
wolverian heh. 11:17
just trying to make things too hard on myself.
Aankhen`` Ah, I see.
BBL, stuff to do. 11:18
integral there's a bug with run-smoke.pl when smoking perl5: run-smoke.pl uses chdir to move to the root directory of the "pugs sandbox", and when it is calling pugs itself it uses ./pugs; but when crude_repl.pl is calling pugs it uses just pugs. But I do not have . in my %ENV<<PATH>>, so it cannot find pugs. 12:41
autrijus nod. so have it remember Cwd first? 12:42
integral maybe. I thought of just getting run-smoke.pl to change $PATH to include . at the front, but does this work on windows? Do I need to use File::Spec? 12:43
gaal integral: . is always first in PATH on windows, implicitly. 12:44
autrijus nod
gaal it's a security feature!
integral gaal: ah, so just test on $^O ne "win32"?
gaal no need -- it would just work there, except that it doesn't since perl5 doesn't work there anyway :) 12:45
integral windows uses : as a PATH separator?
gaal no, ; 12:46
you have that in getConfig...
path_sep i think, let me maek sure
integral ah
gaal yes, path_sep.
but we have a use lib, no? 12:47
oh, that's irellevant
silly me
maybe crude_repl shouldn't just call pugs out of the path? 12:48
integral hmm, the actual call to pugs must be buried in a module
gaal there's also a literal 'pugs' in the getopt block. 12:50
integral yeah.
oh well, the smoke's working now at least 12:51
gaal well, EvalX needs to be made configurable, either by some class method or by peeking at the environment. 12:56
no problem because it doesn't do anything with $pugs at compile time. 12:57
shall i fix?
or did you hackaround already?
integral I've done the adding . to $PATH hack, but not committed it. so go ahead if you've got something better 12:58
gaal the idea is to add a --pugs=./pugs to util/run-smoke.pl, and have crude_repl honor that. looks a bit cleaner, no? 13:02
wolverian hmm. is p6l dead, or is my mail setup fuxored? 13:03
(haven't released anything in a while, and I have changed things so I have reason to suspect my end is broken) 13:04
gaal beh, except that run-smoke doesn't want to know file_sep and exe_ext
integral: your hack it is :)
wolverian: last mail was from you, 13:45:00 +0300 13:05
wolverian gaal, ah, came now. thanks :) 13:11
Juerd Not anymore, gaal
I sent 2 since
wolverian Juerd, you replied directly to me, so I got duplicates :) 13:12
Juerd To you and the list.
That's what mutt does by default, what many clients do by default, and the way I prefer it.
It's reply to all.
wolverian Juerd, I prefer L (reply to list) for mailing lists 13:13
Juerd Ah, it has a key for that?
wolverian yes.
Juerd Thanks! :)
wolverian you're welcome :)
Juerd The reason I preferred (until a few seconds ago) group reply was that I didn't have to type the mailing list address myself
From now on, unless I forget, I'll use L 13:14
Hm, %hash.keys really only buys one character over texas %hash>>.key, and 0 charactors over the real thing :)
Still, typability wins for %hash.keys 13:15
wolverian what is .keys specced to do currently?
Juerd Same as in Perl 5 13:16
wolverian that is really asymmetric with .elems
Juerd (That is: it is not specced, and thus by default what it always was)
.elems is assymmetric with everything, as specced.
.chars and .keys are symmetric
list in list context, number in item context
Or, well, .chars I think still officially is specced a number :( 13:17
Not entirely sure.
wolverian right, that's why I mailed
I didn't find it in the AES though
so I depended on pugs
Juerd !eval "foo".chars
?eval "foo".chars
wolverian ENOEVALBOT
Juerd Hm
Oh
Who maintains it?
wolverian dunno 13:18
pugs> list "foo".chars
3
same without list
Juerd I think that's annoyingly context insensitive.
Especially since splitting to characters is a very, very, very common task.
Also, I expect .chars to return a list of lvalues tied to the string, but that may be too much to ask for at this point 13:19
$foo = "foo"; $foo[1] = "bar"; say $foo; # >>> fbaro
ehm 13:20
$foo = "foo"; $foo.chars[1] = "bar"; say $foo; # >>> fbaro
wolverian right, I was going to add that to the mail but forgot
same for .words
etc.
Juerd I was skipping one step, although I do think Str should does Array, and then pick .chars
wolverian (that was actually discussed, I think, on p6l. maybe could dig it up)
Juerd It was.
wolverian I disagree with .k and .v, by the way
Juerd Do you also disagree with .kv? 13:21
I personally really dislike inconsistent abbreviation.
wolverian I'd rather it be .each or so
Juerd It returns a list of keys and values, not pairs.
Likewise, now that we abbreviate 'string' as 'str', it must be so EVERYWHERE 13:22
wolverian do pairs flatten? for *@foo.each -> $index, $value { ... }
Juerd There must not be any 'string' or 'String' left in the language, if we choose to use 'str' and 'Str'
svnbot6 r7126 | bsmith++ | Modified run-smoke.pl to prepend . to PATH so that crude_repl can find
r7126 | bsmith++ | pugs via PATH.
Juerd No, they do not.
Pairs are objects: scalars, more specifically: references. 13:23
wolverian then rename %foo.each to %foo.pairs
and .kv to .each
Juerd I think .each doesn't specify that you get a zipped list in any way and is because of that a bad name for this feature.
%foo.pairs isn't needed, for the same reason that @foo.elems isn't needed: list %foo does the trick already.
for %foo { my ($k, $v) := (.k, .v); ... } 13:24
Although of course those are still .key and .value
ods15 karma bsmith
how does that work 13:25
Juerd wolverian: I think @Larry deliberately has not published any decision re .chars in list context yet 13:26
wolverian: It was discussed in detail before, and Larry did not rule. Later, a ruling was explicitly requested, and again, there was none. 13:27
wolverian Juerd, oh, right. okay. 13:33
svnbot6 r7127 | iblech++ | * Usual svn props.
r7127 | iblech++ | * PIL2JS: runjs.pl: Added --compile-only option, needed for new -CJS.
r7127 | iblech++ | * Main: Added support for -CJS and unbroke -CPIR.
r7127 | iblech++ | * Pugs.CodeGen: Better normalization so -h looks prettier (it's "PIL", not
r7127 | iblech++ | "Pil", for example).
r7127 | iblech++ | * Pugs.Help: Add JS to the list of backends supported.
wolverian Juerd, I seem to have missed it. :)
svnbot6 r7127 | iblech++ | * pugs::hack: Accomodate for the above changes.
wolverian (it being the discussions before9 13:34
s,9,),
autrijus someone in CUFP is advocating using MDA (www.cotsjournalonline.com/home/arti...id=100276) as a vehicale of marketing code-generation, code templating, macro etc inside one's orgainzation 13:57
nothingmuch is that a good thing or a bad thing? 14:02
;-)
autrijus it's a buzzword thing that I think is genuinely helpful.
this is for pointyhairs: www.omg.org/news/whitepapers/OMG_Ar...-Oxley.pdf
nothingmuch does anybody here know XML::XPath? 14:44
rafl does, of course. 14:46
nothingmuch i'm having trouble with $node->find(...relative_path...) 14:48
i have <page><title>foo</title><content><items><item><title>bar</title><...></item><item>...</item></items></content></page>
given $xp->findnodes("/page/content/items/item") i want to extract the title for each $_ in a mao { } 14:49
rafl: ping? 14:59
rafl nothingmuch: Oh, forget about you. Hold on..
Hm.. curious. It only returns the first item to me. Does the same happen to you? 15:14
nothingmuch i get a big tree 15:16
i suspect it's the complete set of "title" elements 15:17
the problem is that the text is in hebrew, so Data::DUmper doesn't help a lot (it prints \x{decaf} everywhere)
rafl Hm. What works for me us using findnodes('/page/content/items/item/title'), taking the title child node and doing getData. 15:19
nothingmuch right, that's cool 15:20
but not as nice as I wanted 15:21
can you do relative xpath queries from a certain node?
($xp->findnodes('/page/content/items/item'))[0]->find
('title/text()')
s/\n//l
tr/l/;/
rafl $xp->find('/foo', $context); 15:22
findnodes also takes an optional context.
nothingmuch oh, i see 15:23
thanks =)
15:24 khisanth__ is now known as Khisanth
rafl Some Gimp artists in here? I'd like to have only the circle from the "Peoples Republic Of Perl" picture from the state of the onion 9. :-) 15:25
nothingmuch rafl: you'll have your circle in about 2 mins 16:26
nothingmuch.woobling.org/7peoples.png 16:36
should wokr on a transparent background well
err, dark background
i can take away the white 16:37
slightly tweaked 16:38
gaal how do i trace from inside the parser? the following:
nothingmuch rafl: ping
gaal unsafePerformIO $ liftIO $ print (show ne)
yields:
Couldn't match `GenParser Char RuleState t' against `()' 16:39
nothingmuch i think liftIO is unnecessary
integral you could just use trace...
nothingmuch but you have to put the result somewhere
ghci :t trace
i forgot where
gaal integral: trace doesn't work either
nothingmuch it wraps unsafeperform io
integral it doesn't?!
trace (show ne) (return ()) -- normally works in any monad
gaal well, not the way i tried it
oh, w/o the lift and stuff? let's see
nothingmuch yes, you don't need the lift, since trace does not let the data interact 16:40
gaal wow, trace just like that did work. i thought though that print foo was equivalent to trace foo (return())? 16:41
nothingmuch goes to B.S.
gaal anyway thanks ;-) 16:42
integral ah, I think you might have needed a return on the front of yours
integral &
gaal okay, i obviously don't understand something about Env. integral, are you still there? 16:47
in Parser, ruleClosureTrait, i basically do 16:50
env <- getRuleEnv ; putRuleEnv env{ some modifications } 16:51
but a print after that shows the modifications aren't in. what gives?
oh wait! i was overwriting the current env with the caller. yikes! 16:54
only question is, is it kosher to try to modify my envCaller via putRuleEnv env{ envCaller = ...} ? 16:56
is my intuition correct, that this won't work and i have to modifyRuleEnv on the caller myself, and even that might not work? :) 16:59
bah! my approach of writing the TVar in ruleClosureTrait is completely wrong; i have to think of some other way to communicate pragma requests. 17:26
nothingmuch why is it wrong? 17:28
gaal because it only gets called once 17:29
nothingmuch as opposed to?
gaal every 'use' or 'no' on a lexical pragma 17:31
nothingmuch oh, i see
i think 17:32
gaal i should be in ruleUsePerlPackage or thereabouts
getting warmer... 17:40
getting warmer... 17:42
arr, this barfs: 17:55
unsafePerformIO $ liftSTM $ writeTVar (envInitDat env) (idat { initPragmas = [] }
(with an extra ')' at the end...) 17:56
Couldn't match `GenParser Char RuleState t' against `()'
why?
integral gaal: Are you sure you don't need a return in front?
gaal i don't want to return just yet... 17:57
nothingmuch don't STM and IO not like each other
?
gaal just do the write
well, this did work:
integral gaal: umm, unsafePerformIO has type IO a -> a
gaal: so if you're doing that inside another monad, you'd need a return
gaal let idat = unsafePerformIO $ liftSTM $ readTVar $ envInitDat env
integral unsafePerformIO turns a *monadic* thing, into a normal pure value, that can be assigned with a let. 17:58
nothingmuch not any monadic... an IO
gaal hmm. i don't care abot the return value actually
nothingmuch the reason it's unsafe, btw, is that it ruins referential transparency
if let is applied, you don't know if it checks again or doesn't
integral gaal: you still need "return". return doesn't exit from anything, it just wraps a monad around something
gaal tries 17:59
enlightentment will come, one day.
nothingmuch: i will guard against the unsaefness with NOINLINE; i hope that should be enough? 18:01
nothingmuch the unsafeness is probably not relevant to parsing because it isn't likely to happen more than once, right?
gaal whoo! look what i got
pugs.exe: <<loop>>
that must be the haskell runtime telling me something ;-) 18:02
nothingmuch =) 18:04
pasteling "gaal" at 192.115.25.249 pasted "what's loopin'?" (36 lines, 1.5K) at sial.org/pbot/13305 18:07
nothingmuch arrf 18:10
gaal hmmm?
nothingmuch is having a hard time with a box he doesn't own, and the software that isn't installed on it 18:11
elmex %hash.k cool, it's huffmanized
nothingmuch likes it
but doesn't want it to replace .keys 18:12
aliasing++
in fact, we should standarzie on the scheme of abbreviation
gaal urig's arguments are solid, imo
nothingmuch for example, i would like .characters method on a string, but also .chars
elmex %hash.k is ridiculous.... why not abbrev every method by 1 char? why not encode a complete perl-program in spaces and runlenght encode it ??? 18:13
nothingmuch because .k is universally known 18:14
$str.c is not as obvious
though i wouldn't use .k on a hash - it doesn't sound plural... on a pair i would
gaal orthography isn't as interesting as semantics. help me find the loop please :) 18:15
elmex will there be also 'keys %hash' ?
nothingmuch elmex: ofcourse 18:16
i expect as a multisub
elmex ah
k
gaal trace prints show the loop is happenning on the second readTVar 18:22
so that does have something to do with the unsafety :)
nothingmuch gaal: do you think you could help port the mysql utf8 patch to sqlite? 18:23
gaal nothingmuch: sqlite is okay with utf8 last i checked 18:25
nothingmuch it's saving data, but not inflating it
DBD::SQLite, that is
sqlite itself does utf8 happily 18:26
perlmonks.org/?node_id=491787
gaal seems to remember a utf8 app he wrote with DBD::SQLite....
sec, raiding the shallow vaults of my memory 18:28
another sec, they're shallow, but twisty 18:33
nothingmuch =)
gaal darn, i just can't find this. 18:36
okay, can you send me something i can 'make test' and see it fail? 18:37
funny, someone sent me a query recently about a utf8 patch updated for DBD::mysql 3 18:38
nothingmuch $c->add_trigger("select", sub { 18:39
_utf8_on($_[0]);
});
gaal har, please send me a .t :)
nothingmuch Data::Structure::Util
no, that's not a test
that's a solution =) 18:40
s/(solution)/"$1"/
gaal ?eval <har> xx 3
nothingmuch no evalbot =( 18:41
gaal what's $c btw, the dbh? 18:42
nothingmuch the name of a class
cdbi class
gaal oh, right. 18:44
heh, NPOV violation in latest Encode.pm docs (`UTF-8 vs. utf8`) 18:46
so now that you have your solution
s/(solution)/"$1"/; # :-)
help me find mine :)
nothingmuch kay, fill me in 18:47
gaal sial.org/pbot/13305 18:48
nothingmuch 's hair stands
gaal traces show the loop is at the second readTVar
as you said, there's some ref transparency violated.
what?
nothingmuch it's long ;-)
gaal oh, ignore the parts that are too indented 18:49
the skinny on this is:
when the parser sees 'use foo'
it looks for foo::import and calls it
(there's also handling for the case where &import doesn't exist; and for the symmetric &no, but those are immaterial now.) 18:50
what i added was just this:
before calling import, we initialize a TVar
&import possibly writes to it
and i just want to pick up the value it wrote to. 18:51
s/to\././
okay?
the writing side of this, btw, is in Pugs::Internals::install_pragma_value
nothingmuch hmm 18:52
so idat is read correctly
then the import happpens
then idat' is read
and that's where it loops?.
gaal yes.
nothingmuch sounds hairy
gaal brb
nothingmuch what is getRuleEnv? 18:53
current lexical scope?
gaal current Env
i may have gotten some details wrong: for example maybe install_pragma_value needs to write to envCaller, not env, but that's again immaterial. 18:54
nothingmuch what if the value was an IORef?
gaal well, it's very material, only not to the looping problem :)
nothingmuch remember that STM will undo and redo if there was a violation
gaal rather than STM? no idea
nothingmuch do you know exactly how STM works? 18:55
gaal i don't actually understand that bit
heh
nope.
nothingmuch okay, i'll explain in a sec
phone
gaal ok
type ahead when you can; i'll raid the kitchen.
nothingmuch okay 18:56
STM works like this:
every read and write to/from STM is recorded in a journal 18:57
when an atomic { } finishes, the journal is checked
and if it's invalid, it's rolled back, and the block of code runs all over again
since IO is not permitted, it's sane
since haskell is pure, and most data structures don't need journalling it's fast
anyway, when two threads use a single TVar 18:58
and they conflict
then they will be rerun
until there is no conflict
conditions which are known to have not changed since the last fail cause a thread to block till the condition is changed, and the thread should be tried again 18:59
effectively what you get is semantics which are sometimes inefficient but usually good enough, without the need to lock anything
that's why it could loop
try an IORef - they have a similar interface 19:00
also, i think liftSTM in unsafePerformIO is silly
there should be an unsafePerformSTM
unsafePerformSTM $ readTVar $ envInitDat env
now, WRT the writeTVar what's up with the return? 19:01
and i think that maybe the readTVar and writeTVar, due to unsafe/lift are somehow not being finished synchronously, and are thus looping 19:02
but i don't really know
gaal that was just to get it to compile, per integral's suggestion
okay, i'll try an IORef.
nothingmuch try unsafePerformSTM
then try IORefs
gaal k 19:04
with unsafePerformSTM it still loops. trying the other one. 19:07
nothingmuch okay
gaal it wants a Show instance now... 19:12
looking for examples of how to do that...
nothingmuch a show for IORef? 19:13
or for whatever you put in it?
gaal for (IORef InitDat), which is in Env
nothingmuch hmm
that'll be hard
read the IORef, and make 'data InitDat' have a 'deriving Show' in it 19:14
gaal data InitDat already derives Show (and Ord, which was also a problem)
how do I do the "read the IORef" bit? 19:15
nothingmuch hmm
readIORef
i should thin
k
gaal src/Pugs/AST/Internals.hs:1038
nothingmuch nopaste your current code
gaal the problem is not in my new code, it's in the old code where i just s/TVar/IORef/ - src/Pugs/AST/Internals.hs:1038 19:16
nothingmuch svk pulling
haven't done so in 2-3 days
gaal it isn't new
well, i'm off for tonight. see 'ya! 19:38
zzzZZZ &
Juerd good night 19:40
Aankh|Clone goes to sleep. 19:42
G'night.
nothingmuch morning 22:27
or evenning
22:27 khisanth_ is now known as Khisanth
nothingmuch wonders whether he should work on anything in his present condition 22:32
dduncan its about 1600h here 22:54
nothingmuch 2 here.. 0200
nothingmuch is headed to bed
dduncan then good night to you 22:55
nothingmuch you too =)
dduncan later
nothingmuch all i have to do is get up from my current seating accomodation ;-)

Powered by Cro and the Raku® Programming Language.

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

Thank you!