6.2.11 released! | pugs.blogs.com | pugscode.org | pugs.kwiki.org | paste: sial.org/pbot/perl6 | < audreyt> my fingers juxpatosed
Set by nothingmuch on 24 February 2006.
00:11 TMTOWTDIt left 00:37 frederico_ joined 01:23 joepurl joined 01:48 pankaj_ joined 02:16 vel joined 02:29 Supaplex_ joined 02:45 samuel joined 02:49 Bit-Man joined 03:03 Khisanth joined 03:09 samuel joined 03:12 samuel joined 03:14 samuel joined 03:15 samuel joined 03:18 samuel joined 03:19 hcarty joined, hcarty left 03:20 samuel joined 03:23 samuel joined 03:27 samuel joined 03:34 samuel joined 03:38 samuel joined 03:44 samuel joined 03:48 samuel joined 03:51 samuel joined 03:53 samuel joined 03:56 joepurl joined 03:58 samuel joined 04:04 samuel joined 04:06 samuel joined 04:09 samuel joined 04:12 samuel joined 04:13 amnesiac joined 04:14 samuel joined
ayrnieu sigh, I'm lame. 04:21
ayrnieu forgets his feather password.
Khisanth samuel: are you planning on staying or what? 04:33
04:36 mako132_ joined 04:37 samuel joined 04:39 samuel joined 04:41 samuel joined 04:44 samuel joined 04:45 stevan joined 04:47 samuel joined 04:49 samuel joined 04:51 samuel joined 04:54 samuel joined 04:57 samuel joined 04:59 justatheory joined 05:00 samuel joined 05:02 samuel joined 05:05 samuel joined 05:06 samuel joined 05:07 mauke joined 05:10 samuel joined 05:12 samuel joined, drrho joined 05:15 K joined 05:17 samuel joined 05:21 samuel joined 05:25 jisom joined 05:26 samuel joined 05:31 samuel joined 05:33 samuel joined 05:36 samuel joined 05:38 samuel joined 05:41 samuel joined
TimToady putter: it would be relatively easy to make a patch to disable interpolation (except for \\) on regexen. Would that help? 05:43
05:44 samuel joined 05:49 samuel joined 05:52 samuel joined
gaal bsb, lumi: ping 05:53
05:53 samuel joined
gaal pizza saturday, my place 05:54
05:54 xinming joined 05:56 samuel joined 06:01 samuel joined 06:06 samuel joined
jisom is there currently any intention of allowing inlined pir/pasm in perl 6? 06:09
06:09 samuel joined 06:12 samuel joined 06:17 samuel joined 06:18 draconit_ joined 06:19 samuel joined 06:23 iblechbot joined, samuel joined 06:27 beppu_ joined, jisom left, samuel joined 06:28 jisom joined 06:29 samuel joined 06:31 draconit_ left 06:32 samuel joined, draconit_ joined
gaal jisom: eval $pir, :lang<parrot> should work 06:35
$work &
jisom right now I think compiling ghc stalled the computer building ghc...might have to force it to restart....should also say that the memory requirements for building ghc are excessive instead of just "it's slow" 06:36
06:38 samuel joined 06:40 samuel joined 06:42 samuel joined 06:44 samuel joined
GeJ morning folks 06:48
draconit_ boker
morning
06:53 samuel joined
dduncan I think that effing multiple commit is going to happen again ... 06:57
svnbot6 r9230 | Darren_Duncan++ | r2625@darren-duncans-power-mac-g4: darrenduncan | 2006-02-23 16:48:45 -0800
r9230 | Darren_Duncan++ | ext/Rosetta/ : small changes to resync with new 0.721.0 CPAN release
dduncan yep, there it is
06:58 samuel joined
dduncan god, and I don't think that's the only one, either 06:58
svnbot6 r9231 | Darren_Duncan++ | r2646@darren-duncans-power-mac-g4: darrenduncan | 2006-02-24 01:11:26 -0800
r9231 | Darren_Duncan++ | ext/Rosetta/ : incremented main module versions following CPAN release, fixed format bugs in Language.pod
dduncan f*ck
stupid svk 06:59
yet another ...
svnbot6 r9232 | Darren_Duncan++ | r2671@darren-duncans-power-mac-g4: darrenduncan | 2006-02-25 01:39:44 -0800 07:01
r9232 | Darren_Duncan++ | ext/Rosetta/ : added new module Rosetta::Shell, an interactive command shell over Rosetta, to make casual use/try-outs/dev/testing easier
r9233 | Darren_Duncan++ | r2730@darren-duncans-power-mac-g4: darrenduncan | 2006-03-01 22:54:27 -0800
r9233 | Darren_Duncan++ | ext/Rosetta/ : added new file examples/shell.p6, updated Shell.pm and Model.pm
dduncan okay, actually that one was correct 07:02
r9233 is correct, 32 and 31 are dups
and 30 is a dup 07:03
07:03 samuel joined 07:08 samuel joined 07:13 f joined, samuel joined 07:15 samuel joined 07:17 samuel joined 07:20 samuel joined 07:22 samuel joined
xinming flooding? 07:22
07:24 samuel joined
buu So 07:26
07:27 samuel joined
buu How are roles different from multiple inheritance? 07:27
07:29 samuel joined 07:32 samuel joined 07:34 samuel joined 07:37 samuel joined
audreyt greetings 07:37
<- about to take taxi to German Perl Workshop 07:38
Kattana Hallo o/
audreyt the airline (Austrian/Lufthansa) lost all baggage
they are still trying to 1)find it and 2)send it to hotel
hence I'm without most of my stuff :/
buu: roles are flattened, so you don't ever have to call super()
buu: also, if two roles provides the same methods to a class that mixes them in, that's a compile time error rather than runtime breakage 07:39
buu So what happens if one role provides a method another role does?
heh
audreyt conflict
;)
07:39 samuel joined
audreyt okay, bbl in ~30min if the workshop has wireless 07:40
&
07:42 samuel joined 07:44 samuel joined 07:47 samuel joined 07:49 mj41__ joined, wilx`` joined 07:52 samuel joined 07:55 samuel joined
ayrnieu irssi: /ignore samuel JOINS PARTS 07:56
07:56 samuel joined 07:58 meppl joined
gaal audreyt! 07:59
08:00 samuel joined 08:10 Bit-Man joined 08:25 ghenry joined 08:27 jisom_ joined 08:31 pankaj_ is now known as spo0nman, jisom_ joined 08:33 ghenry_ joined 08:39 marmic joined 08:59 nothingmuch joined
nothingmuch hola 09:01
gaal hey nothingmuch
nothingmuch read the first SYB paper 09:05
how does it replace the threading of inherited/synthesized attrs in AGs?
it's like having just synthesized ones
and while impressive, it's completely orthogonal =) 09:06
azuroth curses UML 09:07
gaal I've no idea--am not familiar with AGs--where did you see it claimed that it did? 09:09
nothingmuch everyone said that SYB is >= AGs 09:10
mooseness 09:24
ayrnieu ?eval { my $bal = 0; sub deposit { $bal += @_[0]; $bal } sub withdraw { if ($bal - @_[0] < 0) { "Sorry, you only have a balance of \\$bal" } else { $bal -= @_[0]; $bal } } } [deposit(250), withdraw(5), withdraw(5), withdraw(100), withdraw(200)] 09:29
evalbot_9233 [undef, undef, undef, undef, sub {...}]
azuroth dare I question that? 09:31
ayrnieu ah, oops.
?eval { my $bal = 0; sub deposit { $bal += @_[0]; $bal } sub withdraw { if ($bal - @_[0] < 0) { "Sorry, you only have a balance of \$bal" } else { $bal -= @_[0]; $bal } } }; [deposit(250), withdraw(5), withdraw(5), withdraw(100), withdraw(200)] 09:32
evalbot_9233 [250, 245, 240, 140, "Sorry, you only have a balance of \$bal"]
ayrnieu oops again :-) azuroth - sure, why not?
azuroth ohh, it was.. trying to subscript the scope..?
for fear of being laughed at ;-) 09:33
ayrnieu if you're already afraid of being laughed at, and in no danger of contracting this fear through trauma... 09:36
would you mind rewriting what I just wrote where $bal is an object with .deposit and .withdraw methods? :-) 09:37
azuroth I can try...
?eval class Account { has $:bal; method deposit(Int $amt) { $:bal += $amt }; method withdraw(Int $amt) { if ($bal - $amt < 0) {"Sorry, $bal"} else {$bal -= $amt} } } 09:39
evalbot_9233 Error: unexpected ":" expecting "$!"
09:43 Draconit joined
azuroth nah, I've no idea 09:43
ayrnieu ah well, someone who has one will surely come along. 09:46
curiously, though, why did you say Int $amt ?
azuroth good point; s/Int/Num/ 09:47
or did you mean why prototype at all? 09:48
ayrnieu no, just why the Int 09:49
rafl searches for an easy but practical example of rules. 10:02
I'd like to mention them in my Perl 6 talk on GPW, but I think I would need 40 minutes only to explain them. 10:03
Unfortunately the whole talk needs to fit into 40 minutes.
10:04 rantanplan_ joined 10:12 mako132_ joined
ayrnieu ?eval class Account { has $.bal; method deposit (Num $amt) { $.bal += $amt; $?SELF }; method withdraw (Num $amt) { $.bal < $amt and return "Sorry, you've only \$$.bal"; $.bal -= $amt; $?SELF } }; my $act = ::Account.new.deposit(250); [map { $act.withdraw($_); $act.bal } <5 5 100 200>] 10:14
evalbot_9233 [245.0, 240.0, 140.0, 140.0]
ayrnieu bias towards returning $self^W$?SELF , meet wall. 10:15
?eval class Account { has $.bal; method deposit (Num $amt) { $.bal += $amt }; method withdraw (Num $amt) { $.bal < $amt and return "Sorry, you've only \$$.bal"; $.bal -= $amt } }; my $act = ::Account.new.deposit(250); [map { $act.withdraw($_) } <5 5 100 200>]
evalbot_9233 Error: No such method in class Int: "&withdraw"
10:16 iblechbot joined
ayrnieu ?eval class Account { has $.bal; method deposit (Num $amt) { $.bal += $amt }; method withdraw (Num $amt) { $.bal < $amt and return "Sorry, you've only \$$.bal"; $.bal -= $amt } }; my $act = ::Account.new; $act.deposit(250); [map { $act.withdraw($_) } <5 5 100 200>] # :-( 10:16
evalbot_9233 [245.0, 240.0, 140.0, "Sorry, you\'ve only \$140"]
jp-autark can you defined your own "Num" type, say to be all numbers >= 0? 10:17
define 10:18
ayrnieu doesn't know. 10:19
azuroth I think so - somehow
ayrnieu class Num { has $.bal = 1_000_000 }; 5.bal
?eval class Num { has $.bal = 1_000_000 }; 5.bal
evalbot_9233 Error: unexpected "=" expecting trait, "handles", ";" or "}" 10:20
10:22 clkao joined
lypanov hey clkao 10:24
ayrnieu ?eval class Num { has $.bal is rw }; my $n = ::Num.new; [$n, $n.foo = 10, $n.foo, $n.foo = 5, $n.foo] 10:25
evalbot_9233 Error: No such method in class Num: "&foo"
ayrnieu ?eval class Num { has $.foo is rw }; my $n = ::Num.new; [$n, $n.foo = 10, $n.foo, $n.foo = 5, $n.foo]
evalbot_9233 [Num.new(("foo" => 5),), 10, 10, 5, 5]
azuroth ?eval my subset PosNum of Num where {$^x >= 0}; my PosNum $a = 5; 10:28
evalbot_9233 Error: unexpected "P" expecting variable name or "(" 10:29
azuroth ?eval my subset PosNum of Num where {$^x >= 0}; my $a = PosNum.new(5);
evalbot_9233 Error: unexpected "P" expecting variable name or "("
azuroth ?eval my subset PosNum of Num where {$^x >= 0}; my $a = ::PosNum.new(5);
evalbot_9233 Error: unexpected "P" expecting variable name or "("
azuroth ?eval my Num subset PosNum where {$^x >= 0}; my $a = ::PosNum.new(5); 10:30
evalbot_9233 Error: unexpected "s" expecting variable name or "("
azuroth oh well
ayrnieu there are no subtype tests in Pugs, currently. Want to write a test? 10:34
oops, they are in data_types/ 10:36
somehow, my grepping failed me.
actually, Perl6::Bible's S12.pod just seems out of date. The tests indicate that it's called 'subtype', now. 10:37
azuroth hmm. I'd be inclined to go with the synopses rather than the tests? :x 10:38
ayrnieu (still not implemented)
10:39 chris2 joined
ayrnieu (I mean, the not-implementedness of subtype continues the perceived likely not-implementedness of subset.) 10:40
azuroth true 10:41
ayrnieu but, yes, it looks like I could just use a lexical subtype for ::Account :-) 10:44
where I could make type errors out of <=0 10:45
azuroth interesting indeed 10:47
I'm learning about derivatives of graphs. I really wish I'd done some proper maths 10:48
ayrnieu ?eval multi rev () {}; multi rev (*$a, *@r) { (rev(@r), $a) }; rev 1..5 10:55
evalbot_9233 (\5, \4, \3, \2, \1)
ayrnieu ?eval, (5,4)
evalbot_9233 Error: unexpected "," expecting program
ayrnieu ?eval (5,4) 10:56
evalbot_9233 (5, 4)
svnbot6 r9234 | juerd++ | Committing my Perl 6 Myths talk 11:06
rafl audreyt: Nice talk 11:16
11:18 ntgrl joined
nothingmuch jabbot: seen audreyt? 11:22
jabbot nothingmuch: audreyt was seen 3 hours 41 minutes 51 seconds ago
nothingmuch arf
just missed her then =)
=(
lypanov 01:57 < audreyt> [00:40] okay, bbl in ~30min if the workshop has wireless
now == 04:22 11:23
so at a guess, workshop didn't have wireless
nothingmuch lypanov: Juerd seemed to have access
Juerd Audrey can't get the vpn to work 11:24
nothingmuch i just got retro-spam
spam dated for 2005
oh
Juerd We exchange my lightening talk on a usb stick :) 11:25
11:28 K joined 11:35 Draconit left
nothingmuch מוה! 11:35
11:35 Draconit joined
nothingmuch Juerd: send my regards to balas if she's at the workshop 11:36
Juerd I haven't seen balas 11:37
svnbot6 r9235 | juerd++ | Audrey's changes: present tense, blank lines, a new slide, a better "Perl 6 is not Perl" slide
lypanov umm 11:38
nothingmuch In that case, please don't try to greet her under any circumstance
lypanov that reminds me. stupid question... but
nothingmuch there's no point in trying to do it if she's not there!
lypanov why did perl6 keep $/%/@ ?
ntgrl lypanov++
Juerd lypanov: Because they're useful. 11:39
nothingmuch lypanov: they help disambiguate things
for example &sub is the value
but sub is a call
this allows us to not need to say sub() all the time
lypanov & understand 11:40
but $ % and @?
nothingmuch it also helps reading
lypanov & makes perfect sense, i'm a ruby oldie and have to agree
nothingmuch they're not *really* necessary
lypanov is it still possible to use a % instead of a $ and make the code just /not work/?
nothingmuch but:a
Juerd lypanov: Of course.
nothingmuch a. it doesn't really fudge
i mean, it's not a hindrance of any sort 11:41
lumi nothingmuch: ping
nothingmuch b. it helps readability
lumi Well, not ping
nothingmuch c. it is perlish
lypanov but it requires you to think about it while you are coding more...
nothingmuch lumi: gnip?
lypanov its the reason i don't use perl...
Juerd lypanov: They're *ugly*, but ugly isn't a reason to have or not have something.
nothingmuch i like having to think while i code
lypanov this is kind of sad :(
Juerd lypanov: They're useful and make many things easier.
nothingmuch ;-)
lypanov nothingmuch: i don't
nothingmuch: my test cases do my thinking
nothingmuch actually i sort of agree
eek
that's not a good enough approahc
thinking is required 11:42
Juerd lypanov: Every feature, every character, everything, in programming, requires thinking. If you are unable to think, or don't like to think, programming is not for you.
nothingmuch at least at some level
lypanov nothingmuch: somedays i'm tired
nothingmuch thinking about *language tax* is not
ntgrl you could also ask why we have % and @, instead of just having all containers as objects in $ vars
nothingmuch but
sigils help when reading
lypanov ntgrl: exactly
nothingmuch lack of sigils help with writing
Juerd ntgrl: You can have that if you want.
ntgrl it's easy to think that functions are special to warrant a sigil
lypanov sigils just confuse...
ntgrl Juerd: yeah, you just have to become world president first to force everyone else to do it too =)
Juerd ntgrl: Forcing people has never been Perlish. 11:43
ntgrl Juerd: okay, how do I write perl code *without* being forced to use sigils? :-P
lypanov okay. lets put it another way
is slice syntax still utterly fucked or does it Just Work these days? 11:44
ntgrl it was utterly fucked? (for something other than hash references...)
lypanov ntgrl: i mean for hash refs
sorry should have said that
nothingmuch perl 6 slicing for hashes is OK now 11:45
i think
lypanov does perl6 still have this weird concept of refs vs non refs?
nothingmuch that's a weird concept?
ntgrl my question is: why separate sigils for hashes and arrays? why not just: my @array is vector; my @array is hash; my @array is btree; ?
lypanov: curiously I wouldn't have said that was weird :-/
nothingmuch ntgrl: again, it's a cultural issue more than anything 11:46
perl programmers are more important to Perl 6's design than ruby programmers
in terms of "feeling at home"
lypanov i can code both, well
ntgrl maybe, but I can understand that you can slice an @ or a %, but you can't slice a $, so there's *two* categories there, not the three
lypanov doesn't remove the fact that it is very difficult to change perl code when compared to changing ruby code 11:47
nothingmuch the main diff is that arrays are ordered, hashes aren't
and that implies that some things make sense with one thing implicitly
but have to be explicit otherwise
ntgrl like?
lypanov umm *thinks*
wolverian refs can't be sliced? 11:48
nothingmuch (iteration, splicing, poping, pushing)
refs can be sliced today
ntgrl we cope with "key", "values" and "each" already
wolverian: well you're slicing the container pointed at by the ref... but I'm needlessly pedantic
nothingmuch ntgrl: it DWIMs in perl 6
lypanov nothingmuch: okay. maybe the main reason for my hate is the difficult with finding a variables type, and added to that the insane syntax that is needed in many cases
nothingmuch $array_ref[...];
lypanov but it sure sounds like the syntax has been fixed 11:49
nothingmuch define finding a variable's type
the value class?
ntgrl but it's still slicing the pointed at thing, not the scalar
nothingmuch or the container implementation?
ntgrl: true, sigils are not as neeeded in perl 6 as they were in perl 5 11:50
due to more "automatic" behavior WRT reference
auto dereferencing
auto enreferencing
ntgrl sure, but that doesn't change what's being sliced!
nothingmuch but we still distinguish between list and scalar context
and listy things are @ish
and @ish things behave in a listy way 11:51
like automatically concatenating
etc
again, this is more of a cultural issue
lypanov i'll have to just try it, its all about feel
if i dislike it i'll try and come up with tests and point out how it can be improved
nothingmuch you can probably hack a weird glue layer over the grammar that makes any term an rvlaue
not just sigiled stuff
sure
that's always welcome
lypanov nothingmuch: i think its stupid so say "this is cultural issue"
s/so/to/ 11:52
nothingmuch but since Perl 5 programmers feel at home with this, i doubt it's going to change a lot
lypanov nothingmuch: you can surely make it less of a cultural issue and improve it for others
nothingmuch well, show me how it isn't
most of Perl 6's target audience irrationally thinks that sigils are important
lypanov print Dumper in perl5 doesn't work generically
nor does ref(blah) always work
nothingmuch (just read all the crap P6 has been getting for changing -> to .
the first should be fixable in perl 6 11:53
the second will not
because not everything is a ref
lypanov there is thusly in some (read, many) cases when its simply impossible to find out how you are meant to use a methods returned value
without good documentation (*cough*)
or reading the entire damn source
nothingmuch okay
lypanov this is fixed somehow?
nothingmuch you mean returning an array ref instead of a list for "performance" reasons?
lypanov for example, yes
nothingmuch okay
that behaves better now 11:54
because things coerce more nicely
but frankly, i couldn't care less
this is a low level issue
the reason I want Perl 6 is because of high level things
for example, all the idiots that diss python for using significant whitespace are just white noise to me
it's just syntax 11:55
if the semantics are good enough, than who cares
(i like haskell, and it parses almost like python, but i wasn't very impressed with python... guess why?)
ntgrl lack of overriding the whitespace convention?
nothingmuch ? 11:56
Perl 6's approach to this is: here's a default, the underlyingg API lets you change that
ntgrl you can do: do { f; g } in haskell, so whitespace convention is entirely optional
nothingmuch if you want 'use Syntax::Ruby' i don't think it'll mind
not just for do
also for where
and let
they have significant whitespace
azuroth why weren't you impressed though, nm? 11:57
nothingmuch and nobody uses do { f; g }, that's ugly ;-)
because python has this sort of uninspiredness about it
ntgrl sure, that's because it's defined as a single way to convert {;} <-> whitespace
nothingmuch it is as simple as it can ever be, and getting simpler
read: dumber
azuroth hmm. I think I feel the same way about PHP
a similar way, at least
nothingmuch there is lots of stuff to sto pthe user from being inspired
php is an attrocity
the only think i liked about php was that open returns a file handle, instead of accepting an uninitialized one as a param 11:58
but now we have lexically scoped handles in perl 5 so it doesn't matter
anyway, this discussion is really going nowhere 11:59
bottom line is:
sigils won't go
it's not my fault
lypanov not saying it is 12:00
gaal I am! of course it's nothingmuch's fault
lypanov just feel its very short sighted to not attempt to address these issues while "you" still can
its audreyt's fault. as always ;)
lypanov no longer likes the direction ruby is heading in
lypanov is most impressed by perl6 and pypy
lypanov hates python
so what is my option? ;) 12:01
"fix" perl6...
nothingmuch why don't you like it?
lypanov perl6 or the direction of ruby?
nothingmuch (ruby)
lypanov ah
there are a number of syntax changes that i dislike heavily, feel they are huge regressions
nothingmuch like?
lypanov nothingmuch: can't think of specifics, its a feeling 12:02
nothingmuch syntax can be anything as long as it isn't taxing
Perl 5's syntax for OO and FP is hard
lypanov syntax effects me a lot, if it doesn't look pretty then i really don't want to work with it :)
nothingmuch because it's easier to misbehave than to do the right thing
okay
let's try a few things in ext
lypanov and python made me just want to cry
nothingmuch i feel that most of the code i've written in Perl 6 is very beauutiful compared to what I can dish out in Perl 5
lypanov (first time i had an indentation related bug i ran screaming) 12:03
nothingmuch svn.openfoundry.org/pugs/ext/
lypanov most of the code i've written in perl6 is cute
s/writ/seen &/
nothingmuch write some then =)
lypanov waiting on the arrival of a new machine
nothingmuch something that really made me happy was Class::Event 12:04
and another thing that was very fun was Getopt::Process
Perl 5 really lacked the umph to make these feel dense and meaningful
wolverian Class::Event? 12:05
nothingmuch oh! collect_arguments should be lexical to parse_arguments
wolverian: yes
it's not in ext
wolverian ah. 12:06
nothingmuch because it makes very very heavy use of roles and mmds
wolverian is the module name in Getopt::Process a typo btw?
(module Getopt::Proccesor)
wolverian imagines it could be a namespace clash too
nothingmuch yes, i think it is ;-)
Migo wanted to rename it anyway
to Getopt::Engine
svnbot6 r9236 | gaal++ | r9257@sike: roo | 2006-03-02 14:01:55 +0200 12:07
r9236 | gaal++ | Minor edits to Peek talk. Online version updated.
nothingmuch svn.openfoundry.org/pugs/ext/
sorry
damn clipboard
svn.openfoundry.org/pugs/misc/Class.../Events.pm
wolverian those are long lines 12:09
anatoly ugh, I don't know :(
looks too javaish to my taste
nothingmuch anatoly: interfaces?
gaal yo tolymoose
anatoly interfaces and the general verbosity
method get_subscriptions (Class::Events::Publisher $publisher, Class::Publisher::Event $event) 12:10
this is not a line I will enjoy writing out
perl5 almost never has lines I don't enjoy writing out
yo gaal
gaal: what time at your place saturday?
lumi Yo gaal
nothingmuch anatoly: err, you don't need to =)
get_subscriptions( $thing, $event );
but that's internal, isn't it? 12:11
and interfaces suck
roles are nice because they can actually contain behavior
anatoly nothingmuch: what do you mean I don't need to? When I use the role? Sure, not the point though.
nothingmuch oh, yhou mean how iwrote it? 12:12
you can shorten the class names
and you can not annotate the types if you don't want to
but i have the Ć¼ber power of completion on my side
anatoly by the way, if I'm in A::B::C::D, why do I have to write A::B::C::E and not some kind of ::..E or whatever?
nothingmuch yes, there's something like that
i just don't remember how to write that
anatoly ah, ok :)
nothingmuch also, please look at the code within 12:13
lumi Package names are globally scoped, right?
nothingmuch the idea of Class::Event's numerous roles is that you can milk performance but also get flexibility
lumi Like Java and not like Python
nothingmuch lumi: not necessarily
you can have lexical ones too
they default to global though
gaal anatoly: not sure about time yet 12:14
there may be a jazz gig i want to go to afterwards :)
nothingmuch gaal: which one? 12:15
gaal Third World Love
anatoly gaal: well moose me or something, I'm in Rishon over weekend. 12:16
gaal great, will do
nothingmuch when is it?
gaal trying to check 12:17
hmm, 21:30 but it's w/o daniel friedman who I really like 12:18
nothingmuch what date? 12:19
saturday?
gaal yeah. next thu, 9/3, he will be there.
www.hagada.org.il/hagada/html/modul...&pid=7
nothingmuch whatever happenned to luqui =( 12:23
12:39 sisi joined
sisi what else is new? :) 12:50
12:52 Aragone is now known as Arathorn
sisi . o O 12:53
12:53 sisi left 12:59 elmex joined
Arathorn everything old is new again. 13:00
elmex is stitting on the GPW 13:07
13:08 Draconi1 joined 13:10 Draconi1 left, Draconi1 joined, nothingmuch joined
Juerd ntgrl: Write your own grammar, that is sigilless. 13:10
nothingmuch no, just derive a grammar 13:11
== less repeating yourself
13:11 Draconi1 left, sisi joined
integral except that sigils are stored in the pads, so I have to magic them out of somewhere... 13:11
Juerd ntgrl: This is a matter of changing the syntax for "my", and alike declaration keywords, and having them install lvalue subs.
nothingmuch: Same difference.
nothingmuch integral: what ntgrl wants is to have just scalars
sisi Ah..."everything changed"
Juerd sisi: No
integral nothingmuch: sort of, More like bound values like in haskell really 13:12
but maybe not *sigh*
sisi Juerd: oh? :)
integral ntgrl ::= integral
Juerd integral: Hey, don't fool us into thinking you're some kind of newbie ;)
13:13 Draconit joined
elmex is there a syntax freeze in perl6 ? 13:14
Juerd No
sisi "No means no"?
13:14 fglock joined
Juerd sees audreyt read and edit a talk. Impressive, that speed of reading and typing. 13:15
elmex when is a syntax freeze going to happen? when is the final perl6 interpreter going to be written?
Juerd elmex: Probably not even then. Why do you ask?
nothingmuch elmex: christmas
Juerd elmex: Do you want a freeze?
lypanov Juerd: have to admit i'd love to be there just to see that :)
Juerd Someone's whistling here. Very annoying during a talk.
lypanov: Why? Don't you think that if something isn't perfect, and someone has something *better*, we should use the better thing? 13:16
lypanov: Don't you think that if ambiguous cases are found, clashing syntax is found, we should be able to fix the problem?
I think a syntax freeze before the release is a bad idea.
Eh, s/before/long before/
rafl: Don't forget to diff the usb stick against your local old copy. 13:17
elmex ok, then, when will there be a release of the final perl6 compiler (which targets parrot)?
lypanov Juerd: sorry. "Juerd sees audreyt read and edit a talk. Impressive, that speed of reading and typing."
Juerd elmex: Christmas
lypanov: Ahhh, okay.
lypanov: Sorry for my misinterpretation.
lypanov i don't believe there is ever any such thing as a "final interpreter"
and nor do i believe in syntax freezes
;)
rafl Juerd: I won't. 13:18
Juerd elmex: Perl 5 has known a syntax freeze for very long; it hasn't benefitted from that.
elmex: In fact, it pushed innovations like // far, far away.
We still don't have it, while the feature itself has been wished for for over a decade.
(Originally it was written "??", hookhook.)
lypanov rationals or comments?
oh. i have no idea then? :) 13:19
i like the ?? !! syntax btw
Juerd So do I.
elmex ah, ok, so there will be finally the real perl6 interpreter.... why took it so long? :)
lypanov elmex: its Hard (tm)
Juerd elmex: It's hard to create something that hasn't existed ever before.
fglock audreyt: re packing iterator_engine to CPAN: how about "Perl6::Grammar", and reuse Parse::RecDescent API? 13:20
elmex yes, it is, i agree. but that doesn't explain why the interpreter hasn't been started earlier
Juerd elmex: There is no programming language that has all these new features
elmex Juerd: perl5 has at least a part of it
Juerd elmex: Yes, but we're not trying to re-create Perl 5. Perl 5 already exists, it doesn't need to be created. 13:21
lypanov and its not possible to easily add to perl5 13:22
Juerd elmex: Perl 5 has many neat features, but Perl 6 has more. They cannot be integrated into Perl 5, because Perl 5 is very hard to maintain because (lack of) its internal structure.
elmex yes... Perl6 isn't Perl anymore
rgs and backward compatibility constraints
Juerd Also, for a better Perl, we need *fundamental* changes in syntax. 13:23
lypanov perl sucks. yay for perl6! :P
elmex yes, maybe
Juerd elmex: It depends on your definition of Perl. It still looks and feels like Perl.
elmex: Do you think that Larry Wall's view on the definition of Perl isn't authoritive?
elmex Juerd: yes, but it is not the successor of Perl5
rgs some would say it even smells like perl.
Juerd He thinks the definition *includes* Perl 6.
elmex Juerd: no, i don't think that larrys view is the end of all wisdom
Juerd rgs: Mind if I quote you on that in my lightening talk? :)
rgs heh
nothingmuch elmex: look at my compiling perl 6 ttalk 13:24
Juerd elmex: But he created both Perl 5 and 6.
rgs and the previous ones too.
lypanov and patch!
elmex Juerd: Perl6 was created by the comunity
lypanov elmex: perl6 was defined by larry
nothingmuch nothingmuch.woobling.org/perl6_comp...start.html
note that images willl load very slowly
lypanov and implemented / refined by the community
nothingmuch and only then will S5 kick in
Juerd elmex: Nah, the community will *write* it :)
elmex lypanov: there were RFC's
Juerd elmex: Larry is this ship's captain
elmex Juerd: the ship is traveling very long now.... 13:25
svnbot6 r9237 | fglock++ | iterator_engine - added a talk sketch
nothingmuch elmex: don't take it as "larry is not open to suggestions"
Juerd elmex: Indeed.
nothingmuch in fact, i wasted about 4-5 hours of larry's time just in the OSDC/Hackathon
he is a very good listener 13:26
lypanov elmex: /win 5
elmex thats great
Juerd elmex: However, I think it's good to wait and get something that resembles perfection, instead of rushing things and getting something as horrible as the unreadable as Perl 5's source.
lypanov doh
lypanov kicks irssi
elmex: sorry. ignore that :)
Juerd: i disagree for what its worth
Juerd elmex: It has taken years, and perhaps it will take a few more.
nothingmuch lypanov: you can use esc/meta 13:27
sisi thinks: marriages don't break up on account of infidelity. *^.^*
Juerd elmex: In the meantime, we have Perl 5.
lypanov: Elaborate
lypanov nothingmuch: oh. esc-1 etc is a wonderful idea
elmex Juerd: i've been waiting since 2002, and there is no real spec yet. the synopsis are imho very incomplete and many things are only documented on the mailing list
lypanov nothingmuch: thanks :)
13:27 sisi left
lypanov elmex: the spec == the test cases 13:27
elmex Juerd: yes, but Perl6 does block improvements in Perl5 imho
Juerd elmex: Audrey et all are working on that. Don't expect things to change extremely quickly.
nothingmuch elmex: well, we're not going to scrap it 13:28
elmex lypanov: great, so i have to crawl through thousand lines of test-code to learn perl6 ?
nothingmuch this is a volunteer thing
Juerd elmex: Different people work on Pelr 5.
nothingmuch nobody tells anyone to work on this and not that
or vice versa
lypanov elmex: you learn from a spec?!
Juerd elmex: Perl 6 doesn't block Perl 5 development, it actually improves it at times.
rgs right !
Juerd Perl 5 learns from Perl 6, and gets some of its features.
rgs like //, switch, ~~, say
elmex lypanov: yes, partially. i learn mainly from having an interpreter and then i look at the spec and try out
lypanov elmex: pugs runs now
nothingmuch // is lovely 13:29
elmex ok
lypanov ?eval 5
evalbot_9236 5
lypanov elmex: ^^ see? :P
elmex perl -e 'print 5' 13:30
# 5
lypanov: see? :)
lypanov no... not really
elmex lypanov: yes, me too
lypanov elmex: pugs (one impl of perl6), runs now
elmex: and there is a partial constantly updating spec. and testcases
i also have to admit i'd like a more compact version of the spec 13:31
is one available?
nothingmuch test cases, interpreter, and spec are all complements to each other
none can be good without the other
lypanov the current stuff is verrrry verbose
lypanov doesn't really like reading
nothingmuch lypanov: there's a quick ref
lypanov nothingmuch: ah. neat. uri?
azuroth am I horribly mistaken in thinking fixity is somehow a synonym for associativity?
lypanov or in docs/ ?
nothingmuch lypanov: in docs or something
lypanov nothingmuch: thx. shall check it out 13:32
nothingmuch azuroth: fixity is how manny, where
like, postfix is 'after'
it's always after "one thing"
as with prefix
azuroth ahh, right
nothingmuch circumfix is around any number
infix is between two things
postcircumfix is after, and then around some more stuff
lypanov fixity == arity + associativity?
azuroth damn. is there a short word for assoc?
nothingmuch like [ ] is after @array and around the slice
lypanov: you could say that, yes 13:33
lypanov azuroth: assoc? :P
nothingmuch ass
azuroth uglyyy :-P
lypanov hehe
ummmm ass
13:35 Nouk joined
elmex it's just scary that Perl5 grow from a evolutionary process out of perl1 to perl4 and was mainly done by Larry. Perl6 seems to throw away most things from Perl5 and introduce very many new untested concepts in a very short time and throws away the old codebase 13:35
rgs nope, it doesn't 13:36
Juerd rgs: How does one pronounce your name?
lypanov elmex: its not a bad thing go occasionally throw away the codebase...
rgs translators will happen !
Juerd: I don't know. Spanishish ?
Arathorn elmex: sounds exciting rather than scary to me :) 13:37
Juerd rgs: I can't guess the last part, Suarez
rgs Juerd: "aar - gee - ess" ?
Juerd Does that sound like swahres?
rgs Juerd: soo a ress
Juerd Ah, thank you
rgs yes
azuroth spanishish... sounds an interesting - oil of spaniards? 13:38
13:38 kolibrie joined
Juerd elmex: What do you mean "untested"? 13:38
rgs even *I* pronounce my name differently depending on contexts
elmex imho the perl5 people should improve the old perl5 interpreter a little bit more and just bump the version to perl6
Juerd elmex: Larry *is* actively improving (refactoring) Perl 5!
elmex: Nicholas Clark will integrate his changes back into the main Perl 5.
elmex: This all helps migration to Perl 6. 13:39
elmex Juerd: many concepts were invented before there was Pugs and some concepts aren't implemented yet
Juerd elmex: Indeed.
elmex: Your point, please?
elmex Juerd: good question
Juerd Also, isn't every invention at first untested?
rgs aren't the specs written before the implementation ? :)
lypanov Juerd: i think the point is depth vs breadth wise implementation
Juerd Like, ehm, pseudohashes? Restricted hashes? Formats? All those other things we don't use, but are implemented in Perl 5? 13:40
lypanov Juerd: and afaic there is slightly too much depth in perl6
s/ic/ics/
elmex yes, but wouldn't it be better if all these new concepts were integrated in perl5 from the beginning? then they could've been tested when they came up
Juerd elmex: It's hard for me to understand what you want. A few moments ago, you complained that there was no complete specification, but now you say things should be tested first.
rgs no, because perl 5 needs to be mostly backwards compatible
Juerd elmex: Can you explain how these go together?
lypanov elmex: perl5 is very difficult to extend 13:41
Juerd lypanov: There might be. The Plan is to do it Right.
lypanov elmex: if what you did was the case, it would take /even longer/
s/did/said/
Juerd (See how the uppercase words make things really cute in English, while they just don't even look weird in German? :))
lypanov Juerd: hehe
elmex lypanov: i don't think it would have taken 6 years to have // in perl5 or a 'class' keyword... 13:42
lypanov elmex: the perl6 spec is a lot more than jusrt that
s/rt/t/
rgs // is in perl 5 right now
Juerd Because they capitalize every Noun, you cannot use uppercase Letters to indicate something that is cutely official.
elmex lypanov: yes, the perl6 spec is a big blob
Juerd elmex: You don't NEED those in Perl 5.
elmex rgs: yes, after 6 years 13:43
lypanov when did pugs start?
lypanov is sure it wasn't six years ago
rgs one year ago
Juerd elmex: Why write "class Foo { }" instead of "{ package Foo; }" if they mean exactly the same thing?
The point is that it does something *different* in Perl 6.
lypanov doesn't really care about parrot all that much as he deems it doomed
Juerd lypanov: One year and one month ago.
lypanov right
its come very far 13:44
rgs jumps off the troll train
lypanov good idea
bbl
elmex maybe the only thing that bugs me is that Perl6 is not Perl5+1
lypanov everyone hates change, but its pretty much inevitable 13:45
lypanov personally hates lack of change
azuroth parrot is doomed?
lypanov azuroth: yeah
elmex i don't hate change. perl6 has many great features.
nothingmuch parrot has had lots of management issues
and I think parrot has been prematurely optimized in ways that are not generally useful enough
azuroth hmm, ok
nothingmuch but that doesn't mean it's doomed
people are still participating and helping 13:46
nothingmuch is not so quick to mark it off as a dead thing
rgs it's merely resting !!
lypanov nothingmuch: its slow. and pretty much pointless
azuroth :-)
nothingmuch rgs: when *I* made that joke i was beaten up 13:47
13:47 haggai_e joined
rgs runs and hides 13:47
lypanov ;)
nothingmuch rgs: you're supposed to give lypanov a voucher for your brother's fish store instead
and instead of running off and hiding, running off to become a lumberjack 13:48
lumi Pet store, surely?
13:48 haggai_e left, haggai_e joined
nothingmuch i thiink it was more specialized but can't remember.... googling 13:48
lypanov que? :P
lumi His brother's pet store in Bolton
nothingmuch ah 13:49
lypanov: www.mtholyoke.edu/~ebarnes/python/d...parrot.htm
lypanov ummm
lypanov sucks at quotes
gaal lypanov: I wouldn't say it's slow:
# www.nntp.perl.org/group/perl.perl6....nals/32958
lypanov lumi: second you said "bolton" i knew something was going on ;)
gaal: ackermans makes no difference in the real world 13:50
gaal: real world optimisations == high level things such as malloc coaslescing
gaal: sorry but parrot is just bullshit
i agree entirely with nothingmuch in "in ways that are not generally useful enough" 13:51
and yes. its true, i can not spell coacleslssing
Juerd Hm, would it be safe to mention that I put my talk in the Pugs repository? ;)
lypanov flames Juerd
Juerd (It might be construed as "If you really don't like my talk, go ahead, and change it")
nothingmuch lypanov: did yyou see the perl 6 compilation talk?
Juerd Ah well. We'll see.
nothingmuch i'm very very concerned with Perl 6 being fast 13:52
and most of the ways I want to make sure it's fast have nothing to do with parrot
lypanov nothingmuch: no. any urls?
nothingmuch nothingmuch.woobling.org/perl6_comp...start.html
images will load slowly
lypanov reads
nothingmuch but when they're done S5 will kick in
lypanov k
ummmfunkycheese 13:53
elmex yaho, audrey is having her talk now
Juerd Visual Basic Rocks. 13:54
elmex yes
azuroth for real?
Juerd gwbasic++
:)
lypanov lol
Juerd # Audrey
# Larry
# me
rafl # me 13:55
Juerd gwbasic++ then
elmex elmex--
Juerd # toothpaste
lypanov Juerd: how come? ;)
anatoly gwbasic doesn't really ++
lypanov s5 kicked in! yay!
anatoly but audrey's talk is cute
elmex VB was one of my first languages too 13:56
nothingmuch anatoly: but it seems that vb 9 *does* rock ;-)
anatoly nothingmuch: in a funny way, yes 13:57
Arathorn wonders if audrey's trying to get hired by M$ ;)
anatoly vb is one of those languages which either "no one uses" or "everyone uses" depending on your perspective 13:58
Juerd She already is
;)
Arathorn ah, that makes more sense, then :)
Juerd Not really
But there is another Audrey Tang who is.
anatoly but then again, so is perl, so we shouldn't be snobbish about vb, I guess.
Juerd pugs.blogs.com/audrey/2005/12/the_o...audre.html 13:59
lypanov umm
nothingmuch: s5 is lame
nothingmuch lypanov: i can generate plain spork
i just like working with it better since it's one file instead of N per slide
ughhhhhhhhhhhhhhhhhhhhh 14:00
Juerd audreyt++ # Good speaker
nothingmuch there's a "python rocks, perl sucks, you shouldn't make perl 6 it hurts perl 5" thread on the OSDC list
why do people go that way so easily? 14:01
what point are they fucking trying to prove?
it's so pathetic
webmind nothingmuch, url? :)
nothingmuch webmind: for what?
webmind of the archive?
nothingmuch ah
err
it's not worth it
elmex elmex--
nothingmuch you can google if you like
webmind ok
nothingmuch osdc-discuss is the list
lumi elmex: Keep that up, there'll be nothing left of you
nothingmuch the thread is a subthread of "feedbacks from the conference" 14:02
webmind ok, thanks :)
anatoly nothingmuch: i can't believe you put ninjas and pirates in your talk
nothingmuch starting with a reply by Roman Yakovenko to Dov Levenglick
anatoly: =D
webmind something to fight the boredom ;)
anatoly nothingmuch: you're a psycho! but ++ :)
lypanov nothingmuch: anything but this. is it xul or something? 14:03
svnbot6 r9238 | juerd++ | Some more slides
nothingmuch lypanov: javascript + css
i'll make it plain html 14:04
lypanov nothingmuch: either firefox or s5 is damn lame
nothingmuch: it keeps on reverting to unstyled when i open a new tab
nothingmuch uh
odd
complain to eric s. raymond, he wrote the damn thing
lypanov lol
i just pressed a link
anatoly haha
nothingmuch i must admit i don't optimize for portability when doing that stuff
lypanov and ... dum dum dum. firefox crashed
anatoly that explains a few things
what's the thing that audreyt uses for her talks? use that :) it's also one file 14:05
gaal anatoly: nothingmuch is a DrMcNinja really
lypanov anatoly: only works on firefox. also lame
nothingmuch anatoly: takahashi xul
anatoly lypanov: is not lame
lypanov anatoly: if it runs only on firefox its lame 14:06
nothingmuch: yeah... i could really do with the html version
nothingmuch: ff just died again
anatoly lypanov: anything that lets you write the talk as one text file so easily and fast rocks. bah about firefox, it's for displaying at the conference, who cares.
lypanov anatoly: then don't public the damn things on a public site...
nothingmuch haha
lypanov and why not use powerpoint? 14:07
its /more/ accesible than this shit
nothingmuch i don't have ppt
lypanov sorry. but meh
nothingmuch and s5 is accessible for *me*
anatoly lypanov: because powerpoint blows
nothingmuch again, i was not optimizing for portability
but i *can*
give me 2 mins
anatoly lypanov: powerpoint eats your brain
lypanov nothingmuch: was just bitching at anatoly not you, sorry
anatoly: then buy a mac and use keynote :) 14:08
anatoly lypanov: you have ff already. it crashes at nm's spork but won't at audreyt's takahashi, so what's your problem? :)
lypanov anatoly: i don't have ff on my osx box
nothingmuch lypanov: actually, not spork
s5
spork is just kwid -> slides
and it defaults to plain html
well, kwiki
but we use spork::formatter::kwid nowadays 14:09
lypanov so many funny cute names!
nothingmuch okay
yeah, that's ingy for you
nothingmuch.woobling.org/perl6_comp/slides/
reload that
it's in hotpink
because that's the default
;-)
lypanov yay hot pink!
nothingmuch gaal has a nice scheme for plain spork, but i don't know it 14:10
gaal quite simple really
nothingmuch also, wrt ff/s5 borkedage - it's known
that's why my darcs talk sucked
because i had to use ff unexpectedly
but it didn't crash, it was just very slow
lypanov same thing :)
when i can read 10x faster than my computer can display. i begin to worry ;) 14:11
nothingmuch the old html is in s5.html
gaal for reference:
nothingmuch hehe
gaal banner_bgcolor: lightblue
nothingmuch i use it in safari and it works very well
lypanov okay
nothingmuch so there
=)
lypanov *opens his mac up*
nothingmuch no need
s5 is no longer the default for that URL
lypanov i just got s5 again, maybe cached 14:12
nothingmuch think i'll delete index.html
there is start.html which is plain spork
and s5.html which is s5
lypanov nothingmuch: thx. works absolutely perfectly in safari
nothingmuch =) 14:13
again, i'm optimizing for my comfort as I give the talk
and write it
lypanov nods
azuroth nothingmuch++ # like the slides
nothingmuch if we could have Spork::Keynote i'd be even happier
because Keynote has mad-presenter-skill-toolz
like a timer, and the next slide, and whatnot
but i don't work well with graphical tools for this stuff
just for real graphics 14:14
14:14 haggai_e left, feng123 joined
fglock can I use 'return' inside a BEGIN block? 14:17
nothingmuch fglock: you should be able to, yes
when a begin block is used in rvalue context then it's return value is serialized and reused as a constant 14:18
e.g. 'print BEGIN { localtime }'
although, wait
return might not work
because BEGIN isn't a sub
but i guess that ought to be fixed?
fglock nothingmuch: I read it is a sub
nothingmuch since there is no other possible use for return at begin time, not in a macro?
fglock nothingmuch: I need to eval a string that has a BEGIN block, and I don't want BEGIN to be executed twice 14:19
nothingmuch oh wait, is this perl 5 or perl 6?
14:20 Draconit joined
fglock it's p6 compiled to p5, so it's p5 14:20
14:20 Draconit left
nothingmuch ah 14:20
fglock I need to evaluate a BEGIN block during parse, and then again when executing
nothingmuch in that case I don't think that's specccced
fglock nothingmuch: it works: 14:22
perl -e ' BEGIN { print "begin\n"; return; print "not\n"; } '
nothingmuch yay!
14:23 Draconit joined
Juerd just had his lightening talk 14:23
And I invited people to come here (just in case there's a big stampede... ;)) 14:24
nothingmuch yay
we need to have a lightning talk about TTH btw
welcome back, Draconit 14:25
Juerd What exactly is TTH?
integral that'd be why it needs a lightning talk :-P
Juerd :)
So what is it? 14:26
Heh, rafl just asked me "What's TTH" :)
So now two people want to know :)
nothingmuch Test::TAP::Harness 14:27
Juerd Ah
nothingmuch basically say:
hi, i'm X
Juerd Different from T::H?
nothingmuch at pugs we have many many unit tests
sorry
Test::TAP::HTMLMatrix
Juerd Ahh
nothingmuch i always confuse the name
find t/
find t/ | wc -l
Juerd Right.
nothingmuch then say:
so we wrote an HTML visualizer
show the lovely output
Juerd I actually did a count yesterday, 35000+ (see my lightening talk, p6myths2.html) 14:28
nothingmuch and have another 2 mins about how to actually run it
Juerd Right
14:28 wilx`` is now known as wilx
Juerd Get over here, and have it :) 14:28
14:29 samuel joined
gaal nothingmuch: I think TimToady just mentioned BEGIN was, in fact, a sub on p6-l just the other day 14:33
or here, or somewhere
Draconit הי לא חשב×Ŗי שמישהו בכלל הבחין שנכנה×Ŗי לפה 14:34
nothingmuch yay 14:35
Draconit sorry dont know how to say that in english
nothingmuch Draconit: well, we are very sensitive people
and we get a little line saying 'Draconit ..... has joined #perl6' whenever you joined
gaal Juerd: there are another 8-odd k in ext/
Draconit yes, and clorfall too!
sorry for my spelling.. 14:36
14:41 nnunley joined 14:45 fglock left
kolibrie I did not see TimToady in AUTHORS. Who is he? 14:46
integral you've never met him on perlmonks? 14:47
gaal he's someone who puts moose in his talks. gotta love him!
kolibrie has never been to perlmonks
azuroth I've heard he can write an IRC client in less than ten characters 14:48
gaal kolibrie: look at all his replies here: perlmonks.org/?node=perl+oddities&a...ton=Search
GeJ cat /dev/urandom | perl <-- more than 10 characters 14:49
gaal useless use of cat! 14:50
Juerd rafl: Hm 14:51
GeJ yes... realized that just after typing <Enter>
kolibrie gaal: I still haven't solved the riddle 14:53
Arathorn blames CPM 14:54
where you had to TYPE FOO.DOC | MORE to paginate it
iirc
hence not useless abuse of s/cat/type/
Arathorn mumbles
kolibrie: if you want the riddle spoilt: pugs.blogs.com/pugs/2006/02/hackathonil_12_.html 14:55
kolibrie gaal: by the way, I got a little 'hello world' parsec parser working, but don't think I quite understand the AST definition syntax
Arathorn: thanks 14:56
kolibrie almost wanted to guess that, but knowledge generally serves me better than my guesses 14:57
gaal kolibrie: ooh I need to thank you again BTW because a lot of what we talked about made it into my talk
kolibrie gaal: you're welcome - it was mutually beneficial 14:58
14:59 Draconi1 joined
kolibrie gaal: do you have time a little later today to help me comprehend a little more? 14:59
gaal ignoring what AST means for a moment, look at: perlcabal.org/~gaal/peek/slide17b.html#end
how about now? :)
and perlcabal.org/~gaal/peek/slide19.html
kolibrie gaal: have a half hour meeting starting seconds ago 15:00
but after that is fine
gaal ok
15:02 Draconit_ joined
Draconit_ i hate gaim 15:03
it is evil
when u r not a geek
anatoly don't use gaim for irc 15:04
it's ok for im i think
azuroth I don't particularly mind it for IRC..
lypanov colloquy!!!!
Draconit_ it has mood swings
lumi Sadly, there's no Colloquy or Adium for [wl]in 15:05
Draconit_ whats colloquy?
azuroth not for win or lin? what is it on, then? 15:06
lumi Mac OS X
Which is a BSD, of course
azuroth oh, heh, whoops
integral irssi is pretty standard for *nix
lypanov admits to being in irssi right now
lumi aol 15:07
integral gaal: you've got a missing return on slide23a before the Syn 15:08
gaal integral: nope, look at the soource 15:09
oh wait no 15:10
you're right
nice catch :)
integral np
gaal that's what I get for rhetorical simplifications (it uses a trivial helper retSyn in the real source) 15:11
integral I wish the source wouldn't use things like retSyn sometimes. With return $ Syn, everybody knows what return does 15:12
gaal it is kind of a trivial function, yes
anyone know what might start making vim look for X all the time? it used to work fine, but now I always need to vim -X to avoid sloooow loads 15:14
fixed
svnbot6 r9239 | gaal++ | r9261@sike: roo | 2006-03-02 17:11:16 +0200 15:15
r9239 | gaal++ | * fix a missing `return`. Reported by integral++.
Juerd rafl: jitcrunch.cafepress.com/jitcrunch.a...bnQvZ2xvYm
gaal yikes
Juerd ;)
xrl.us/j9rm
rafl Juerd: It incomplete 15:16
Juerd Your sentence also incomplete, it verbless.
(Hi, Leo :P) 15:27
:) 15:28
LeTo *g*
Juerd rafl: gather { ...; take $something; ... } 15:30
Where $something is in list context
azuroth gee. he does seem to have broken english. I'd never really noticed, or it's just worse than usual
Juerd @foo = gather { for @numbers { take $_ + rand() } }; 15:31
azuroth: Who be he?
azuroth err, doesn't matter :-)
Juerd I'm sure it doesn't, but I want to know ;) 15:32
azuroth when you said "hi, leo" up there I thought you were referring to leopold, so I backlogged #parrot and noticed he was talking noticably foreign-english 15:34
svnbot6 r9240 | bsmith++ | Factored out the localisation of the environment from ruleBlockBody.
Odin- Juerd: You want to end like the cat?
15:34 bd_ joined
Juerd Odin-: ...cat? :) 15:36
azuroth: I see :)
Odin- Juerd: Yeah, the one that got killed by curiousity. :) 15:37
15:37 justatheory joined
rafl Juerd: That was caused by the us_intl layout. 15:37
Juerd Odin-: I'm not curious, I just want to know everything.
rafl: You and your silly umlauts
Sorry, *U*mlauts.
Odin- Juerd: I think one is the definition of the other... :p
Juerd Odin-: Nahhh
Odin- Ć¼mlĆ¤Ć¼ts! :p
Juerd ?ml??ts? 15:38
rafl Juerd: They are nice to have in german presentations.
Juerd Ć¼mlƤs
Hm
rafl: Disagreed.
kolibrie gaal: so, the 'data' block is very concise. Can we go over that line-by-line
integral argh, why does the pod parser need to call ruleStatementList?! 15:39
Juerd Why is anything that has to do with perl called xxxxXxxxxxxxxXxxxx? 15:40
kolibrie Juerd++
gaal kolibrie: sure 15:41
kolibrie listens 15:42
gaal look at both those slides... Exp is the name for the entire data type, each variant is one of the possible types in an actual box
it's a *little* like inheritance, but not really 15:43
kolibrie so the AST Syn 'if' matches the line: Syntax String [Exp]
gaal when you're talking in general, an AST is a tree of Exps
but each Exp needs to be of a particular variant 15:44
you can have functions that demand one variant, or you can have functions that are more general
the slide with the long names is just for pedagogical purposes :) 15:45
kolibrie wants to see how the data definition yields a tree, but does not see it
gaal the real code defines Syn String [Exp] as one variant
and "Syn" is called a "constructor" because it's one of the various ways to construct an Exp 15:46
:t Syn
Syn :: String -> [Exp] -> Exp
this means, that in your code you can have Syn "for" [body]
and that returns an Exp (that is of the Syn variant), whose String element is "for" 15:47
kolibrie what are :: and -> ?
gaal that's haskell typing syntax
:: reads "has the type" 15:48
kolibrie so Syn has the type String
gaal the type String -> [Exp] -> Exp
Juerd Does that mean it turns a string into a list of exp, and that into an exp?
gaal which means that if you give it a String and an array of Exp, you get an Exp 15:49
Juerd Oh, almost
kolibrie why do they both use an -> then?
gaal when coming from a non-functional point of view, read type signatures this way:
a -> b -> c -> d -> e
function taking a b c d and yielding e 15:50
but of course there's a trick:
Juerd Hm
Multiple arguments?
gaal if you call this function with fewer args than it "needs", it doesn't blow up
Juerd Why not a, b, c, d -> e or something like that? 15:51
gaal it returns a function waiting on the rest of the arguments!
this is called autocurrying
so for example:
(+) 1 5
"add 1 and five"
(+) :: Int -> Int -> Int
lumi Read it: a -> (b -> (c -> (d -> e)))
gaal but if I say 15:52
(+) 1
I get back a function
azuroth is impressed. looks like I have a parser that supports variable associativity and precedence in 40 lines
gaal its type is Int -> Int
and I can apply that function to an int, and get.. another int, one higher.
(+) 1 is "the function that adds one to its argument" 15:53
kolibrie so any number you pass it will have one added to it and be returned
gaal this is also known as partial application, or sometimes, a section.
you can give this function a name, too:
let inc = (+) 1
you can also write this
let inc = (+1)
lumi gaal: section is the name for the (+ 1) syntax, I think 15:54
kolibrie do all haskell functions autocurry?
gaal lumi: I think you're right.
15:54 vel joined
gaal kolibrie: yes. 15:54
kolibrie ok
gaal in fact, all functions in haskell only take one arguemnt!
so when you see "f 1 2 3 4"
azuroth and sections can let you curry the second arg, instead of the first
gaal it's like saying 15:55
((((f 1) 2) 3) 4)
(well, they all take one argument except for those that take one.) 15:56
lumi Like (** 2) is raise-to-the-power-of-two, and (2 **) is raise-two-to-the-nth-power
gaal so how is Exp a tree? look at the slide with the graphic again
kolibrie looks
gaal the node for Syn, for example, had a name to it, and also an array of Exps
so for example the "if" node has two or three children 15:57
Syn "if" [condition, truecase, elsecase]
kolibrie ok
gaal The node for function application is defined as 15:58
15:58 TMTOWTDIt joined
gaal App Exp (Maybe Exp) [Exp] 15:58
that means it has (1) an Exp representing the function to be called
kolibrie so condition = 42, truecase = { say "hello" }, elsecase = { say "oh no!" } 15:59
gaal (2) an optional invocant, defined only in method calls
(3) an argument list
exactly
more precisely, condition == Val (VInt 42)
because the condition must be an Exp also, not a haskell native type 16:00
you can see that in the graphic:
under the if there's an extra Val box, not immediately 42
and note that Val is one of the variants of Exp
kolibrie and so the condition, being an Int, uses line: Value Val
gaal yeah. of course, it could have been a much more complex condition 16:01
for example, an equality test
42 == 42
which in Perl really means &infix:<==>(42, 42)
that is, a function application 16:02
in that case, the condition would be replaced by a subtree which was a little more complex:
App (Val (VStr "&infix:<==>")) Nothing [Val (VInt 42), Val (VInt 42)] 16:03
see if you can mentally parse that
kolibrie concentrates really hard
gaal you know that App A B C 16:04
start by distinguishing A, B, and C 16:05
kolibrie A = (Val (VStr "&infix:<==>"))
B = Nothing
C = [Val (VInt 42), Val (VInt 42)]
gaal exactly
kolibrie but I forgot what A B and C are 16:06
B is invocant?
gaal then look at the data definition again
oh, in terms of what they mean
yes
that can use a little commenting :-)
A == Exp evaluating to whatever function we're invocing 16:07
B == invocant, maybe
C == argument list 16:08
*invoking
now, it may be the case that the expression resulting in the condition is even more complex
kolibrie wait 16:09
gaal if 42 == get_answer() { say "..." }
kolibrie so for that line, the function we call is (Val (VStr "&infix:<==>"))
there is no invocant
gaal yes
right
kolibrie and we have been given two arguments
or a list with two elements as our argument 16:10
gaal yeah. it's somebody else's job to check that == really wants two args.
yes
kolibrie ok, /me is ready to continue
gaal cool. have a pugs handy? 16:11
kolibrie yes
gaal run it with nothing on the command line
kolibrie ready
gaal and write ":d 42 == 42"
then ":D 42 == 42"
the first one is a little abbreviated 16:12
the second is actual runnable haskell code
kolibrie nifty
gaal (which also contains annotations, like in the slide after the graphic we've been looking at)
kolibrie so, I see the stuff we were just talking about at the end of the second one 16:13
but a lot of stuff at the front I don't understand
gaal there are other representations for this AST, I think we saw the YAML one together last time we talked, no?
kolibrie yes 16:14
gaal yeah, that stuff is Ann nodes, which add more info about parts of the tree
mainly position
so that if there's an error message, it knows to tell the user where the error was
MkPos "<interactive>" 1 2 1 8
that means "in the interactive pugs, from line 1 col 2 to line 1 column 8" 16:15
kolibrie ok
gaal it also happens to be parsed as a bunch of statements, and there's a spurious Noop in there, but you can ignore that 16:16
kolibrie on some of those lines in the data definition, there a more pieces, and some have less
gaal yes; but haskell doesn't get confused because for each type there is a precise number of pieces. 16:17
kolibrie and those types are all defined somewhere else
gaal yes, learning to find them quickly takes a bit of getting used to 16:18
if you know the general data type name (like Exp), you can grep around for /^data Exp/
kolibrie and before this parser is run, strings have already been converted into tokens by something else 16:19
gaal but if you didn't know that Syn belonged to Exp, you may have had a little harder time finding that
I don't have a great solution to that
but when you see this kind of thing, start from the beginning
in this case, you see Stmts Noop (...) 16:20
we happen to know that Stmts is a constructor in Exp
Stmts :: Exp -> Exp -> Exp
that is:
it takes Noop and another (long!) Exp, and together they are one Exp of variant Stmts 16:21
if you have a Perl block with some stuff:
{ A; B; C; }
then the statements inside the block can be arranged to look like Stmts A (Stmts B (Stmts C Noop)) 16:22
kolibrie Stmts has the type Exp -> Exp -> Exp (or, it takes two Exps and returns an Exp)
gaal yes.
integral do I need to do anything to add a new .hs file to the build? 16:23
gaal this isn't really important, except that it provides us with a way to talk aobut a bunch of statements together.
integral: add it to Pugs.cabal.in
integral ah!
gaal and to the clean directive in Makefile.PL
the second bit may not be needed 16:24
integral ack! why? Makefile.PL already uses glob to *search* for .hs files automatically
gaal integral: why what, why can't this only be defined in one place? I don't know the good answer to it, but I do see why cabal wants to know about files 16:25
integral well it's actually wanted to know module names, so I guess it's slightly different :-/ 16:26
*wanting
gaal we can probably engineer things so as this is more automatic, but that'd add some more complexity to our glorious build system
commits welcome, in other words :) 16:27
integral I hope that "glorious" was sarcastic ;-)
gaal kolibrie: so, have things started to sink in?
integral: sarcastic? well, we do use Haskell during build... 16:28
kolibrie gaal: somewhat. I'm going to paste an example 16:29
pasteling "kolibrie" at 66.239.158.2 pasted "gaal: this tokenizes, but how do I define what a sentence must contain" (47 lines, 1K) at sial.org/pbot/16114 16:30
kolibrie I got that from an example, and don't really understand parseToken 16:31
16:33 Draconit_ joined
mauke do you know choice? 16:33
gaal let me rewrite it in hopefully more readable form
kolibrie choice is like alternation, I think 16:34
16:34 kulp joined
mauke yeah, choice [a, b, c] is a <|> b <|> c 16:34
do you know >>=?
kolibrie no
mauke ah, that's the problem 16:35
azuroth not quite related, but how does <- work? it seems like magic
integral mauke: no, it's mzero <|> a <|> b <|> c, it foldls with mzero, not foldl1
mauke d'oh
azuroth: you mean in do blocks? 16:36
azuroth yeah
mauke m >>= \x -> y is the same as do { x <- m; y }
<- is syntactic sugar for >>=
kolibrie ok, that almost makes sense 16:37
kolibrie let's it sink in
azuroth ohh! I thought the ; was, for some reason
mauke (>>=) :: (Monad m) => m a -> (a -> m b) -> m b 16:38
gaal don't worry about that monadic stuff yet :)
kolibrie hmm, don't know => either 16:39
mauke => is special syntax in type declarations
the things before it are class constraints 16:40
in this case it says m must be an instance of class Monad
in our example it's like (>>=) :: Parser a -> (a -> Parser b) -> Parser b 16:41
gaal that's probably a mouthful right now
kolibrie so it could also read: (>>=) :: m a -> (a -> m b) -> m b
mauke p >>= f runs the parser p, calls f with the result, then runs the resulting parser
gaal consider parseToken to mean something like 16:42
parseToken = choice [ A, B, C, D, E ]
16:42 kulp left
gaal where the return value of each of these is what's on the right of the >> or >>= 16:42
this example is a little golfy. 16:43
kolibrie so parseToken defines which subparsers to run on the input
and each subparser tokenizes what it can
gaal kolibrie: they are tried in order, and the parses backtracks when one fails
in Pugs style, that would have been
parseToken = choice [ ruleSpaces, ruleEnding, ruleName, ruleWord, ruleChar ] 16:44
mauke huhu, fmap ((: []) . TokenEnding) ending
azuroth what does { e <- getInt; f <- getInt; return (e - f)} monadify to? {getInt >>= \e -> (getInt >>= \f -> return (e - f))} or something similarly crazy?
mauke azuroth: no { }
gaal and ruleSpaces woudl be
16:45 Nouk joined
azuroth and no return? 16:45
mauke return stays, it's just a function
gaal ruleSpaces = do
many1 space
return []
mauke getInt >>= \e -> getInt >>= \f -> return (e - f) -- works
gaal ruleEnding = do
s <- ending
return [TokenEnding s]
ruleComma = do 16:46
char ','
return [TokenComma]
that's hopefully clearer
take ruleComma
the line char ',' is a sort of an assertion 16:47
if it fails, so does ruleComma
that behaves like throwing an exception
azuroth cool, thanks mauke! it makes much more sense now
kolibrie gaal: so I could replace me parseToken with those definitions
gaal yes 16:48
svnbot6 r9241 | bsmith++ | Factored out inline documentation from Pugs.Parser to Pugs.Parser.Doc.
r9241 | bsmith++ | Had to leave ruleDocBlock in Pugs.Parser for the moment.
gaal ooh integral++
mauke a >> b = a >>= \_ -> b
integral gaal: that was the easy part, now to figure out ruleDocBlock and why it needs to parse statement lists :)
kolibrie and then to make the parser fail if a sentence is not correct, do I change the definition of the new parseToken? 16:49
gaal kolibrie: what does a parser do now if a sentence is incorrect? 16:51
integral *argh* I see why it's so annoying. it's the silly recursive way ruleStatementList works :-/
kolibrie gaal: right now it tokenizes anything that matches a subparser 16:52
I want it to fail if no ending punctuation, for instance
gaal oops, I have a "ride" home 16:53
kolibrie: add assertions
in this case, it looks like you want choice [ ruleEndOfInput, ruleEndOfSentence ] 16:54
but i gotta go :)
kolibrie gaal: thanks so much!
gaal integral: worst case pass ruleStatement as an implicit param.
or an explicit one :) 16:55
&
integral best case: rewrite ruleStatementList to use these higher-order functions we keep talking about :-P
16:56 fglock joined
svnbot6 r9242 | fglock++ | - iterator_engine_p6grammar.pl - precompiled Prelude! 16:56
r9242 | fglock++ | loads in less than 1 second
integral *sigh* it's so nice how emptyExp and Noop are the same thing, but it can be confusing to read 16:59
svnbot6 r9243 | fglock++ | - iterator_engine_p6prelude.p6 - this is the Prelude file 17:02
azuroth night, all 17:03
fglock iterator_engine now has dynamic grammar - iterator_engine_p6prelude.p6 defines pod, and then immediately uses pod for documentation 17:04
svnbot6 r9244 | fglock++ | iterator_engine - cosmetic fix
r9245 | fglock++ | - iterator_engine_p6grammar.pl promoted to p6compiler.pl 17:07
Juerd rafl: When could you re-enable the automatic installation of Pugs on feather? 17:14
rafl: beep
rafl Juerd: Will do so. 17:22
17:31 chris2 joined
svnbot6 r9246 | fglock++ | - iterator_engine - p6compiler can be invoked with: 17:37
r9246 | fglock++ | perl iterator_engine_p6compiler.pl iterator_engine_p6sample.p6
17:42 putter joined, Shabble joined
putter audreyt, gaal: so is it time, and still the plan, to move functionality from Prim.hs to Prelude.pm, now that its compile-duration issues have been fixed? 17:43
fglock: ping?
17:44 szabgab left 17:45 macli joined
putter TimToady: re would a patch help, I don't think so. I'm at this moment trying to work out the design constraint space - more in a bit. 17:46
gaal putter: we can definitely start, but there's a small matter of packaging to perform before pulling the switch 17:54
precompilation works but is not engineered yet 17:55
that is to say, it has a good design but that isn't implemented yet
17:57 calanya joined
wolverian typo on peek slide37, missing } 17:59
in case that's still relevant :)
gaal sure, thanks :)
putter: right now, pugs looks only at src/Pugs/PreludePC.yml and ./Test.pm.yml[.gz] 18:01
we need the per-user cache folder
integral does Test.pm.yml get created at the same time as PreludePC by the build?
gaal integral: no, only on "make smoke" or "make test"
18:02 rantanplan_ joined
gaal this means incidentally that "make install" suckifies ./pugs :) 18:02
If I'm lucky I'll fix it over the weekend, but possibly not as there's some pizza and a family dinner both waiting to happen 18:03
fglock putter: pong
gaal fixed. 18:06
err, I mean the slides. :)
svnbot6 r9247 | gaal++ | r9268@sike: roo | 2006-03-02 20:02:42 +0200 18:07
r9247 | gaal++ | * fix missing closing brace. Reported by wolverian++.
gaal nothingmuch: ping 18:11
svnbot6 r9248 | gaal++ | r9275@sike: roo | 2006-03-02 20:08:53 +0200 18:13
r9248 | gaal++ | * perl6.vim: hilight `rule` the same way as `method`
fglock is there a way to set the filename/line in an eval(), for warnings to behave properly ? 18:17
meeting & 18:21
gaal fglock: #line? 18:29
?eval eval "say 'mmm'; \n#line 44324\nwarn 42" 18:30
evalbot_9248 OUTPUT[mmm ] undef
gaal ?eval eval "say 'mmm'; \n#line 44324\nsay Carp::longmess"
evalbot_9248 OUTPUT[mmm ] undef
gaal oof unsafe
18:31 Draconit joined
gaal but that oughta work 18:33
i have to go now... &
18:47 OuLouFu joined 18:48 axs221 joined
axs221 I've been trying to decide on learning either perl or python. With Perl6 coming out, will learning Perl5 still be useful, or will 6 be too different so that I'd have to relearn most of what I would learn? 18:51
mauke learn all three of them 18:52
axs221 eventually i might, but i'm wanting to just learn by book right now, i bought a python book 18:53
i'm thinking about taking the python book back and getting a perl book, mostly because of the greater demand in the market for Perl, atleast in job search sites, although from what I know Python sounds easier to learn for someone relatively new to programming like me 18:56
mauke yes, probably 18:58
19:02 Nouk joined, mauke joined 19:03 dduncan joined 19:12 vborja joined 19:17 beppu joined 19:21 vborja is now known as handon 19:22 nothingmuch joined 19:23 ruz joined, Draconit joined
putter gaal: I don't quite understand. are you saying src/Prelude.pm still has constraints on size (maybe on install??), and thus shouldnt be non-trivially enlarged at the moment? or that it can be? 19:24
fglock: #line N "filename" which I think the next line is considered N, but I'm not sure. 19:25
yes, now sure. ;)
(used in the "I really should put a copy in common" somewhat cleaned up spike backtracking regexp engine thing - makes syntax errors in evals much clearer. ) 19:26
svnbot6 r9249 | bsmith++ | * Rewrote the looping logic in ruleStatementList so that the recursion is 19:27
r9249 | bsmith++ | factored out into a higher-order function.
r9249 | bsmith++ | * The above allows ruleDocBlock to be moved from Pugs.Parser to its proper
r9249 | bsmith++ | place in Pugs.Parser.Doc.
putter axs221: if you dont have experience coding in other languages, learning python is more likely to start you off with good habits than learning perl. 19:28
integral *sigh* only 50 lines smaller though
putter ;) 19:29
were expecting a greater code shrinkage? 19:30
integral just wishing that it'd suddenly turn into one when I commited ;-)
putter lol
nothingmuch hola 19:31
putter hola, como estas?
nothingmuch estas beseder
putter goes to google language tools...
nothingmuch a friend of mine always asks me como estas
she knows some spanish/portugise phrases from TV soaps 19:32
i always try to spoil her fun
19:32 handon left 19:33 Draconit joined
putter :) google translates to "to beseder". had to check OneLook to make sure that was a googlism, and not a nifty word which is similar in both spanish and english :) 19:33
hmm, so I still dont know what it means. 19:34
nothingmuch beseder is hebrew for "OK"
i don't know what "estas" means
she would always laugh at me and say "enough nonsense" or something of the sort
19:34 Draconit joined
nothingmuch anyway, my sister's birthday party was a big success 19:34
and now I'm off to visit a friend 19:35
putter aaahhh. como=how estas=are you esta=i am etc
fglock nothingmuch: I speak portuguese in case you need :)
nothingmuch i hope i'll survive, i haven't seen her in weeks and she is touch about that kind of stuff =)'
putter yay. was surprise?
nothingmuch fglock: ah, good thing
putter: yes
we made her a Moomin cake house
putter win
nothingmuch thanks to elaine ashton
and we bought her an ipod
and she is delighted 19:36
putter Moomin... putter goes back to google... later...
nothingmuch en.wikipedia.org/wiki/Moomins
fglock hi putter - bye putter
putter ah, it's great when a plan comes together :)
nothingmuch www.axis-of-aevil.net/archives/2005...omins.html
putter hi fglock, bye? fglock
fglock quick question
ping? 19:37
fglock pong
putter are you doing any <expr> expression/operator parsing stuff?
fglock (I said bye because you said later...)
putter oooohh.
putter goes to look at moomins :) 19:38
fglock mm - yes, more or less
nothingmuch putter: it's a series of children's books for adults
fglock I didn't implement operator precedence yet, but it does espressions
s/expressions/
see: iterator_engine_p6sample.p6 19:39
putter ah, en.wikipedia.org/wiki/Moomins
19:39 vborja joined
nothingmuch ciao guys 19:39
putter: yes, just like i linked you =) 19:40
fglock putter: it compiles: '1' infix:<*> '1' infix:<+> '1';
putter oh, duh. saw the first link, explored, ... 19:41
fglock if you try 'print' on that it might work
putter which there was more description of... content? perspective? ie, not cast and geography, but what they do/believe/etc. 19:42
fglock: err, so where is precedence handled? 19:45
fglock putter: there is no precedence yet - first come, first compiled. 19:47
putter got it.
fglock first priority was bootstrapping, which is mostly done
now there are many possibilities - adding precedence is one of them 19:48
I'm not sure what to do next - would like to hear some opinions
performance is quite good
there is the object system work, 19:49
completing the rule syntax,
putter ok, so is a correct one-liner "interator is like the regexp/parser spike, but bootstrapped to self hosting, and using p6 rather than p5 syntax regexps"? other key differences?
fglock packaging to CPAN,
the whole syntax is defined in p6 19:50
see: iterator_engine_p6prelude.p6
I could move the compiler to p6 now :) 19:51
putter right. parser_spike.pl cheats a bit to get parser_spike_target.p6 working. 19:52
:)
hmm... what to do next... 19:53
fglock there is some cheating going on in making { return } blocks work before bootstrapping - but after boorstrapping it can go away - it uses a source filter 19:54
putter I'm afraid I haven't even looked yet at the new -CParser-YAML (or somesuch) output. is it something you could derive from your parse tree, so there is a cross check? maybe a cross check. the trees may be too tied to parsing strategy to compare easily. 19:55
fglock I could generate YAML if I knew the nodes, but I'm afraid it is too complex for now - it can be left for later 19:56
being able to read YAML would be nice, you could reuse compiled things from pugs to i_e
i_e nneds a name, maybe a directory :) 19:57
s/nneds/needs/
putter you could start pouring Parser.hs into p6 rules... 19:58
start working through the sanity tests, eyeballing for reasonable parses...
misc/IteratorEngine ? ;) 19:59
easy enough to rename. svn++ 20:00
fglock Puggie - Pugs little brother/syster
putter *shudder*
:)
PerlParserPrototypingPlatform 20:02
fglock I'd like to have more opinions on the subproject, before working too hard on it
putter lol
fglock it was very successful as an experiment 20:03
putter ooo, PerlParserPrototypingPlatformProject
fglock that's java-ish
putter makes sense. audreyt said something about attempting a pass over the t/ and ext/ code using... something. a new parser? implementing the "sandwich" parser concept to proof it? 20:05
re very successful, indeed. congrats. :)
20:05 larsen joined
putter I seem to have backburnered my backtracking engine (still mostly p5 syntax - next step was to be an operator parser), when the spec for one of my "solid places to stand touchstones", Match, started fluxing. Think I may fork a Natch class, that I can keep stable and do whatever the engine needs for now, without worrying about spec. 20:08
Hmm, and perhaps more importantly, the operator precedence parsing turned out to be a messier proposition than I anticipated. 20:11
In retrospect, I should have / should now hack something, to get back to a parser running, and some semblance of p6, so we could have two implementations chasing each other. maybe. or maybe that would simply be a waste. 20:12
fglock putter: operator precedence is easy - it just take a while to optimize (if needed) 20:13
:) not a waste, but we could have a common API such that we can interchange engines
putter not in the presence of pre-, post-, and especially circum- fix.
re simple.
fglock putter: that's a grammar problem (solved) - all we need is to present options to the grammar in the proper order. it should be about 50 lines of code 20:14
putter I hypothesize pge is not quite doing the right thing re close tokens on cirumfix ops. and there are assorted "make for extra work" bits like operators having a "no whitespace allowed before me" flag. assorted cruft. 20:15
fglock bbiab 20:17
putter so your parser, given ... * ,,, with * rule defined early, looks ahead through the input to match *, with two unprocessed strings ... and ,,, , and then recurses on them? 20:19
Regarding using an alternate regexp engine in perl5, I did a design space grovel - it would take something like four pages to write up - isn't going to happen unless someone _rteally_ wants it. 20:21
Most important point is most of the difficulties lie in trying to swap engines in a way which is _transparent_ for _humans_ coding _perl5_. 20:23
for p6 _on_ p5, rather than _in_ p5, with a compiler emitting arbitrary p5 code, the only(?) residual issue is exporting qr//'s to native modules. 20:24
So in a p6 context, I don't think the exercise is worth the pain. Questions about "what pain?" entirely welcome. :) 20:25
My key uncertainties/questions are:
fglock re: ... * ,,, - yes, and it backtracks if fail
re: "no whitespace allowed before me" 20:26
it can be specified in the grammar, I believe
20:27 calanya left
fglock re: qr// - I could write an emitter for that, so we can choose between rule or regex (besides, we can use regex instead of rule when possible, which is faster) 20:28
putter Can XS somehow set $1 to a non-string (an object)? Is there some way to create an arbitrary variable which gets scoped just like $1 does? Regards reentrancy - is it strictly capture which is broken? does everything not dependent on captures still work? I think those are my big fuzzies.
fglock re: $1 - it's good to have TimToady around :) 20:29
putter :) and japhy.
and bsb
and ... :)
fglock bbiab 20:33
putter Oh, and the tech for playing with regexp's that I know about is, well, source filters of course, bsb's "regexp which calls matcher, gets capture offsets, and correctly sets up capture vars", the use re 'debug'-ish regcompp and regexecp hooks, and... I dont know of any other tools to get leverage around here, short of playing with opcodes (p5, and regex) or patch p5. 20:34
If anyone knows of any other candidates, I would be most interested.
buu Um, Regexp::Parser? 20:36
putter I should mention the hooks are basically providing a way to fake up regexp struct's, both as an api into the perl matchvar machinery, and as a way to do qr// which _is_ reentrant (the hooks are reentrant, its just the engine which isnt - thus the one downside of bsb's device). 20:37
buu: Yes, a good thing to mention. Thanks. (and there is a copy with some bug fixes in misc/pX/Common/Regexp-Parser, also with some additional tests, some failing...) 20:39
20:39 SamB joined
PerlJam I know it's been a while since I've paid attention to pugs when I look at #perl6 and I haven't the foggiest idea what's being discussed :-) 20:41
putter This is the p5p perl5-porters sidebar on #perl6. ;) 20:42
PerlJam Oh. I haven't read p5p in *years*, so that makes sense.
fglock PerlJam: we are porting p6 to p5 (kind of) 20:43
buu Yay!
PerlJam fglock: good luck! I'm of the uninformed opinion that it's Too Hard :-)
fglock PerlJam: no it's not :) 20:44
putter Somewhat off topic. You dont need it for pugs. Nor for a pugs backend running on p5. Nor for a completely independent p6 parse-compile-emit implementation on p5. You only need these silly time-absorbing games if you attempt to give human p5 programmers p6-ish regexs they can use transparently on p5. eh.
fglock it's just that there are _too_ many ways to do it
putter: right 20:45
putter though to be fair, some of the stuff turned up could end up being useful for the p6 on p5 stuff interoperating with native p5 modules.
PerlJam putter: twould be easyish if PGE were a library you could embed. 20:46
fglock putter: I'll translate i_e to p6 anyway
putter I'm basically puttering. Could be working on moving Prim.hs to Prelude.pm, but it's not clear it time for that. Could getting the somewhat cleaned up bactracking stuff at least checked in, or continue work on operator parsing, but with fglock making progress, the presure is off, and it's just not -Ofun. So I've spent 10-ish hours being random. 20:48
fglock: re _too_ many ways, no. the number of possible tools certainly makes analysis harder. but I believe the real problem, at least for a full bore "use EverythingRegexyNowLooksLikeP6; ...", is that you simply cant do it. 20:51
fglock goes read Prim.hs 20:53
putter PerlJam: re PGE, the engine is actually the easy part :). sort of. or at least the part we've sort of kind of already done. the long-term as-yet-unresolved puzzle is the pragmatics of integrating it with unpatched perl.
fglock putter: OT - do you know of an example of using '&' in a rule? (conjunction) 20:54
PerlJam Hmm. unpatched perl? How incestuous must the engine be?
putter though there are some low hanging fruit, like a reentrant p5 regexp engine with rules. that one could do now as an XS module for cpan. but that's a bit off-perl6-topic.
fglock: an example? any example? 20:55
abc&\w**{3} ;)
fglock putter: what is it useful for?
putter ooooohh
PerlJam fglock: Are you asking what & is generally useful for? 20:56
putter one sec
fglock PerlJam: in the context of p6 rules 20:57
PerlJam The things I can think of all involve increased specificity.
putter PerlJam: the regexp engine and the perl compiler/runtime unfortunately know each other quite well.
And p5's extensibility has never extended to permitting the user to, for example, create a Qr class which emulates qr//, overloading =~. 20:58
And there are no hooks, even at the C level, for creating the $1,etc suite. 20:59
PerlJam putter: sounds like the goal should be to get those patches in before the next release of perl5 :-)
putter One of the key values of the hooks, and their ability to create "struct regexp"'s, is you finally have a way to twiddle some of those internals. But I'm afraid it may not go as far as "please, can my $1 be an SV?". :( 21:01
PerlJam: doing p5 guts modifications to better support p6-like stuff... is something which could be added to perltodo. ;) volunteers welcome. making the regexp engine reentrant is listed there. in the "hard" section (though I'm not clear on why it's hard exactly. but I dont see perlguts hacking in my future). 21:04
Oh my, scary last words.
man perltodo (i only discovered it the other day:) 21:05
fglock: re conjunction,
fglock putter: other option - help me doing the i_e thing :P 21:06
putter you know how you sometimes use lookahead to assert some property on the following characters, in addition to the regex which actually matches them? the one gotcha is you cant assure the lookahead matches exactly... (well, you can, but it takes, oh wait, I'm not sure you can) anyway, that it matches exactly what you intended. 21:08
<hiaku>&[<line>+] 21:10
hmm, the grouping [] was unnecessary
fglock putter: thanks 21:11
putter [<mail_header>&.*?^^From: (\N+).*] $<body>:=(.*) 21:12
PerlJam: what's an increased specificity example? both of mine were assert-spec&unpack pairs. 21:16
fglock putter: .* would make it fail, isn't it? 21:18
putter you mean the .* after the (\N+) ? 21:19
fglock yes
PerlJam putter: yes, mine are similar, just that the "unpack" part isn't necessary: [<mail_header>&^^<'From:'>] && do_something_only_with_from_lines; 21:20
putter I don't believe so. I would expect the & to force the * to backtrack.
PerlJam putter: btw, beware the cut-colon! :-) (unless I'm mistaken that you meant for the : to be matched)
putter lol
right
god, both my p5 and p6 are going to be so buggy. colons and brackets and sigh, oh my. 21:21
when writing p5 regex's i now do a double-take each time I use {, "wait, that's... no, it's p5, that's ok". ;) 21:22
PerlJam The conceptual and syntactic overlap is annoying if you have to do both. 21:24
(ergo, I've sworn off writing perl6 for a while)
putter I don't think that last & example actually works... 21:25
ah
<mail_header> would have to match exactly the 5 chars "From:". 21:27
putter wonders what the big-picture status and critical path of pugs currently are. 21:28
putter goes to check audreyt's blog ;)
21:35 DesreveR joined
PerlJam putter: Are you *sure* it would have to be that way? Must they both match the exact same characters completely? Granted, if it didn't I have no idea what state it would leave the current byte/char position pointer in. 21:36
fglock PerlJam: it must match exactly the same substring
putter has anyone groveled over Parser.hs and created a p6 grammar? either a direct translation, or one using <statement> and <expr> modifying constructs like statement_control:<if> and infix:<+>, etc? 21:37
PerlJam Seems like you'd have to worry about end points then rather than just "matches at this position"
"foobar" ~~ /<'foo'>&<'foobar'>/ # What happens? Does the match fail? 21:38
putter PerlJam: just doing zero-length lookahead assertions (?=...) and <before ...> satisfies the "they both start here, and I dont care where they end" need. & exists entirely to provide the "i do care where they end". 21:39
yes, that match fails
PerlJam putter: that seems entirely reasonable.
fglock putter: re: grammar - I'll start doing that soon
PerlJam (Well, that seems reasonable as along as it's adequately documented :-) 21:40
putter translation or self-modifying (what's a better name for that?)
:)
fglock putter: I can write a grammar to parse Haskell, and use it as-is 21:41
putter "foobar" ~~ /<before <'foobar'>><'foo'>/ succeeds
any ideas on what to call a grammar which uses non-local definition of <statement> and <expr> ? 21:42
fglock putter: non-local?
putter ok, that was either you suggesting "non-local" was an ok name, or you being confused about what i meant by "non-local", thus eliminating as an option. i'm not sure which... 21:43
:) 21:44
fglock what do you mean by non-local?
putter ah, eliminated as an option. meant
something which, instead of an explicit rule statement { subrule1 | subrule2 | ... }, 21:45
PerlJam putter: a fully qualified rule name? 21:46
fglock putter: that's <@subrule> in i_e
putter one uses multi statement_control:<if> (...){...} multi statement_control:<while> (...){...} etc to fill it in. 21:47
fglock putter: you mean, statement_control is represented by an array? (or namespace thing)
21:48 pdcawley joined
putter fglock: yeah, but there are some issues... like how do (pause) 21:48
PerlJam must not have been enough svk talk here ;-) 21:49
fglock putter: please not that it is an array of rule - it is very flexible
putter statement_control is a grammatical category. it defines one of the <statement> subrules.
(my yeah, was directed at <@subrule>, not represented by an array ;) 21:50
fglock: yes but. when writing a grammar, in a first match wins engine (like | is), you carefully craft the order of the subrule list. when subrules get added by statement_control defs, someone other than the human has to do the crafting. either the statement_control infrastructure assures the @array has a nice order, or <statement> can't use <@array>. 21:53
fglock putter: you can opt to use longest-match instead of ordered-match 21:54
putter yes
fglock that would be <%statement_control> 21:55
putter and the real parser can play games like trying to massage the @array into a trie, so it doesnt have to repeatedly reparse the same stuff the same way.
fglock putter: it is cached 21:56
21:57 Arathorn is now known as Aragone
putter re statement_control, dev.perl.org/perl6/doc/design/syn/S04.html has a little bit in Statement parsing. 21:58
fglock putter: nice 21:59
22:00 ruz joined 22:01 Limbic_Region joined
Limbic_Region is there a known problem with Feather and https ATM? 22:01
fglock why is 'if' a macro? (S04)
22:02 chris2 joined 22:03 pmurias joined
putter re hash, "An interpolated hash matches the longest possible key of the hash as a literal, or fails if no key matches.", which doesnt help you distinguish between /if <expr> <block> [else <block>]?/ and /if <expr> <block> [else <block>]? [wrap_around_both_branches <mumble>]?/. comparing "if" and "if" isnt going to help. 22:04
pmurias Does any one think that useing Smart::Comments in iterator_engine.pl would be a good idea? 22:05
fglock pmurias: I like Smart::Comments, but I'd like to keep it simple (that is, no unnecessary dependencies) 22:06
putter (there shouldn't have been a ? on the <mumble> clause)
pmurias It would be a dependency only for debugging :)
And casual users don't do that often :)
at least they shouldn't have to :) 22:07
fglock pmurias: I'll check that (you mean, disable 'use Smart::Comments' when not in use?)
pmurias the use line should be comment out by default 22:08
fglock putter: re /if.../ - I don't understand the mumble part, what would it be?
pmurias and if the debuging messages are needed you just delete the #
putter re if macro, well, you need to add a regex to <statement> so you can parse it. and we can currently hang regexs off of, well, rule, and macros macro statement_control:<if> (...) is parsed(/heres the regex/) {...} 22:09
pmurias i'll commit it tommorow if youd don't mind
fglock pmurias: sounds good - I'll check the pod again
pmurias i have to sleep now, will be back tomorrow, bye
fglock putter: ok. so the macro is just a simplification of the rule thing I'm using 22:12
putter re mumble, oh, maybe a block/closure which is handed the block/closure which if would normally execute itself, but it gets to do some stuff, then maybe call it, then do other stuff. but it was just a strawman to point out that with random cruft being added to the parser, some of which will conflict with existing rules, an alt-like parse is insuffient.
putter goes to look again at what fglock is using... and then has to remember to search.cpan Smart::Comments... 22:13
fglock putter: I think there must exist just one 'if' - if you need something more complex, you define it inside the if-macro or define it with a rule, which gives you full flexibility. but there is only one entry in the statement tabl 22:14
putter: re using: only rules, no macros 22:15
s/tabl/table/
unlike multis, which are defined in another category 22:16
putter yes - the push'es, and on to what, are implied by the grammatical category stuff. statement_control, infix, term, etc. 22:17
fglock putter: yes - that would be handled by macros, if we had them
fglock thinks about adding macros - it goes to the TODO list 22:18
putter well, the whole issue of what's a macro/rule/method/sub is still a bit unspeced (or such is my recollection)
22:23 DesreveR is now known as r0nny
putter perhaps in this case, a macro (would have to be a string macro, yes?) could match and rewrite the input. a rule statement_control:<if> {...} could do whatever ruleish things the rest of the p6 grammar rules are doing, perhaps to build an ast. (caveat, I'm not sure I've seen a spec example of a "rule statement_control"). 22:26
svnbot6 r9250 | fglock++ | iterator_engine - updated TODO list - putter++ 22:28
fglock putter: a macro is just like a source filter; macro statement_control:<if> {...} is probably setting an element in the hash %statement_control in the p6 compiler namespace 22:31
%statement_control:<if> actually 22:32
it will look like: %statement_control:<if> = rule { ... } 22:33
putter it has to be a is_parsed and string-returning (not code-returning) macro, yes? 22:34
fglock in i_e, it is supposed to return AST. but a macro can return either string or AST (no code) 22:35
macros that return string will be more portable - I don't think there will be an official AST for all compilers 22:37
so AST is for internal use only
putter ah. have to go. enjoy. 22:40
&
fglock &
22:42 fglock left
svnbot6 r9251 | fglock++ | iterator_engine - pasted some comments 22:42
23:03 nothingmuch joined
nothingmuch moose 23:04
audreyt: ping
beppu stupid question perhaps, but will perl6 allow us to use chars like '!' and '?' in subroutine names? 23:11
mauke probably yes, because perl5 does it 23:14
integral yes, but perl5 allows "\0" in subroutine names. 23:16
23:23 frederico joined 23:29 TMTOWTDIt left
beppu mauke, how does perl5 allow it? I didn't think it did. 23:31
can you post an example?
mauke *{"!"} = sub {"y helo thar"}; 23:32
beppu how do you call it afterwards?
mauke *{"!"}->()
beppu I see... what I really wanted to be able to do was say things like: if ($block->is_filled?) { ... } 23:33
with the ? being part of the method name./
I didn't want to have to use typeglobs for this. 23:34
mauke heh, scheme style
beppu yeah. :)
mauke wouldn't that clash with operators? 23:35
beppu I think it would, but I figured it wouldn't hurt to ask.
mauke hmm, it shouldn't be too hard to write a lisp skin for perl 6 23:37
or perl 5, for that matter
beppu no source filters for me. 23:39
23:39 azuroth joined, Southen joined
avar integral: perl is pretty liberal about the names of everything 23:39
${'Woo, variable with spaces!'} = 5; 23:40
beppu but you have to turn strict refs off, right? 23:43
avar no 23:45
mauke yes
er, wait
avar bleh, yes
beppu ;) 23:48
we should steal back from ruby and allow '?' and '!' at the end of method names. 23:49
PS: has anyone noticed the similarity between `perl -h` and `ruby -h`. I think that's cool. 23:50
obra beppu: or just go hardcore and use _p (predicate) instead of ? 23:53
beppu that's not as cool. ;) 23:54
perl and sigils belong together.
23:57 samuel joined
avar I like ! a lot 23:57
.chomp! str
eh
str.chomp!
what does ? do again? 23:58
mauke returns a bool
avar perl 6 has that
if (?foo) ..
IIRC
along with ~ for a scalar cast.. 23:59

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!