The Return of the Journal : pugs.blogs.com/ | pugscode.org | pugs.kwiki.org | paste: paste.lisp.org/new/perl6 or sial.org/pbot/perl6
Set by stevan on 23 December 2005.
svnbot6 r8405 | luqui++ | Large, gut-mangling change to fix: 00:44
r8405 | luqui++ | my $a = [1,2,3]; my @b = $a; @b.elems # 3
r8405 | luqui++ | Incidentally, unary * works correctly now, too.
luqui ?msg svnbot6 ?check 00:45
svnbot6 r8406 | luqui++ | P6GE now has captures, in their fully-specified glory.
r8406 | luqui++ | Once backtracking controls are in, then it will be a complete combinator library for perl 6 rules!
nothingmuch luqui: please note to @Larry that your latest discussion, wrt rules reimplemented in perl 6 relates to circular prelude, etc 00:54
and that this is a very nice example of why it's so important to think of this early
luqui I guess so... 00:55
on the other hand, you can just think of the rules implementation as the beginning of the post-bootstrapping process for rules 00:56
(not to say that PGE won't be used---it most definitely will, because my implementation is a dog)
nothingmuch well 01:00
the point is that there's a reference implementation for everything 01:01
which gives us:
a. something users can refer to in doubt
(when the docs aren't good enogh)
luqui has used that for Haskell a few times :-)
nothingmuch b. something slow, but that still works
(very quick backend writing)
c. something that allows each backend to implement what they really need to implement
even in a profile oriented way
implement an opcode
profile your code 01:02
luqui I think the biggest block is the graph theory involved
nothingmuch see what the next opcode to add should be
we can simplify it a bit
make it non circular
a bit more ad hoc
or whatever
no, make it circular
luqui uh... make the circular preluded noncircular?
nothingmuch no
that was a brain fart
luqui I don't think you're going to get away from the graph theory... somebody just has to figure it out 01:03
it's probably related to a well-known algorithm
nothingmuch aye
but I have no time to think of it right now
job to demanding
luqui the JS backend is pissing me off
nothingmuch when my Catalyst stint ends, i'll get back to pugs land
right now rewritten sessions and authen/authz are getting less problems 01:04
better docs
daily
luqui cool, understood
nothingmuch things are solidifying
and after that I can just leave it to the community
so anyway
audreyt wonders if it's possible to lure nothingmuch to Jifty6 01:05
nothingmuch Jifty6?
audreyt speculative idea of perl6 port of Jifty
luqui why the heck can't I use modules from JS?! 01:06
audreyt Jifty is the thing I've switched from catalyst to for $job
nothingmuch what is Jifty?
oh
luqui for some reason it's looking for the "Parse" module, even though I'm using "Parse::Rule::Combinators"
audreyt it was once known on my journal as "Bamboo"
nothingmuch i see only Jifty::DBI on CPAN
is it still a layer on top of Catalyst et al? 01:07
? 01:08
obra It's never had anything to do with cat 01:09
It comes out on XMAS
Jifty::DBI is just the object-relational mapper we've got
nothingmuch what does it do differently? 01:12
audreyt nothingmuch: it's this continuation-capable, page-region-based, scriptaculous-powered, declaratively-modelled, absolutely-no-sql-needed-for-schema-generation, REST-action-based, buzzword-compliant thing
nothingmuch: I think mostly the idea of page regions, decomposition to REST actions, and web continuations, are quite different 01:13
obra It's also much more in the "one good way" than "everywhich way" camp than Catalyst. 01:14
audreyt also, it has ajax_autocomplete, ajax_validate, ajax_link, ajax_tangent, ajax_everything that happens to degrade happily on w3m.
obra (autocomplete doesn't work in w3m ;)
hm. 01:15
it could.
audreyt I did say "degrade" :)
and it could.
however, its default presentation layer library is really primitive (the same level as Catalyst's default Scaffold); it can definitely be improved 01:16
nothingmuch hmm
sounds cool
i don't have anything to do webapp wise in the short term future though 01:17
so my answer is no ;-)
audreyt that's fine :)
obra audreyt: yes. the presentation-layer library wants love. lots of love.
also a coherent theory behind it 01:18
audreyt and also it probably wouldn't hurt to adopt catalyst-style attribute based controller
justatheory clears his throat
audreyt if not as a default, certainly as an option
to avoid dhandler overflow
obra attribute-based controller?
justatheory: when are you releasing your web framework? 01:19
audreyt sub details : Regex('^product/(\w+)/details$') {
....
}
obra ah. *nod*
I sort of wonder if a dispatch table is a better answer than that.
at least to fit with the jifty model. 01:20
audreyt well, but it is a dispatch table :)
justatheory obra: No idea. Will probably launch some apps first.
Hopefully in 2006, though.
obra sorry. but _just_ a dispatch table.
audreyt ah. hm.
obra how many apps do you have in the pipeline?
audreyt maybe. I have no idea. have to experiment
obra audreyt: nod.
justatheory obra: One at th emo.
audreyt but that's one single thing I miss from Cat.
obra agree that we want it.
*nod* I don't like the dhandler game
nothingmuch okay, i'm off to bed 01:21
also, i must warn you guys
the reason i like catalyst is it's complete lack of one-good-way 01:22
obra it'll get a lot easier when we get a working mason plugin.
audreyt nothingmuch: indeed, that's what I liked -- and still like -- about perl5 as well
nothingmuch because that makes flexibility be thought out very hard in advanced
*advance
audreyt but sometimes it doesn't hurt to experiment with bright white castels :) 01:23
nothingmuch anyway, i'll take a long hard look at Jifty when it's out, and i'm doing web stuff again
audreyt castles, even
nothingmuch *yawn*
shit, i wanted to bike tomorrow morning
luqui audreyt, do you think Parse::Rule should do its own rule parsing, or should it just provide combinators for the perl 6 parser to call?
nothingmuch luqui: two layers
luqui nothingmuch, clearly
audreyt luqui: Parse::Rule::Parse::Rule
luqui :-) 01:24
nothingmuch luqui: i mean, two modules
luqui I was just wondering if the top layer was even necessary. Like lemon meringue without the meringue
nothingmuch make them pluggable
and then make them easy
anyway, i'm off
audreyt luqui: you can do the pmichaud thing and use a precedence parser frontend that has no concre te syntax 01:25
and bootstrap rule parsing from there
luqui yeah, I figured. So the answer is "yes you should do your own parsing" 01:26
It kinda makes sense if we want perl 6 to be self-hosting
something's got to parse the grammar spec
audreyt *nod* 01:30
porting 323 lines of PIR
probably isn't that difficult 01:31
luqui right
svnbot6 r8407 | luqui++ | Added test for [ [ x ]* ]* infinite loop, and made it pass. 01:35
luqui there, now I'm officially better than PGE ;-)
audreyt rofl 01:36
hm, I wonder how long it can take to port that 323 lines to Hs 01:37
audreyt goes trying
luqui why?
audreyt for fun, mostly 01:38
luqui ah. I wrote a regex engine in like 30 lines of Hs using the list monad
audreyt also parsec doesn't handle circumfix cleanly
luqui to try out my approach...
luqui noticed with sadness that you couldn't define circumfix operators 01:40
audreyt why? 01:42
luqui er, in pugs
audreyt er, that's another reason to write Text.Parser.OPTable now
luqui makes sense 01:43
audreyt parsec can emulate bottom up, but it's not going to be fast
in fact that's one of the slowest parts
luqui really? how strange
audreyt due to a lot of shifting in circumfix/postcircumfix
currently we dynamically construct an prec parser every time we enter BEGIN 01:44
and cache it inside the parser
which does speed up things
but I'd like something less dependent on unsafePerformEval
luqui aye
well, I guess we'll all be writing OPTable at the same time then 01:45
audreyt it's a race!
luqui however, I have to sleep now.
oh darn, I'll lose
stupid sleep
audreyt :)
good night, sleep well
if I manage to pull it off, you'll probably have a much easier time porting it to p6
luqui because it's possible to read haskell?
audreyt sir, you are really mean when you're sleepy :) 01:46
luqui what?
audreyt just go sleeping already
luqui oh, mean towards PIR. but it deserves it. :-)
mwahaha
sayonara 01:47
audreyt (yes, because haskell makes algorithm flow more explicit. but no, I don't think PIR is impossible to read)
luqui &
yeah, but you're audrey
luqui fg;bg
oh, audreyt, one more thing 01:54
how is pugs's support for continuations these days
audreyt luqui: limited to coroutines and nonreentrant &?CALLER_CONTINUATION 01:57
again, the new runcore is designed with full serializable continuations 01:58
luqui ooh, serializable... shiny
using continuations, Parse::Rule could be made quite a bit more readable
audreyt but the JS core is fully resumable in $?CALLER_CONTINUATION 01:59
luqui but I'll wait on the new runcore to convert it to that
audreyt oh? you can't just use exceptions?
try { die $obj }
luqui hmm
audreyt because you the throw out whole coros or closures, that's sometimes easier than continuations
(sometimes it's even easier just to code things as coro, as parrot hackers seem prone to do)
luqui I tried as coro 02:00
lots of repetition
I ended up explicitly looping over results way too much
CPS was a good fit
but explicit CPS is inherently hard to read
(on the plus side, It can be transliterated easily to perl 5 with the current implementation) 02:01
anyway, off to bed
luqui &
avar fg
audreyt strangely I've got the API implemented in no time 02:03
now to implement the shift/reduce rules...
dduncan commit pending ... 02:18
stevan audreyt: ping
dduncan done 02:22
svnbot6 r8408 | Darren_Duncan++ | r1549@Darren-Duncans-Computer: darrenduncan | 2005-12-23 18:20:09 -0800
r8408 | Darren_Duncan++ | /ext/Rosetta-Incubator : in Language.pod, rewrote the 'Native Primitive Types' section, particularly about numeric types, and edited some other parts
dduncan I did a fair amount of thinking about that numerics stuff ... feedback appreciated
considering I tend to buck the trend's there that I know about 02:23
stevan obra: re: Jifty, how are you doing continuations? Coro? something else? 02:24
obra we're doing app-level continuations. 02:25
"We're faking it"
stevan ah
fake it until you make it :)
obra but, it seems to work pretty well
stevan is Jifty public yet?
audreyt managing one's own heap is definitely the way to go
obra Sunday
audreyt if you control both the controller _and_ the presentation
stevan excellent
obra audreyt: we should talk some time about gc strategies for continuations 02:26
stevan has to find a new web framework & ORM solution for $work
audreyt obra: I still advocate stashing it all to the webpage
obra: zero gc on server side
stevan $client is not into the ide of using ruby
audreyt obra: so, SHA1_HMAC
obra audreyt: chmrr had some good reasons not to do that 02:27
stevan: how soon?
audreyt obra: I'd like to hear them
obra audreyt: after sunday morning ;)
but yeah, I'd like you and him to debate it.
:)
stevan obra: well, the current project ends Feb. 1, and we need to decide what to use before the next project (around March)
obra ok. so not a need-a-test-release-cut-tonight 02:28
stevan not at all :)
obra cool 02:29
dduncan fyi, from my reading, the term ORM is a bit of a misnomer 02:30
considering that many objects are essentially just multi-part types
and the plain relational model allows for those
stevan dduncan: I agree, but its recognizable :)
dduncan sure
obra it maps from Objects into a Relational Database 02:31
dduncan but I'm just pointing out that people mention object-relational as if it were different from relational
whereas plain-relational supports objects
obra In standard use, has nothing to do with an object-relational database
it's talking about first-class in-your-language objects
avar dduncan: doesn't ORM refer to stuff like defining a table as extending the definition of another table? 02:32
dduncan of course, the relational model was designed to be embedded in another programming language, rather than being something separate
the relational model says that you have sets which are related to other sets
it provides common ways of organizing and processing those
its loosely like how Perl has hashes as a first class variable type 02:33
a relation is like a hash where the key is multi-valued, and there is no hash value
and all hash keys have the same number and type of values 02:34
so one hash key is a tuple
the whole hash is a relation
of course, that assumes no duplicates allowed or undefs/nulls
a multiset is like a relation but that you use the hash value to count occurances where the key occurs in the set more than once 02:35
adding nulls complicates things
put another way, 02:36
a relation is like a junction of array refs, where each array has the same number of elements
and corresponding elements are of the same type
I'm not sure if Perl supports junctions of array refs, though 02:37
oh yes
and each array element / hash key part can be arbitrarily complex type
either a scalar or multi-part object 02:38
audreyt stevan: er, sorry, pong. 02:40
stevan audreyt: :)
did you see my ramblings earlier
the new infix ops dont work since there is no send defined (at least not one I or ./pil could find) 02:41
audreyt $ ./pil -e '1`add(3)'
4
stevan :/
lemme svk up again
everything still works though,.. 02:42
Bootstrap.pil still compiles and the tests pass 02:43
is that intentional?
do_if, do_for, etc are not subject to this right?
audreyt currently .send just dispatch to ` if it's not userdefined 02:45
stevan audreyt: ./pil -e '::Class`is_nil`cond(-> { true }, -> { false });'
audreyt so it's operationally the same
stevan I am sorry... I didnt look deep enough,.. that last bit fails 02:46
other ones work
but it seems it fails when chained
audreyt fixed 02:47
stevan :)
audreyt brb
stevan resumes reasing about R/OM while he waits for svnbot
svnbot6 r8409 | audreyt++ | * allow chained `method calls. 02:49
obra R/OM? 02:55
stevan obra: this article www.onjava.com/lpt/a/6366
it is an interesting point of view on the idea
basically the idea is that in some cases,.. the R should lead the O 02:56
where most ORM tools (especially Java ones) force the other way around
audreyt: the null instance now is ::`new_opaque({}) 02:57
I kinda like that
r0nny yo 03:01
howz the state of the object system atm ?
stevan r0nny: if you dont need imperitive features like assignment and loops,.. it is pretty good :)
r0nny: we have the entire class system, and a slightly broken (but usable) role system, and the foundation of the boxed type system 03:02
dduncan stevan, that article has good ideas
stevan but it is all still in PIL
dduncan: yes, i think so too,.. it is particularly applicable to the work I do, where the data model really should lead 03:03
r0nny n8 then, and marry christmas
stevan r0nny: same to you :) 03:04
dduncan its certainly safer for the data model to lead, since that lets the database do what it's good at, and helps make the data safer
stevan dduncan: yes, and it helps in seperating responsibilties too,.. if the DBA owns the database, rather than maintains the mess the ORM system created,.. he/she is more likely to be happy 03:05
stevan has decided that he is a really really *really* bad DBA, and should never be trusted with control of anything but a really simple schema 03:06
so $coworker is now in charge of it since he is a good DBA 03:07
audreyt: I have just converted everything to use ` and . appropriately, and I am running the tests to make sure all is well 03:09
if all goes well, can we make it so that ` and . /only/ work for prims and user-defined respectively? 03:10
I would like to make sure there is nothing I missed which might bit me later 03:11
svnbot6 r8410 | audreyt++ | * Text.Parser.OpTable -- Haskell port of PGE::OPTable. 03:19
r8410 | audreyt++ | All data structures and APIs are there; need to fill in
r8410 | audreyt++ | implementation now, but it should help luqui++ a lot already :)
audreyt yes 03:28
what about !
does it not consult MRO?
stevan no, that should
it is for user defined private methods right?
audreyt yes
my method foo 03:29
stevan yes, that doesnt use the MRO anyway
it consults the current class, if it doesnt find the method,.. it fails
sorry,.. (re-reading previous statements),... ! should *not* consult MRO 03:30
audreyt good, then we are still on the same page
stevan no ;) 03:32
audreyt oh? ;)
stevan ok, you convinced me
audreyt heh :p 03:33
do you think $closure`() is hard to type? should we introduce sugar $closure() ?
audreyt feels like a sugar dealer in the past week or so 03:34
stevan well, if we did $closure(), then we need to do -> { }do_if(...) 03:35
to keep things consistent
audreyt no, it's just for `()
stevan or, we can do $closure() and -> { }`do_if(...)
audreyt yeah
but then $closure.() will mean different things as $closure()
stevan hmm 03:36
maybe it should be $closure`() then
audreyt k
stevan I think you and I will likely be the only ones ever writing this anyway (at least for the forseeable future)
audreyt yup 03:37
let me finish OpTable.addToken first 03:40
pmichaud's clever algorithm is a bit too clever
I'm not sure if it is actually correct on precedence handling
stevan ok, take your time,.. I need to think/read about roles anyway
stevan wanders over to Scala 03:41
audreyt fwiw, while our dynamic side calculus resembles YARV, I'd like the static side to resemble vObj/lambdaCube 03:43
(vObj is what Scala uses internally) 03:44
they have this new self-hosting compiler for Scala written in Scala that I kept meaning to check out but didn't got the timeslice
stevan do you have links for vObj lambdaCude papers? 03:45
stevan needs to just get an ACM membership so he can read all the good papers 03:46
audreyt all vObj papers are in scala.epfl.ch/docu/related.html
one implementation of lambda cube is in Jhc's "data E" type: 03:48
repetae.net/john/repos/jhc/E/E.hs
lambda cube related papers: www.cs.ru.nl/~henk/papers.html 03:49
Jhc's take on reified types: repetae.net/john/computer/jhc/jhc-r...class.html
stevan which paper is most relelvant to Lambda Cude 03:50
audreyt well... as usual we turn to the great educator SPJ 03:51
research.microsoft.com/~simonpj/Pap...henk.ps.gz
azuroth test 03:53
audreyt stevan: henk is coauthored by emeijer, who brought much of the type calculus to VB's soft typing which I hope to emulate 04:08
stevan ok
this stuff is a little over my head,.,. but I will attempt to grok as best I can (well at least the vObj stuff so far) 04:09
audreyt I suggest you focus on vObj for 6.28.0 :)
we won't need \Cube until 6.2813 04:10
stevan ok
can you explain "dependent types",.. I am confused by the term
audreyt ok
stevan the "nominality" stuff sounds similar to @Larrys "class but undef" thing 04:13
audreyt normally we have functions, which is a term whose value depend on other terms 04:14
then we also have higher-rank types like "Hash of ::t", which is a type that depend on other types
stevan ah 04:15
ok, this makes sense then in the context they are talking about
audreyt there are also term depending on types, which is typecasing
stevan they are talking about the SML module system... where stuff like that is easy to express
audreyt or compile-time ref()
so far so good?
stevan so far so good 04:16
typecasing or typecasting?
audreyt typecasing... think want()
stevan ok
audreyt which is a term whose value depending on the (return) type
finally we come to dependent types
which is a type that depends on terms
for example, the "n-Tuple" type 04:17
depends on a concrete value "n"
stevan Hash of foo()? where foo() returns some kind of type?
audreyt think shape()
:shape, even 04:18
stevan ok
SML functors come to mind
types which depend on types whose value may not be known at that moment
am I on the right track? or smoking crack? 04:19
audreyt that's still tyop
type operator
you have to depend on some concrete runtime term's value
it's not expressible in ordinary ML or Haskell98
arrayOfThreeInts :: Array 3 Int 04:20
here the "3" is not a type, but a normal term (literal value)
that's a dependent type
stevan oh 04:21
hmmm
how is this useful? is there a real world context for this?
audreyt is arrays of three ints not useful?
boundedNumber :: IntBetween 3 4
stevan ok 04:22
audreyt with GADTs haskell gains the ability to express dependent types
stevan now I am thinking about Ada subtypes
audreyt yes, but does Ada check them at compile time, or are they all turned into assertions?
stevan I think Ada checks at compile time,.. but thats because it's Ada, and planes are better left in the sky :) 04:23
audreyt also, bounds are easy... what's harder is
stevan but conceptually it is helping me make the leap
audreyt n = 4
foo = 2 :: IntBetween 1 n
or even
makeFoo n = 2 :: IntBetween 1 n 04:24
in perl6:
stevan so runtime type construction?
audreyt sub makeFoo ($n) { my Int where { 1..$n} $x = 2 }
yeah
stevan oh
very cool
audreyt yup, very cool. 04:25
I think it's definitely 1)useful 2)sexy 3)finally somwhat understood
stevan what would the return value of that p6 function be?
audreyt it will be 2
unless $n is < 2
where it is an error 04:26
stevan ok
audreyt and presumably you can check it at compile time if you see makeFoo(0).
stevan but it seems that this brings the power of compile time type-checking to runtime
or does it bring the flexibility of runtime to compile time (where it can be checked)? 04:27
audreyt both. 04:28
it... blurs the boundary between runtime and compile time.
stevan even better :)
audreyt which perl has been doing (very crudely and dangerously) for ages.
stevan yes 04:29
I see why this is important for us
audreyt and now there's finally some light in formalizing this :)
stevan one more questions 04:35
when they say in the vObj paper ... "objects and classes with type members"... they mean the object has attributes which are types? 04:36
audreyt yes. 04:39
stevan like inner classes?
class Foo { has Class $.bar = Class.new(name => 'Bar'); } 04:40
audreyt something like that, yes
stevan or maybe
class Foo { my class Bar { } }
audreyt aye 04:42
stevan class Foo[::T] { my class Bar is ::T { } } 04:43
would that be a "dependent type"?
since the value of T would be deferred until later in the program 04:44
audreyt well, if all this happens at compile time, no, that's just a type operator 04:45
a type depending on type
class Foo[&f] { ... }
is another matter 04:46
stevan my $foo = new Foo[@ARGV[0]] :P
audreyt now, allowing that syntax makes it effectively dependent typing, yes.
stevan how would &f be used in the body of Foo?
audreyt it would presumably be invoked. 04:47
stevan and its return value be used in a type annotation/definition?
audreyt presumably, yes. it's not really p6, just for illustration 04:49
I'll brb
stevan ok
stevan sleep & 05:08
asavige anybody there to answer my dumbass p6 questions? 05:24
azuroth I'll give it a go, if they actually are dumbass :-) 05:26
asavige ok, thanks i've got @lol = ( [ '1a', '1b' ], [ '2a', '2b' ] );
now I want to iterate for @lol -> $t { say "1st='$t[0]' 2nd='$t[1]'" } works 05:27
but i want to write: for @log -> $x, $y but I don't know how to flatten
05:28 wolv is now known as wolverian
dduncan yes, "Laugh Out Loud" sounds like typically dumbass conversation :) 05:29
azuroth ?eval my @a = (['1a','1b'],['2a','2b']); for @a -> [$x,$y] { say $x ~ $y; };
05:29 evalbot_8404 is now known as evalbot_8410
evalbot_8410 Error: unexpected "f" expecting ";", statements or end of input reserved word 05:29
azuroth ?eval my @a = (['1a','1b'],['2a','2b']); for @a -> $x,$y { say $x ~ $y; };
evalbot_8410 OUTPUT[1a 1b2a 2b ] undef
azuroth ?eval my @a = (['1a','1b'],['2a','2b']); for *@a -> $x,$y { say $x ~ $y; }; 05:30
evalbot_8410 OUTPUT[1a 1b2a 2b ] undef
azuroth hmm. sorry, I've got no idea
asavige azuroth: thanks (I groped around with those variations too;-) 05:32
azuroth that's alright. I'm gonna try installing nvidia drivers, brb! 05:33
svnbot6 r8411 | audreyt++ | * Text.Parser.OpTable: addToken fully implemented. that's 50% of the functions... 06:17
r8412 | audreyt++ | * Text.Parser.OpTable: Helper classes to construct optables easily 06:59
obra having fun, audreyt 07:04
audreyt yeah, cloning PGE. 07:07
almost there now.
(the shift/reduce part; the full grammar part is luqui's job)
bbiab & 07:14
svnbot6 r8413 | audreyt++ | * Text.Parser.OpTable: DynStr node, which covers callbacks
r8413 | audreyt++ | to PGE::Match and other callback-code nodes.
obra Next task: implement PGE in Sendmail.cf 07:17
dduncan question: 07:59
svnbot6 r8414 | asavige++ | added some more split tests
dduncan what is the perl 5 equivalent of aliasing a lexical variable from another, like := does in perl 6? 08:00
a bit of experimenting I tried with '\' and '*' didn't seem to work
perhaps the * only works on globals
'*' meaning glob 08:01
asavige dduncan: S2 says "typeglobs are gone. Use binding (:= or ::=) to do aliasing" 08:06
typeglobs don't work on lexicals though
dduncan yes
but I'm trying to make some perl 5 code act like perl 6 08:07
in perl 6, I can say %foo := %bar
what is the perl 5 equivalent of that?
and note that %foo and %bar are both lexicals, declared with 'my'
asavige k. sorry, don't know how to do aliasing on lexicals
dduncan specifically, I want to do something like this: my %hash; \%hash = $hash_ref; 08:11
buu dduncan: Tie, generally.
dduncan: Or scary glob vooodooo
dduncan if anyone else knows, glad tohear it
buu try
dduncan yes, what is the scary glob vodoo?
buu eval: %h1 = qw/foo bar/; *h2=*h1; $h2{foo} 08:12
dduncan it seems that *foo = \$bar doesn't work
buu pokes eval
oh
dduncan that may only work for globals
buu eval: %h1 = qw/foo bar/; *h2=*h1; $h2{foo}
p5evalbot buu: Return: bar
buu eval: %h1 = qw/foo bar/; *h2=*h1; $h2{foo}='wtf'; $h1{foo}
p5evalbot buu: Return: wtf
buu You can make pseudo lexical globs with local. 08:13
Obviously you can't refer to the globs of lexicals.
asavige maybe look at Lexical::Alias on CPAN by Japhy?
integral lexicals don't have globs. There's also Data::Alias
dduncan my situation is that I'm trying to emulate 'trusts' in perl 5
jdv79 trusts?
dduncan I have a function that exports \%foo from one package, and that func is called by the other
what the func returns is a hash-ref 08:14
I want a lexical hash in the calling package to point to what the hash-ref does
buu Abuse tie.
It's fun and profitable!
integral I think you'll have to
dduncan it should be a non-tie solution
buu Too bad.;
dduncan I was hoping some simple symbol manip would work 08:15
buu Obviously symbol manipulation works. /me points upwards
integral not for lexicals. there's no symbol to manip unfortunately. But PadWalker, Lexical::Alias, Data::Alias might help
dduncan I'll look up the Lexical::Alias
asavige Lexical::Alias is done at XS level 08:17
jdv79 are we there yet?
buu Yes.
dduncan oh ... don't want to use xs 08:18
buu aliases here to there.
You are whiny.
integral dduncan: umm, you'll have to...
jdv79 the man knows what he wants and he wants it now!
:)
buu Data::Alias doesn't seem to have xs
Oh wait, I lie 08:19
dduncan if I can't find a solution, then at least I have the workaround of using the $hash_ref->{foo} instead of $hash{foo}
buu It's there at the top.
dduncan: Yes.. workaround..
dduncan but I wanted to try something new, make the code look nicer
buu dduncan: What earthly reason do you have to be so picky?
dduncan making similar-acting code look similar 08:20
its just a preference
buu dduncan: So about my question?
dduncan will look at data::alias
I wanted to do this aliasing thing as a style preference 08:21
buu So use tie
Where's the problem?
dduncan the problem was that I wanted it to be simple, more importantly 08:22
if its not as simple as $hash_ref, I'll just use $hash_ref
buu Uh
tie %x,$y; $x{y} 08:23
dduncan is that all that's necessary?
buu Yes
dduncan I thought I had to declare accessors
buu Well, yes, in the module.
But the using code is clean.
Tie actually needs a third argument.
dduncan thank you for your help 08:24
buu Hrm. sub FETCH { $self->{$_[0]} }
dduncan okay, commit pending ... 08:59
... delayed ... have to find my password again ... 09:08
azuroth tell me when you find it 09:11
dduncan I just have to so rarely enter it ... don't know why I'm being prompted now of all times 09:12
oh, that's why
the server's down
azuroth yeah, auto-remember password things always end in disaster 09:13
dduncan audreyt, it appears that the svn service is down, but the physical server's up, so maybe you can fix it remotely?
well, the perl 5 version is committed to utsl anyway ... the perl 6 version will have to wait until tomorrow, unless the server's back up shortly 09:18
luqui buu, somebody had to tell him :-) 11:56
?eval my %hash = (a => 1, b => 2); my $h = { %hash }; $h.perl 12:37
12:37 evalbot_8410 is now known as evalbot_8414
evalbot_8414 "\\sub \{...}" 12:37
luqui ?eval my %hash = (a => 1, b => 2); my $h = { b => 2, c => 3 }; $h.perl
evalbot_8414 "\{(\"b\" => 2), (\"c\" => 3)}"
luqui Umm..... why can't I commit? 13:01
azuroth dduncan was saying earlier that svn is down, even though the server is up 13:02
luqui I can update...
but the openfoundry webpage is giving me internal server error
so maybe their auth database is causing problems
good think I have svk ;-)
azuroth hmm, yeah 13:03
luqui s/k/g/
?eval sub hr() { { a => 1, b => 2 } } hr.{'a'} 13:23
evalbot_8414 \undef
luqui ?eval sub hr() { { a => 1, b => 2 } } hr{'a'}
evalbot_8414 \1
luqui ?eval sub hr() { { a => 1, b => 2 } } hr .{'a'}
evalbot_8414 \undef
Mahmoud hi 14:41
luqui hola 14:43
Mahmoud what is perl six 14:47
luqui do you know what perl 5 is ?
Mahmoud a scripting language 14:48
compiled and interporated language 14:49
i already use perl five but i head a lot about perl six.. but too lasy to read all what was said
luqui perl 6 is the next version of perl the language
okay
you can play with it in here:
Mahmoud it will be like a new language right?
luqui ?eval say "hello, world"
evalbot_8414 OUTPUT[hello, world ] bool::true
luqui yeah, pretty much
Mahmoud close to c and c++? 14:50
luqui uh.... farther than perl 5 was
azuroth ?eval my @a = (1, 2, 3, 4, 5); @a >>+<< (5, 3);
evalbot_8414 (6, 5, 3, 4, 5)
Mahmoud wtf?
luqui azuroth, don't scare him right of the bat, geez
azuroth sorry. I think it's cool
luqui j/k 14:51
Mahmoud perl six will be better for OOP?
luqui yep
Mahmoud but little slower and more resources needed?
azuroth ?eval my @a = (1, 2, 3, 4, 5); @a >>*<< 2; # double all
evalbot_8414 (2, 4, 6, 8, 10)
luqui ?eval class Foo { has $.bar; method say_bar() { say $.bar } } my $foo = Foo.new(bar => 4); $foo.say_bar()
evalbot_8414 OUTPUT[4 ] bool::true
Mahmoud print "foo";
?eval print "foo"; 14:52
evalbot_8414 OUTPUT[foo] bool::true
luqui Mahmoud, at the moment it's a lot slower
azuroth hopefully not, Mahmoud! parrot's our dedicated VM, and it's pretty fast
Mahmoud why there is no ';' at the end? canceled?
luqui but the current implementation is just a bootstrap
the idea is that it will be faster, because it will be more aggressively compiled
Mahmoud ?eval my $var; 14:53
evalbot_8414 undef
luqui (but that may be a dream given all the new features. I imagine it will end up being about the same speed overall)
Mahmoud i think OOP is always slower
just like C and CPP..
luqui well, Perl 5 had OO
it was just a pain to use
Mahmoud C has OO too
i should not talk about OOP.. 14:54
luqui it's really more about the style of programming when you're in an OO language
azuroth s/has/can do/ ?
Mahmoud i didn't do it with perl
will they include multi line comments?
c style comments
luqui heh, no
but there's always POD
Mahmoud ?eval #comment
evalbot_8414 undef
Mahmoud larry wall hates c style comments for personal reasons 14:55
luqui yeah. I don't recall the argument
Mahmoud any reason?
luqui yeah, he had some reasons, and they weren't that bad
Mahmoud i read something like 'c guys don't use it so much so why would perl guys use it?'
luqui you could write to [email@hidden.address] and ask 14:56
Mahmoud i'm afraid from perl six... it may kill the language and no one write on it
luqui yeah, that's a common fear
Mahmoud just like why people use ASP, because it's easy and lots of programmers know it
luqui but perl is losing ground as it is to python, ruby, and php 14:57
people don't seem to have a problem learning new languages
Mahmoud losing ground? explain?
luqui a lot of perl programmers are using those languages instead because they're better at certain things
python looks cleaner to many people, ruby is better at OO, and php is better for quick web stuff 14:58
Mahmoud php is faster than perl with mod_perl?
luqui you know, that ends up not mattering very much
well, faster for the programmer 14:59
not necessarily for the web server
Mahmoud coding in php is faster?
luqui uh huh
coding web sites, that is
Mahmoud because of embeding php in the html you mean?
luqui yeah. it has a high "whipupitude"
azuroth heheh. 15:00
Mahmoud it think it makes it faster for the server side too.. no?
luqui with perl you have to put something in cgi-bin, know about modules, etc.
Mahmoud, whatever
maybe, maybe not
Mahmoud the server doesn't send any thing to mod_php unless it sees <?
luqui that's not what's important to people, even though they think it is
php is certainly not faster than Mason 15:01
Mahmoud i use mod_perl by extention
luqui but Mason is a bitch to use, so people use php instead
azuroth mason?
Mahmoud < files *.pl> directive in my httpd.conf for perl-script
luqui azuroth, it's a Perl 5 web templating module
azuroth ahh
Mahmoud i hate the idea of embeding perl/php/xxx with html.. better to use template engines
luqui yeah, experienced programmers think that way 15:02
Mahmoud thanks :P
luqui but there are many unexperienced programmers for whom templating engines are crap
and it's easier to embed code in html
Mahmoud for quick and dirty tasks it's better to not use template
saving time 15:03
i think we were talking about perl six
luqui ah yes
well, that's why we want to make perl 6
Mahmoud ?eval print "foo"; 15:04
evalbot_8414 OUTPUT[foo] bool::true
Mahmoud OMG..
shoult it be capital OUTPUT?
luqui that's just the evalbot
if you did it on the console, it would be what you expect
Mahmoud i thought evalbot is a perl five to perl six convertor
luqui oh, no
audreyt that would be nice
luqui it's a perl 6 evaluator
Mahmoud heh..
audreyt but no it is not
Mahmoud how would i say print "foo"; in perl six? 15:05
luqui print "foo"
audreyt you say
integral ?eval say "foo"
evalbot_8414 OUTPUT[foo ] bool::true
audreyt print "foo"
Mahmoud no ;?
audreyt well, if you like semicolon, you can add that
just like in perl 5.
luqui they're optional on the last statement, just like in perl 5
I'm hearing an echo
except it's coming before me
Mahmoud how does perl divide statements if no ;s are used?
audreyt it's an ohce
luqui semicolons are used! They're just optional on the last statement of a group 15:06
?eval say "foo"; say "bar"; say "baz"
evalbot_8414 OUTPUT[foo bar baz ] bool::true
Mahmoud it is optional in some cases with perl five too
azuroth ?eval say "concat " ~ "strings";
evalbot_8414 OUTPUT[concat strings ] bool::true
azuroth ?eval say "concat " . "split"(""); 15:07
evalbot_8414 Error: unexpected "." expecting operator, ":", ",", term postfix, postfix conditional, postfix loop, postfix iteration, ";" or end of input
azuroth ?eval say "concat " . split("");
evalbot_8414 Error: unexpected "." expecting operator, ":", ",", term postfix, postfix conditional, postfix loop, postfix iteration, ";" or end of input
Mahmoud wtf?
audreyt ?eval "concat".split;
evalbot_8414 ("concat",)
luqui we changed string concat from . to ~
audreyt ?eval "concat".split('');
luqui so that we could have . for method call
evalbot_8414 ("c", "o", "n", "c", "a", "t")
Mahmoud heh why?
azuroth ?eval say "concat " .split("");
evalbot_8414 OUTPUT[concat ] bool::true 15:08
azuroth ?eval "concat " .split("");
Mahmoud so . is for only method call?
evalbot_8414 ("c", "o", "n", "c", "a", "t", " ")
azuroth ?eval "concat " ."split"("");
evalbot_8414 Error: unexpected "." expecting operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
luqui Mahmoud, yes
azuroth, what are you trying to do?
Mahmoud nice..
azuroth I dunno. can you do sort of method pointers? 15:09
?eval $a = &String::split; "blah".$a(""); # kind of like this maybe?
evalbot_8414 Error: unexpected "." expecting term postfix, operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
luqui yeah, but they're harder in perl 6 (you have to really shout at perl to say I'M USING A SYMBOLIC LOOKUP)
but I don't think it's implemented in pugs yet
azuroth ahh 15:10
audreyt mm? pugs does do symbolic lookup.
luqui ?eval my $a = "say"; "foo".::($a)()
evalbot_8414 Error: unexpected "." expecting term postfix, operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
luqui then is my syntax wrong
audreyt ?eval my $x = 1; $::('x')
evalbot_8414 \1
luqui audreyt, how do you do it for methods?
azuroth ?eval ("a", "b", "c").join(" ") 15:11
evalbot_8414 "a b c"
audreyt luqui: I'm sorry but, I don't think that is in the spec.
azuroth ?eval ("a", "b", "c")>>.~(" test")
evalbot_8414 Error: unexpected ">" expecting term postfix, operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
luqui audreyt, oh. let me look. I'm pretty sure that's the way it's done.
so the spec may need to be updated ;-)
audreyt yes, please do. 15:12
azuroth ?eval ("a b c", "b c d", "c d e")>>.split(" ")
evalbot_8414 Error: unexpected ">" expecting term postfix, operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
azuroth ?eval ("a b c", "b c d", "c d e") >>.split(" ")
evalbot_8414 Error: unexpected ">" expecting operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
azuroth hmm. is that not how you do hyper-method, or is it not implemented?
audreyt azuroth: yeah, unidirectional hyper is not yet parsed.
we currenly only do bidirectional ones.
luqui: have you seen my OpTable.hs?
azuroth aww. that's still cool :-D 15:13
audreyt ?eval (1, 2, 3) >>+<< 4
evalbot_8414 (5, 6, 7)
luqui audreyt, well, it was in A12, but there's nothing in the synopses
audreyt luqui: A12 is not normative so I pretend it does not exist :)
luqui thinks that's a good plan
audreyt please do a carry over
luqui audreyt, I did look at OpTable.hs
but not very much 15:14
I'm pretty much done with the basic combinators
audreyt k. I think mostly you just need the data structures
spent quite some time getting them concise enough
the parse algorithm itself can be implemented in every which way.
luqui sure
Larry says "OP parsing is not quite so simple as bubble sort, but not quite so subtle as quicksort" 15:15
and I don't think quicksort is subtle
so... should be fine
audreyt heh
it's definitely more complicated than quicksort
because we allow dynamic alternation of optable 15:16
luqui doesn't see how that makes anything more complicated
audreyt (which PGE::OpTable doesn't yet do, so I'd have to invent my own callback system)
luqui: some terms may be functions that alters the table itself
so when you parsed the first half of circumfix
it may introduce something that clashes with the closing half 15:17
luqui but that's okay, you'll get the clash when you get there
azuroth ?eval sub postfix:<!> (Num $a) { if $a == 1 return 1; return $a * ($a - 1)!;} say 10!;
evalbot_8414 Error: unexpected "i" expecting ";", statements or "}" conditional missing body or reserved word
luqui but I guess I won't really understand until I try to do it
azuroth, I don't think postfix works 15:18
audreyt $x[ BEGIN{our &prefix:<]> := ...} ]; 15:19
?eval sub postfix:<!> ($x) {[*]1..$x}; 10!
evalbot_8414 3628800
audreyt it does work.
it's the "if" syntax
luqui oh right
azuroth ohh yeah. :-p
luqui I just remember playing with postfix and being frustrated 15:20
azuroth ?eval sub postfix:<!> (Num $a) { if $a == 1 { return 1; } return $a * ($a - 1)!;} say 10!;
evalbot_8414 Error: unexpected "}" expecting term postfix, operator, ">>xx<<", ">>+&<<", ">>+<<<", ">>+><<", ">>~&<<", ">>~<<<", ">>~><<", ">>*<<", ">>/<<", ">>%<<", ">>x<<", "\187xx\171", "\187+&\171", "\187+<\171", "\187+>\171", "\187~&\171", "\187~<\171", "\187~>\171", "\187*\171", "\187/\171", "\187%\171", "\187x\171", "xx", "+&", "+<", "+>", "~&", "~<", "~>", "*", "/", "%", "x", ":", ",", postfix conditional, postfix loop, postfix iteration, ";" or "}"
luqui heh, I find myself writing this in Perl a lot:
if $foo {
then bar
oh crap
it's perlskell
azuroth ?eval sub postfix:<!> (Num $a) { if ($a == 1) { return 1; } return $a * ($a - 1)!;} say 10!;
evalbot_8414 Error: unexpected "}" expecting term postfix, operator, ">>xx<<", ">>+&<<", ">>+<<<", ">>+><<", ">>~&<<", ">>~<<<", ">>~><<", ">>*<<", ">>/<<", ">>%<<", ">>x<<", "\187xx\171", "\187+&\171", "\187+<\171", "\187+>\171", "\187~&\171", "\187~<\171", "\187~>\171", "\187*\171", "\187/\171", "\187%\171", "\187x\171", "xx", "+&", "+<", "+>", "~&", "~<", "~>", "*", "/", "%", "x", ":", ",", postfix conditional, postfix loop, postfix iteration, ";" or "}"
azuroth ?eval sub postfix:<!> (Num $a) { if ($a == 1) { return 1; }; return $a * ($a - 1)!;} say 10!;
evalbot_8414 Error: unexpected "}" expecting term postfix, operator, ">>xx<<", ">>+&<<", ">>+<<<", ">>+><<", ">>~&<<", ">>~<<<", ">>~><<", ">>*<<", ">>/<<", ">>%<<", ">>x<<", "\187xx\171", "\187+&\171", "\187+<\171", "\187+>\171", "\187~&\171", "\187~<\171", "\187~>\171", "\187*\171", "\187/\171", "\187%\171", "\187x\171", "xx", "+&", "+<", "+>", "~&", "~<", "~>", "*", "/", "%", "x", ":", ",", postfix conditional, postfix loop, postfix iteration, ";" or "}"
azuroth eek. time to give up
luqui azuroth, it may be a problem with using ! inside your definition of ! 15:21
I think that's a currently half-working feature
azuroth ahh
audreyt ?eval sub postfix:<!> (Num $a) { if ($a == 1) { return 1; } return $a * ($a - 1)!; }; 10!
evalbot_8414 Error: unexpected " " expecting term postfix, operator, ">>xx<<", ">>+&<<", ">>+<<<", ">>+><<", ">>~&<<", ">>~<<<", ">>~><<", ">>*<<", ">>/<<", ">>%<<", ">>x<<", "\187xx\171", "\187+&\171", "\187+<\171", "\187+>\171", "\187~&\171", "\187~<\171", "\187~>\171", "\187*\171", "\187/\171", "\187%\171", "\187x\171", "xx", "+&", "+<", "+>", "~&", "~<", "~>", "*", "/", "%", "x", ":", ",", postfix conditional, postfix loop, postfix iteration, ";" or "}"
audreyt weird
luqui (azuroth, though in perl 6 it is legal)
Mahmoud looks like larry wall created a new hash algorithm
luqui Mahmoud, ?
azuroth ?eval [~]("a ", "b ", "c") 15:22
avar ?eval <Ʀ ư>
evalbot_8414 "a b c"
("\\", "\\")
audreyt ?eval sub postfix:<!> (Num $a) { if ($a == 1) { return 1; } return $a * &?SUB($a - 1); }; 10!
evalbot_8414 3628800
Mahmoud i was away, and when I returned back to the irc window, i shocked with all the symbols on the screen
luqui haha
avar $ pugs -e 'say <Ʀ ư>.perl'
('ƃĀ¦', 'ƃĀ°')
luqui yeah, azuroth has been making perl barf all over the screen
azuroth ^_^;
Mahmoud heh
avar $ pugs -e 'say for <Ʀ ư>'
Ʀ
Ć°
why does this happen
Mahmoud i was googling for 'mod_perl vs mod_php' .. i hope i'm not trolling
luqui avar, it's probably something with the IRC interface 15:23
avar pugs normally prints out multibyte characters properly but the .perl method doesn't
luqui oh
avar luqui: that's on my local machine...
luqui yeah, I just figured that out ;-)
Mahmoud any benchmark you guys know about mod_php vr mod_perl?
avar ;))
luqui avar, good question.
luqui experiments
avar looks like I need to write another t/pugsbugs/file ;) 15:24
luqui avar, works for me
audreyt please do :)
luqui something with your config maybe
Mahmoud i'm afraid that i wasted my time with Perl :D
devorce time..
:P
luqui Mahmoud, huh? Why? Why? Don't leave!!
:-)
Mahmoud no sorry.. i'll leave..
avar luqui: what locale are you using? 15:25
luqui: I have en_US.utf8, maybe that's it..
luqui en_US.utf8
avar mm
Mahmoud rides his horse and leaves
luqui avar, oh, by the way
your example above, where you quoted $ pugs -e 'say for <...>' # I can't type those characters 15:26
looked completely fine to me
the output was precisely those two characters
Mahmoud luqui, don't worry.. i'll not leave.. i'm staying with you.. don't be afraid
avar luqui: it looks fine here too, what doesn't look fine is say <Ʀ ư>.perl
luqui oh
luqui looks
avar $ perl -Mutf8 -e 'print ord for split //, "say <Ʀ ư>.perl"; print "\n"' 15:27
115971213260230322406246112101114108
just to make sure we're giving it the same input;)
luqui ("\d[230]", "\d[240]")
yep, it's the same 15:28
that line is what I get
avar $ pugs -e 'say "Ʀ".perl; say "Ʀ"'
'ƃĀ¦'
Ʀ
simplified testcase
do those two come out the same for you? 15:29
luqui avar, try putting some input in a file
"\d[230]"
Ʀ
that's what I get
avar $ pugs say.p6
'ƃĀ¦'
Ʀ
$ cat say.p6
say "Ʀ".perl;
say "Ʀ";
same stuff
$ file say.p6
say.p6: UTF-8 Unicode text
luqui avar, what terminal are you using?
well, your term is probably fine 15:30
given that you can cat correctly
avar I'm sure my terminal is fine with regards to spitting out UTF-8, I work on UTF-8 all day;)
xterm with -en utf8 and LC_ALL=en_US.utf8
(the LC_ALL is important)
-en UTF-8 actually..
luqui avar, that's very strange then 15:31
avar, and this is the latest svn?
avar nah, 6.2.10
luqui oh, that could be it
avar I should compile the latest maybe;)
luqui I put in a patch a month ago or some that dealt with this sort of thing
s/some/so/ 15:32
avar I'll make a testcase for it regardless, just to make sure there aren't any regressions 15:33
luqui thanks
oh, avar 15:34
you probably can't commit
openfoundry is being a bastard
what kind of business isn't open on christmas eve! Geez!
avar christmas day;) 15:35
luqui christmas day? You're saying you're 16 hours away from me?
avar I'm saying that CD isn't the 25th everywhere in the world;)
luqui doesn't know much about timezones, but he didn't think that that was possible
oh
oh, I get it
avar It's the 24th here 15:36
luqui that's cool
where are you?
avar Iceland
luqui interesting
avar It's like that in .dk too
and probably .no,, .sv
although I'm not sure
mm 15:38
the perl6.vim file needs fixing
It doesn't autodetect perl6 for some reason
plus it defaults to that awful coding style *ducks*
luqui perl6.vim isn't the thing that autodetects 15:39
you need a filetype detector, and I'm pretty sure there isn't one in the pugs repo 15:40
avar mm
the std vim syntax files do that somehow
I think
luqui yeah
.vim/filetype.vim is where you put custom ones
vim has a global filetype.vim that it uses
Mahmoud going to cut my beard heh :P
1+ cm long 15:41
15:41 Mahmoud is now known as Mahmoud-Afk
pasteling "luqui" at 67.165.197.242 pasted "My .vim/filetype.vim" (19 lines, 655B) at sial.org/pbot/15110 15:42
avar heh, .wiki ;) 15:43
does that actually work for anything non-simple?
luqui does what actually work?
avar mediawiki syntax highlighting 15:44
that file works a charm, thx;)
luqui it seems to be pretty good
avar mm
avar MW dev
our syntax sucks;/
luqui it doesn't make any sense if you dont have those filetypes in your ftplugin
of course
by which i mean in your syntax/ 15:45
avar hah
It's kind of like php
it wasnt designed, it kind of just grew like cancer
luqui you guys should abruptly change to kwid
15:45 Mahmoud-Afk is now known as Mahmoud
avar what's that? 15:45
luqui kwid is a wiki syntax designed by ingy 15:46
luqui likes it a lot
Mahmoud heh.. php is cancer LOL
avar well, actually the main problem with it is not the syntax itself, but the parser we're using
luqui avar, really?
avar it doesn't use a parse tree because php would be too slow for that, it's just a regexp hack
luqui yuck
like Pod::Parser
yeah, I always thought that MW was a fairly nice piece of software 15:47
I wanted to submit a patch, and found that it was only nice on the outside
with a heart black as coal
avar It's very different, 15:48
that parts I write are nice & clean;)
15:48 Mahmoud is now known as Mahmoud-Afk
luqui of course. all those other parts are hard to understand ;-) 15:48
avar since I'm pretty fanatical about cleanness...;)
I guess it suffers from pretty much the same problem as most php apps 15:49
1. written as a hack
2. still a hack
luqui is being a good boy at the moment and documenting his code
avar 3. hard to extend hack
luqui php makes it hard to exercise good software engineering practices
whereas perl makes it easy to do so... (but perl also makes it easy to write complete crap) 15:50
avar php5 is slightly better
luqui (which java claims it makes hard to do)
avar you can't make private members for classes in php4 for example, which is .. insane
luqui (but I claim that it's impossible for any language to do that)
(except Haskell. it's very hard to write crap in haskell, because it's hard to write anything in haskell) 15:51
avar hahah
luqui (if you don't really think through the problem, Haskell's vagueness-killer will get you)
(end of long parenthetical)
(really)
avar Working on MW is kind of fun because you get to please the crowd, few FS projects are like that, I guess;) 15:52
luqui what is that acronym FS?
avar free software
luqui oh, of course
well, you get to do that to a certain crowd in Perl 15:53
except that is a crowd that seems not capable of being pleased
avar ;)
luqui avar, spent any time on p6l?
avar perl6-language? 15:54
luqui yeah
avar I subscribe to it since ~1wk ago, haven't posted anything
luqui ahh
so you missed the "self", um, "debate"
that's good 15:55
avar self.* ?
luqui coming up with *optional* syntax to call methods on self
even the people who said they would never ever use it fought against it
and then when we said "okay, there's no syntax" 15:56
everybody said "no, there needs to be a syntax"
which is much more like "it needs to be *my* syntax"
avar you mean something like $this-> and this. in php, ruby & C# ?
luqui yeah
the one I liked best was ./method
oh, and in C# at least, can you just say method() as shorthand for self.method() ? 15:57
luqui hasn't done much C# 15:58
Khisanth not really like $this-> and this. ... since that would just be $self.
or whatever depending on you prototype and pragmas? :)
luqui use Perl6::Attributes; ./method(...) :-) 15:59
avar I like self.method
Khisanth that is irrelevant for that particular thread :P
luqui yeah 16:00
so Larry made a decision that pleased nobody at all
and then later he quietly changed it to something else
without telling anyone about it
"if you have an objection to it, you'll have to read the spec to find out what you're objecting to first!"
Khisanth our $Larry is sneaky; 16:01
luqui and all the whiners are too lazy to read the spec
Khisanth perl6 is just a plot to make writing poetry easier!
luqui doesn't get it 16:02
luqui knows about perl poetry
luqui still doesn't get it
Khisanth I mean it adds a lot more usable vocabulary
luqui oh, you were referring to "our $Larry is sneaky;". Now I get it. 16:04
is, but, does, when, ...
Khisanth lazy :) 16:06
audreyt whew. 4 journal entries today. 16:17
I think I deserve sleep :)
luqui should read those 16:18
luqui thinks it's funny that you can write pages of code and still need more
but write 4 journal entries and be wasted
luqui works the same way, though 16:19
audreyt I think English is a really sloppy programming language.
but maybe that's a virtue :) 16:20
gaal either that, or humans are a hell of a complex platform.
luqui Well, you think that, until you meet an expert English programmer.
audreyt it's probably both :)
avar audreyt: mm, url yo journal?;)
*to
audreyt avar: see topic
(pugs.blogs.com)
ingy: see you in RL tomorrow :) 16:21
audreyt waves &
gaal bye audreyt :)
avar I wonder if there are any rss viewser (preferrably for firefox) that don't suckl 16:26
I'm using gmail's ATM 16:27
luqui openfoundry is back 16:31
clkao zzz 16:32
svnbot6 r8415 | luqui++ | Fixed interpretation of curlies to make a hash. The following:
r8415 | luqui++ | { %foo }
r8415 | luqui++ | { %$foo }
r8415 | luqui++ | { %foo, a => 1, b => 2 }
r8415 | luqui++ | are now interpreted correctly.
r8416 | luqui++ | Backtracking controls.
r8417 | luqui++ | Change "cut"'s name to "commit", even though what I'm doing is a little more general than commit.
r8418 | luqui++ | A little documentation.
rafl_ Who wrote the smoke/yaml_harness stuff? 16:33
luqui dunnos
clkao iblech? 16:34
rafl_ He only wrote the smokeserv, afaik. 16:38
audreyt yaml harness is nothingmuch and gaal's doing I believe 16:40
svnbot6 r8419 | luqui++ | A little bit more documentation, for the slightly tricky quantify combinator. 16:44
gaal rafl_: yup, I wrote the initial version based on Straps and nothingmuch did most of the refactoring into the TAP modules. 16:46
svnbot6 r8420 | luqui++ | Minimal quantifiers. 16:53
r8421 | luqui++ | Making sure that quantifiers are maximal by default.
Khisanth rafl_: couldn't you ask the svn server? :) 17:08
slavik so, when is perl6 supposed to be released? 17:10
Khisanth EOT 17:14
integral Christmas! :-P 17:22
Khisanth tomorrow?! 17:29
slavik sweet :P 18:12
avar my 1st post to perl6-language;) 18:54
luqui except you're wrong :-) 18:58
no, not really. but you could have been a little more right 18:59
@Larry (or, more specifically, @Larry - lwall - luqui) think that there should be an "official" implementation 19:00
which is written in PIR (ugh) and runs on Parrot
and all other implementations try as hard as they can to conform to that one 19:01
but I'm ashamed of that position, so I didn't want to reply...
avar @Larry? 19:02
luqui design team
avar ah
luqui clever pun...
avar explain;)
luqui the array of Larrys
anyway...
oh, cool name, by the way 19:03
er, cool looking. I have no idea what it sounds like.
avar ;)
I have an ogg vorbis of the pronounciation somewhere...
so what *is* the status with regards to specs? 19:06
svnbot6 r8422 | luqui++ | Optional, and some untested anchors.
luqui well, the reason there are specs is because larry didn't want to program
avar hah
luqui (he did this for the sake of the code... or so he says)
Larry creates little languages wherever he goes, so it's often hard for people to grok his code without immersing themselves in it 19:07
there are a couple more reasons that it's good that larry doesn't program
for instance, when you're writing the language, it's easy to specify the behavior that's easiest to implement 19:08
rather than easiest to use
avar Of course there are downsides to that as well, you're likelier to specify things that might conflict if you're not implemnting it too 19:09
luqui that's actually not true... 19:10
we get plenty of conflicts in pugs
and nobody realizes them for a long time
until somebody says "hey, this doesn't seem like what it's supposed to do"
There is one downside that I see 19:11
and it's the same as the upside
I have a strong belief that what's easiest to implement is often the right thing to do
(as long as there is a solid design)
so you don't get some of those wacky over-dwimmy situations that we've seen throughout the design process 19:12
(all of which eventually get boiled down into something consistent and predictable, thankfully)
avar an assembly language would be easiest to implement;) 19:40
19:41 DesreveR is now known as r0nny
SamB the trouble with DWIM is that it is overly hand-wavey; sometimes you still haven't figured out what you mean yourself, how can you expect the computer to do it? 19:58
buu SamB: If you haven't figured out what you want, why are you typing code? 20:04
Mahmoud-Afk hi buu :) 20:23
20:23 Mahmoud-Afk is now known as Mahmoud
Mahmoud IRC Rule #586: Bosses never replay 'Hi's. :/ 20:24
reply* 20:25
buu ZOMG.
Mahmoud looks like calture difference.. here it's very rude to not reply 'hi','salam' or any welcome message. 20:26
here means UAE
tewk ls 20:27
obra Mahmoud: around here, it's not particularly rude. With the number of folks coming and going, that would be..a lot of traffic 20:28
But welcome :)
Mahmoud I see..
reasonable :)
practical people ;)
obra That is the perl thing, yes ;) 20:29
azuroth good morning 20:37
dduncan mawnin 20:57
foundry's up, so will push momentarily 20:58
azuroth hey hey
dduncan and the same person who was here just before I left, and had the f trouble
done; now the p6 and p5 repos are in sync 21:00
svnbot6 r8423 | Darren_Duncan++ | r1562@Darren-Duncans-Computer: darrenduncan | 2005-12-24 01:04:36 -0800
r8423 | Darren_Duncan++ | /ext/Rosetta-Incubator : major update to lib/SQL/Routine.pm; for the Document and Node classes both, added full attr decls in code and docs, added full BUILD submethod, and export_as_hash() method, to code and docs
dduncan ?eval my %h = { 4 => 5 }; my $bar = { %foo }; $bar; 21:42
21:43 evalbot_8414 is now known as evalbot_8423
evalbot_8423 Error: Undeclared variable: "%foo" 21:43
dduncan ?eval my %foo = { 4 => 5 }; my $bar = { %foo }; $bar;
evalbot_8423 {("4" => 5),}
Juerd dduncan: Don't assign a *reference* to a hash. 22:02
luqui that actually seems to be a bug
his code shouldn't work, but it did.
dduncan oh
Juerd { } returns a reference. Hash assignment works with lists. 22:03
dduncan I have had issues with related matters for awhile
luqui ?eval my %a = { a => 1, b => 2 }; say %a<b>
evalbot_8423 OUTPUT[2 ] bool::true
luqui yep, wrong
Juerd ?eval my @foo = [ 1 ];
evalbot_8423 [[1,],]
luqui I fixed it for lists, though
Juerd That works better.
luqui: Arrays :)
dduncan so is %hash = $hashref wrong then?
luqui it didn't before yesterday
Juerd, :-) stupid words. why couldn't they be more like numbers? 22:04
dduncan there was a checkin about this yesterday
Juerd dduncan: Not if %hash is declared to have non-string keys.
%hash = $hashref just means %hash = ($hashref => undef)
luqui with a warning
dduncan so what is the proper way to copy a hash
Juerd Yeah, odd number blah blah :)
dduncan: my %hash2 = %hash1.copy; 22:05
luqui dduncan, you copied the hash correctly
Juerd Alterantively, just:
my %hash2 = %hash1;
dduncan so that works then
Juerd Sure.
dduncan and %!hash2 = %hash1 works
Juerd %hash1 is in list contex
dduncan and $foo!hash2 = %hash1 works?
Juerd What's %foo?
eh
What's %!foo? 22:06
luqui private member
Juerd Oh my, when did that happen?
luqui dduncan, the latter is doing something quite different
because %hash1 in that case is in *scalar* context
Juerd Does %foo in list context eval to .kv or to .pairs?
luqui .pairs
Juerd Great :)
dduncan my last 2 lines were both referring to object attributes that were hashes
Juerd No more while-each. I hate while-each. 22:07
dduncan but one specified the object name
luqui dduncan, it's all about the sigil
if you write it with a %, then it's a hash
otherwise it's a hash reference
Juerd I tend to write short subs, and to return as early as possible. while-each sucks so much that I'm accustomed to for (keys %foo) { $foo{$_} )
dduncan so if $obj is an object ...
luqui Juerd, me too
Juerd Which is ugly, but at least works.
dduncan what is the syntax to set its %!hash attr? 22:08
Juerd I don't get why people love iterators. They SUCK :)
Lists rock, iterators suck.
luqui $obj!hash = {...} # note the curlies
Juerd And we don't need iterators now that we have lazy lists. Hurrah!
luqui that is, you assign it a reference
dduncan okay
luqui doesn't mind iterators 22:09
dduncan so with %!foo I assign as a %hash, and with $obj!foo I assign as a $hashref?
luqui hates implicit iterators
yep
dduncan okay ...
Juerd luqui: Okay, that works for me :)
dduncan I have committed code that isn't sure how to do it ... will fix
one more thing ...
Juerd I've abused scalar glob yesterday, though ;)
dduncan the syntax to generate one hash ref that is a copy of another hash ref?
is it {*%{$hashref}} ? 22:10
luqui Juerd, isn't the fact that you're using it in the first place a good indicator that you're abusing it :-)
dduncan, unfortunately, that doesn't work
Juerd luqui: If there's no ? or * in there, yea :)
luqui { %$hashref } will work fine
dduncan and { %{$hashref} } too I assume? 22:11
luqui also in perl 6 (but not necessarily in pugs), $hashref.copy
Juerd glob "{a,b,c,...,z}{0,1,2,...,9}{0...9}{0...9}", but in full, can be SO useful :)
luqui ah, that glob
dduncan called away ... back in 15m or so
Juerd my @ids = lazy_ids(); shift @ids; 22:12
Lalala :)
luqui yes, but there really needs to be a more general mechanism
list comprehension
Juerd Agreed
Though syntax is hard.
dduncan I was using the '*' there to mean splatty, or something
luqui yeah, that's the biggest problem 22:13
everything's already taken
Juerd For more advanced stuff like list comprehension, I'd think WORDS as opposed to SYMBOLS, would help a lot.
luqui dduncan, yeah, but if there's anything besides a hash or a pair in curlies, it's taken to be a coderef, not a hash
dduncan say if you want to populate a sub's arg list from a %hash
luqui that's the problem with the *
Juerd, you might be onto something there
Juerd Because symbols are taken, but also because much of the Perl target audience (sysadmins, one time scripters) doesn't regularly use these techniques.
luqui "where", the best word, is already taken though
dduncan back in a few mins
Juerd luqui: "where" isn't clear anyway 22:14
luqui true
any suggestions?
Juerd It may read naturally, but it communicates very little semantic information
luqui: This is a good moment to admit I don't really know what list comprehension consists of :)
luqui it's just a notation that mixes multiple maps and greps together 22:15
for example:
(in haskell, but I'll keep it simple)
[ x + y | x <- [1..10], y <- [1..10], x > y ]
[3,4,5,5,6,7,6,7,8,... 22:16
(had to go to ghci to figure out what that would actually output)
Juerd Oh, right, audrey showed me in the very first days of pugs. Now I remember. 22:17
luqui map { f($_) } @list <==> [ f(x) | x <- @list ] # Haskerl 22:18
Juerd Well, remember... Not really, but I do recognise the example.
Eh, s/example/syntax/
haskerl; heh
luqui grep { f($_) } @list <==> [ x | x <- @list, f(x) ]
and it's all just generalizations from there
[ expression | x <- sourcelist1, y <- sourcelist2, ..., condition1, condition2, ... ] 22:19
Juerd Right
And why not using foreach?
luqui you mean: for @sourcelist1 -> $x { for @sourcelist2 -> $y { push @result, expression($x,$y) if condition1 && condition2 } } 22:20
that's why
Juerd my @res = for @foo -> $foo, @bar -> $bar { next unless condition; ...; f($foo + $bar); }
What exactly is why? Notation? 22:21
luqui we could let for return the list of its results instead of the last thing evaluated
and then that might work
Juerd Otherwise, simply:
luqui the main reason for mine was that it's imperative
but for can't do that yet
Juerd my @res = gather { for { for { take ... if ... } } };
Larry doesn't want if and for to return things. I want them to. :( 22:22
??:: should simply desugar to if/else
luqui Juerd, you can
Juerd But so far, it's just in my head.
luqui ?eval my $x = do if 1 { 4 } else { 5 }; $x
evalbot_8423 \4
Juerd I see.
I don't think I understand "do" there.
luqui you can "do" that with any statement-level construct
it just turns statementy things into expressiony things 22:23
Juerd Why is the word needed?
We have context!
dduncan question, if the sigl's the thing ...
luqui because there's a syntactic difference
Juerd void context: statement, other context: expression.
What difference?
dduncan say we have @list_of_obj ...
luqui first, semicolons
Juerd Semicolons were fixed with the whitespace hack
luqui only sortof
dduncan how do we set a hash attribute of one array element that is an object?
luqui anyway, Larry thinks it's important to have a difference between statements and expressions
Juerd I know he does :( 22:24
dduncan eg, @list_of_obj[0].hash = ?
or also
Juerd dduncan: Have you ever used Perl 5?
dduncan yes
but there we have $ all the time
Juerd Because in the current state of Perl 6, it may not be wise to use it to learn Perl.
luqui Juerd, braces. if if $foo { } else { } { } else { } 22:25
dduncan Juerd, ...
luqui or rather
if if $foo { } else { }
dduncan I am very experienced with perl 5, but for some strange-seeming aspects
Juerd dduncan: % is for hashes, @ is for arrays, $ is for scalars (including references)
dduncan and I'm also trying to write correct code in perl 6
Juerd dduncan: This does not change.
dduncan: Even if the hash is used for a slice or a single element, it's still %.
luqui dduncan, okay, it's not all about the sigil there...
Juerd However, if you have a reference, even if it points to the same hash, it's $. 22:26
dduncan so I'm learning perl 6
luqui it's more like, it's all about the variable
or something
um... @foo and %foo expand themselves in list context
that's the only thing that distinguishes them from scalars
Juerd luqui: But the braces thing is the same thing that happens with any bracketed construct: if it becomes complex, more parens are needed. We have precedence for paren-thingies, but more importantly: warnings. Nesting curly-thingies could simply be made impossible without parens. 22:27
dduncan back in a bit
luqui "it's all about the sigil" was an overgeneralization
Juerd if (if $foo { } else { }) { ... } else { ... }
luqui no, the second one
Juerd With "if if" being a parse error
luqui why is if if a parse error?
Juerd, do you know about parsers?
Juerd Because it uses to statement: things without encountering parens :) 22:28
luqui so when can you use a statementy thing?
Juerd statement:* would then be allowed only on statement level, and within parens.
luqui so only after parens?
then you can't say: my $x = if 1 { }
you'd have to say: my $x = (if 1 { })
which is kinda what we did with do, except do looks nicer 22:29
Juerd I don't know much about parsers. I do believe that parsing should be the last, not first, thing to consider when inventing syntax.
luqui yes, but thinking about how the parser does it is kinda like Haskell in that it defoggies your thinking
Juerd I see your point re my parens suggestion.
But I don't get why do-if would be possible while if would not. 22:30
luqui Chomsky believed that human languages were context-free. So if a parser can't do it, then humans can't easily do it. But Chomsky was full ot it anyway. ;-)
Juerd The first thing I'd be tempted to do is write an error handler automatically inserting "do" whenever parsing an "if" failse...
s/failse/fails/ 22:31
luqui well...
then you wouldn't get missing semicolon errors
and we experienced programmers never get those anyway
but noobs get them all the time
Juerd If you say so :)
luqui do you understand why you wouldn't get missing semicolon errors?
Juerd Javascript fixes the missing-semicolon thing the same way I'd fix the do-if thing: insert them automatically whenever parsing differently didn't work :) 22:32
luqui sub foo { } my $x = foo \n if condition {...} else {...}
Juerd (Well, almost)
luqui except my example is a genuine error, and it parses fine
foo will be getting a weird argument, though 22:33
Juerd But without the else { }, that's already not a trappable error.
luqui Juerd, how so?
Juerd my $x = foo if condition({ ... })
luqui but that's been specified 22:34
when you see a block in operator position, it's the statement-level block
22:34 Mahmoud is now known as Mahmoud-Afk
Juerd But back to list comprehension. I've read some more, and now understand it's not magic, but actually quite simple :) 22:34
luqui yep
Juerd One thing I don't understand yet. Are the input lists walked in parallel? 22:35
Or do you get @x*@y iterations in total?
luqui no. they're done glob-style
yes @x*@y
Juerd ok
luqui in Haskell, they're last-first
(which makes the most sense to me, but could be argued either way) 22:36
Juerd comprehend @foo; @bar -> $foo, $bar { next unless $cond1 and $cond2 and $cond3; f($foo + $bar); }
Which can be written in a minute or two? 22:37
Or am I not understanding it correctly yet?
luqui Juerd, that's what I was thinking
Juerd That I'm not understanding it yet?
luqui no, the comprehend syntax
the thing I like about list comprehension syntax from haskell and python is that it reads just like "set-comprehension" in maths 22:38
in fact, it *is* that syntax :-)
Juerd I think the easy solution is to make for (be able to) return lists, even if it's only with do.
You can always introduce macros for the nice syntax later. 22:39
luqui basically making for a statement-level map with reversed arguments
luqui could live with that
Juerd What would we still need map for?
Isn't map just list-returning for? :) 22:40
luqui reversed arguments. reads better sometimes. also, statement-level.
Juerd And can't one already do for { ... } <== @foo, if they really want arguments to be reversed?
luqui but statement-level isn't going to sell *you* :-)
Juerd No, statement level does nothing to me. Everything's an expression, and I'm happy. :)
sub foo { ... } should just return \&foo. 22:41
luqui Juerd, you should take up haskell
let x = if 1 then 4 else 5
Juerd I know.
If only I had a big bag of tuits
luqui yeah. It took me until we were driving out to Portland to get tuits
avar x = 1 ? 4 : 5;
;)
luqui yeah, but Juerd is objecting to the fact that there are things that aren't expressions 22:42
oh, avar, and you mean $x = 1 ?? 4 !! 5;
Juerd I've began programming again, partly ignoring pain (CTS-like, but not CTS), because I'm a little more than out of cash.
avar luqui: that was C
luqui uh huh
Juerd avar: I thought it was Perl.
luqui but this is #perl6
luqui notices that that's perfectly good perl
Juerd { my $x; sub x : lvalue { $x } } x = 1 ? 4 : 5; 22:43
Heck, who needs sigils anyway.
avar are subroutines expressions or declerations in p6?
luqui I believe there was a module that gets rid of sigils precisely by doing that
avar, if they have a name, then they are declarations
if not, then they are expressions 22:44
avar mm
Juerd luqui: Getting rid of sigils (optionally!!) is one of the ideas in Perlego.
avar too bad,
luqui avar, por que?
Juerd Perlego is my reference Perl 6
avar I kind of like def x(y) puts y; end if 2 in ruby;)
luqui avar, you mean you like how it looks, or you like the ability to do that?
Juerd my foo = 3; # from now on, we know what "foo" is, but we can use $foo too, if we want. 22:45
avar how it looks, I know I can do it just by wrapping stuff in another block or something..
luqui (to me, that looks like the "end" is under control of the "if", but nothing else)
Juerd It solves 50% of the line noise issue :)
luqui Juerd, should be pretty easy with a grammar munge
Juerd luqui: I thought so :)
luqui Juerd, might even be pretty easy with a source filter
you should write it in perl 5 22:46
Juerd But I keep striving to get things like that into standard Perl.
luqui yeah, but remember who I am...
I'm that guy who hates when people talk about standard perl...
Juerd I think that configurability is GREAT, but that the default should be sane.
And that syntax is one of the few things that needs specal sanity. 22:47
special
luqui too. he personally doesn't think that my foo = 3; is all that sane
avar mm, no sigil?
Juerd Well, that's okay. You're far from alone in thinking that, which just means that my idea wasn't a good one in the first place. 22:48
Which in turn means that I can save a lot of time writing a module that does it too.
luqui :-)
Juerd There are only two things left in my wishlists that I find really important. For the other things, I've been convinced that my ideas aren't sound, or they are already standard Perl 6. 22:49
avar what does my x = 5; do ?;)
Juerd The two things left are <-> and parenless method calls.
luqui avar, nothing.
Juerd, wants it to do the obvious thing.
s/,//
Juerd avar: Nothing, but in the hypothetical Perlego, it'd declare $x and a sigilless alias for it, called x.
luqui Juerd, you do know that we have listop method call 22:50
avar eh, what's the point?
luqui ?eval "foo bar".substr: 1, 3
evalbot_8423 \"oo "
Juerd luqui: That's nice, but substr as a listop?!
luqui thinks he agrees with Juerd about <->
avar what would <-> do? 22:51
luqui The : turns any method call into a listop
Juerd I see.
I can very well live with that I think :)
Although.
luqui avar, <-> is like ->, but it declares its arguments rw
Juerd Does it imply list context too? That would be unfun :(
luqui Juerd, well, I think "turns into a listop" they meant "parses like a listop" 22:52
Context is callee-side now anyway
so it definitely doens't imply list context
Juerd Oh, that's interesting. How does that work?
luqui Juerd, it kinda doesn't
that's one of the big gaping holes in the design
Juerd An @foo is flattened on the callee side?
(when in slurp context) 22:53
luqui if foo is a multi: foo(bar());
we can't know what context to call bar() in until we know which foo() to call
avar luqui: isn't there an existing syn for that?
luqui but we can't know that until we call bar()
Juerd Hehe
luqui because it's a multi
Juerd Hmmm
luqui avar, existing syn for <-> ?
yes: -> $x is rw, $y is rw { } 22:54
Juerd I've always assumed the context would be a disjunction
avar like for %hash.kv -> $k is rw, $v is rw { or somthing...
luqui yeah, that's legal
it's just not as nice as <->
which is very mnemonic
avar I like is rw better...;)
Juerd multi foo (Foo $x) { ... }; multi foo (Bar $x) { ... }; foo(bar()); # bar is in Foo|Bar context
And the foo is chosen according to what bar() actually returned.
luqui Juerd, that's one way to do it 22:55
avar or were you planning to munge "<->" into subroutine argument lists ?
or would they be a special case?
luqui Juerd, but if you're in Scalar|List context, what do you return?
avar, -> declares a sub as it is
?eval (-> $x { $x + 1 }).(10)
evalbot_8423 11
Juerd luqui: That's a context that doesn't exist in my head, for two reasons: 1. there are no List objects; 2. Main context type (item versus list) shouldn't be mixable across multis. 22:56
luqui Juerd, the first one is just semantics. The second one probably ends up being right
Juerd List objects just don't work.
avar luqui: ew;)
luqui you don't use it unless it looks nice
that's how a lot of perl programming goes 22:57
you can abuse the syntax, but you can also use it to make things very readable
Juerd And I still think we should rename "list context"
luqui Juerd, it is renamed
avar I meant, if you want to replace for %hash.kv -> $k is rw, $v is rw { with for %hash.kv <-> $k, $v { what do you want to replace sub x(Num $x is rw) with?
luqui it's "Slurpy" context
avar nothing?
luqui yes, nothing
Juerd luqui: That's Pugs, not Perl 6.
avar or sub x(Num <-> $x or something
luqui Juerd, I believe it's perl 6, too
Juerd luqui: A conscious stray from Larry's path. 22:58
luqui avar, certainly not!
Juerd No, Larry recently confirmed it's item versus list.
luqui Juerd, really?
Juerd Yes.
luqui messageid?
Juerd Hold on
avar luqui: I don't like it, two totally different ways to declare something as rw?
Juerd Mailbox is growing too large
...
luqui avar, it's just sugar
dduncan okay, I'm back ... will fix and checkin stuff now ... 22:59
luqui look at it for for:
for @a <-> $x { $x++ }
@a goes into $x, then $x goes back into @a
leo is pondering if perl6 {is,will be} released today;) 23:00
Juerd That Larry doesn't To: people makes finding answers to your own message hard.
Normally I'd ~f larry ~t juerd ~b context
(I've stopped To:'ing people myself too, though)
luqui leo, yes
obra leo: no ;)
luqui hehe
avar luqui: could I do for @a ā‡” $x { ++$x } ?;) 23:01
Juerd luqui: Message-ID: [email@hidden.address]
luqui: And thread
dduncan I never To: people myself if they're on the mailing list
when people do that to me, I get 2 copies
Juerd dduncan: Which really is better. 23:02
dduncan: But it does make finding replies harder.
dduncan I just look at all the replies to threads I participate in
luqui avar, you might have to declare that yourself
dduncan particularly ones I start
Juerd As searching has no threading functionality.
dduncan: Yeah, but that isn't quite a viable alternative to searching archives
luqui unless that symbol is latin-1, in which case we might (but probably won't) make it a synonym 23:03
Juerd dduncan: Mind you, of the over 7500 p6l messages in this mailbox, I wrote over 600
dduncan so I always To my list responses to the list itself and only that ... most of the time ...
I only add a To for a specific person in rare chances I want it to stand out to them / have them see it sooner 23:04
luqui Juerd, but brent has a good point
not a great point
dduncan, I wonder if that works 23:06
Juerd luqui: It's a good point indeed. 23:07
dduncan with some mailers, they format a message list so that direct mails stand out more than indirect mails
Juerd luqui: But it's a definition without argumentation. *why* nouns? I agree that if any one is a noun, they should all be.
But they can also all by adjective.
avar ?eval "Ʀ" 23:08
evalbot_8423 *** unexpected "(" expecting term postfix, operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input at -e line 15, column 6
Juerd In fact, since the main contexts both (excluding void) have the same tree of types underneath, adjective may make sense.
avar ?eval say "Ʀ"
evalbot_8423 *** unexpected "(" expecting term postfix, operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input at -e line 15, column 6
avar ...
?eval 1+1
evalbot_8423 2
Juerd "singular Dog context", "plural Dog context"
avar ?eval "\303\246"
evalbot_8423 "\d[303]\d[246]"
avar odd
?eval "\303\246".perl 23:09
evalbot_8423 "\"\\d[303]\\d[246]\""
Juerd avar: Please try such things with pugs directly. Use the evalbot if you want to show the channel something specific.
avar Trying to figure out if there is something to show;) 23:10
Juerd I just had a toilet revelation 23:19
item context should be "one" context.
With a low precedence "one" operator, and a high precedence "1" operator.
"1", the number, differs from "1", the operator, in that it has 0 arguments.
And 1foo is unambiguous because it's an invalid identifier anyway 23:20
Ignore all that :)
Toilet revelations aren't the greatest anyway
luqui: A bare closure should just start iterating if it gets any input lists. 23:21
luqui: { @foo.push: $x + $y if condition and condition } <== @x, @y; 23:23
Though where exactly $x and $y come from, I don't know ;)
nothingmuch $^x and $^y ?
Juerd Oh, of course, they're $^y and... what nm said :)
Hm 23:24
nothingmuch this concludes mmy helping pugs today ;-)
Juerd comprehend { $^x + $^y } <== @x, @y, { condition }, { condition };
nothingmuch now i'm off to sleep.... i'm frozen stiff from a walk in the rain
Juerd nothingmuch: Awww
nothingmuch it was voluntary
i *like* the rain
we get it very seldom 23:25
Juerd nothingmuch: Still, I wish to you a warm evening
nothingmuch thank you =)
Juerd I absolutely hate rain.
We have way too much of it.
nothingmuch it's always like that =)
Juerd It's unfortunate that weather can't be exported easily.
Other natural resources, like gas and oil, can.
nothingmuch if we had a lot of rain i wouldn't have liked it either ;-) 23:26
dduncan ?eval sub bar () { my %foo = (4=>5); return { %foo }; } say bar();
evalbot_8423 OUTPUT[45 ] bool::true
Juerd No, but a healthy world market would balance things out a bit :)
nothingmuch hehe
then the US would try to attack seattle
Juerd Haha
dduncan ?eval sub bar () { my %foo = (4=>5); return { %foo }; } say bar().ref;
evalbot_8423 OUTPUT[Hash ] bool::true
Juerd dduncan: Do you have a local pugs installation? 23:27
dduncan: Or access to feather?
dduncan yes
but in my situation, its treating the sub's return value as a block
Juerd dduncan: Please use that. Or, if you were trying to make a point, don't forget the point itself.
luqui: How about: comprehend { $^x + $^y } <== @x, @y, { condition }, { condition }; 23:28
Ehhh 23:29
luqui: How about: comprehend { $^x + $^y } <== @x <== @y <== { condition }, { condition };
luqui: It'd be nice if there was a way to carry lexicals out of a block. 23:30
luqui: Because $^apple doesn't sort well with $^pear, if @pear was the first input list 23:31
luqui: Or maybe this all just means we need postfix ->, then <-, as well. 23:33
luqui: comprehend { $x + $y if condition } <- (@x; $y); 23:34
luqui: Which for all I care, can just replace map. map { $x + $y if $x > 10 } $x, $y <- (@x; @y); 23:35
luqui: Though I'd write it args-first: map (@x; @y) -> $x, $y { $x + $y if $x > 10 }
I think that's nice enough syntax to work with :)
Hm, no, map of course walks arrays in parallel 23:41
comprehend then.
It's a shame -> can't be placed in the middle of something. 23:42
@foo -> $foo, @bar -> $bar feels more logical, syntax wise, than @foo; @bar -> $foo, $bar
dduncan okay, here's a demo of the problem ... 23:43
?eval class M { has %!h; submethod BUILD () { %!h = (4=>5); } method geth () { return { %!h }; } } class T { method tra ($m) { .useh( $m.geth() ) } method useh (%h) { say "got {%h}"; } }; my $gm = M.new(); my $gt = T.new(); $gt.tra( $gm );
evalbot_8423 OUTPUT[got <SubBlock(<anon>)> ] bool::true
dduncan that's distilled from the actual failing code
what I demo'd earlier, a simpler case of the same, does work
so I'm trying to figure out what to fix in the newest example so that the copy of %!h gets to useh() intact 23:44
Juerd Hmmm
dduncan but somehow it becomes an anon block
Juerd What if -> (and <-> of course) would just declare parameters for "the next bare block" 23:45
And without statement control keyword, it'd just pre-alias 23:46
avar $ ghc -v
Glasgow Haskell Compiler, Version 6.4, for Haskell 98, compiled by GHC version 6.2.2
\o/
no more 5.* ;)
Juerd $foo + $bar -> $x; $xyzzy * $quux -> $y; { $x + $y }; 23:47
There's VERY little point in that, though. 23:48

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!