»ö« | perl6.org/ | nopaste: paste.lisp.org/new/perl6 | evalbot usage: 'perl6: say 3;' or rakudo:, alpha:, pugs:, std:, or /msg p6eval perl6: ... | irclog: irc.pugscode.org/ | UTF-8 is our friend!
Set by lichtkind on 5 March 2010.
00:01 Schwern left 00:04 xinming joined 00:06 justatheory left 00:07 Trashlord left 00:09 Trashlord joined 00:17 sahadev left 00:20 alester joined 00:23 lestrrat is now known as lest_away 00:27 Chillance left
sorear pmichaud: I'm confused about the function and relationship between Perl6Scalar and ObjectRef 00:33
I can see why we need one of them, to implement the distinction between binding and assignment. But both? 00:34
pmichaud at one point we needed "soft references" versus "hard references" 00:35
we may still need that, which is why it hasn't been completely taken out
00:36 masonkramer joined
jnthn pmichaud: iirc, we're currently only using one of them? 00:36
pmichaud but you're correct that the function is not a purely bright line in rakudo at the moment
jnthn: we have instances of both in the codebase, iirc
jnthn OK
pmichaud I was going to see about cleaning it up as part of fixing the binding semantics
jnthn OK, cool.
masonkramer I read somewhere that in perl 6, it will be possible to ask for machine ints instead of sv's...is that still in the cards? and if so, could anyone point me towards more information about it
pmichaud masonkramer: my int $x; 00:37
jnthn masonkramer: It's spec'd in S09
pmichaud (note lowercase 'i')
jnthn masonkramer: Rakudo doesn't implement it yet, sadly.
pmichaud CNYI in Rakudo at the moment.
jnthn On the cards post-Rakudo *.
00:38 yinyin joined
sorear note, however, that it won't be helping performance for a while 00:39
even once it's implemented
because MMD optimization is a completely orthogonal problem
masonkramer Hmm...not sure I follow you
sorear has int $.field is good for memory usage regardless
masonkramer I thought one of the problems is that double indirection requires double dereferencing
sorear when we implement native types, $int1 + $int2 will work like this:
1. Wrap $int1 and $int2 into freshly constructed Int object 00:40
2. Fetch the multisub list for &infix:<+>
3. Filter out candidates that can't handle two ints
4. Call the best one
once we have local type inference / MMD optimization: 00:41
1. Wrap $int1 and $int2 into freshly constructed Int objects
jnthn sorear: 3 isn't really right fwiw.
sorear 2. Call &infix:<+>:(Int,Int)
jnthn But you probably know that. :-)
sorear once we have that + inlining
1. addl %ecx, %edx
jnthn: actually I haven't studied the MMD kernel all that much yet. I know it does funky things with posets and inheritence metrics, but that's about it 00:42
jnthn (The algorithm is just a case of looking through the sorted candidate list until we find one that matches, then calling that. Once we find a match, we (usually) ignore the rest of the list)
posets? 00:43
Gee, I hope I wasn't meant to know what they were when I implemetned it. :-)
It just makes a DAG based on an n-by-n smart-matching of type objects.
sorear posets are basically a kind of directed graph
jnthn It's a DAG.
sorear when you look at a poset, your frame of reference is the ordering
two nodes are > < or incomparable 00:44
jnthn If that view implies topological sorting, then it's probably fine.
jnthn is not a graph theorist :-)
sorear topological sorting is the link between the poset view and the DAG view
jnthn Ah, OK.
sorear they're the same concept seen from different angles
jnthn I hadn't heard the term "poset" before. Nice to know. 00:45
sorear (note that, since topological sorting is strictly a finite algorithm, infinite posets and infinite DAGs are not *quite* the same) 00:46
00:46 autin joined
jnthn
.oO( and now I'm glad I'm not a graph theorist :-) )
 00:48
sorear neither am I
diakopter sorear: I thought 2/3/4 can happen at compile-time
sorear diakopter: that's called an optimization
diakopter: and we don't have nearly enough manpower to do everything at once :/
diakopter :P 00:49
why so defensive?
sorear so, there will probably be a revision of rakudo with native ints but no MMD optimization
Personal failing. Please make me aware.
diakopter ok; I was just verifying that the model in my head was correct-ish 00:50
(that MMD can happen at compile-time)
for variables with type annotations.
sorear diakopter: Yes; that was in the second version of my explanation
diakopter (and for ones without, given inference)
sorear 17:41 < sorear> once we have local type inference / MMD optimization: 00:51
17:41 < sorear> 1. Wrap $int1 and $int2 into freshly constructed Int objects
jnthn and you know all the candidates and that they don't require runtime subset checks... :-)
sorear 17:41 < sorear> 2. Call &infix:<+>:(Int,Int)
could I have done this clearer?
diakopter well
sorear IMO the natural way to do this is specialized MMD trampolines.
"Call through this if you have an Int and an Int" 00:52
diakopter my question wasn't addressing your statements' clarity
sorear it might point to the only candidate
or, if there becomes more than one, it can be redirected to the general dispatcher
diakopter I don't see why an int would become an Int... why wouldn't there be an &infix:<+>:(int,int) 00:53
masonkramer I'm sorry to ask, but what does MMD stand for? (naturally, I understood very little of what just transpired in this room :))
sorear masonkramer: multi-method-dispatch 00:54
basically
you have a normal sub - there's one function, direct call
masonkramer ohh
this is same name, different function sigs
sorear you have a method (a virtual one, if you go that way) - one function per slot in the inheritence hierarchy
yes 00:55
00:55 jonrafkind left
diakopter sorear: did you see my question above? ^^ 00:55
sorear diakopter: Primitive types cannot be passed to polymorphic functions 00:56
So the MMD engine would have to receive Int and Int
diakopter I assume you're going to follow that up with an explanation of why Primitive types cannot be passed to polymorphic functions
sorear I'm not sure if it makes sense to have &infix:<+>:(int,int), in light of the fact that the MMD engine can never see an int 00:57
diakopter: main reason is that it screws up the garbage collector
the GC needs to know what's a pointer
that means that a given stack slot should either be able to hold Any, or int, but not both 00:58
diakopter ok. on my VM, it knows the difference between primitive types and reference types
sorear exactly
diakopter right, so it wouldn't "screw up the garbage collector"
sorear you can't write a single function that accepts Object and int in the JVM; I'm sure CIL is the same
diakopter eh? 00:59
sorear it doesn't "screw up the garbage collector" because the runtime manager /doesn't let you do it/
diakopter why can't you write a single function that accepts Object and int in the JVM?
masonkramer "you can't write a single function that accepts Object and int in the JVM" - but the dispatcher accepts a function name and a signature and finds the method that accepts two ints... 01:00
jnthn diakopter: I think sorear's point is that if you have a Capture and it's just a list of objects, things have to be auto-boxed. OTOH, you may be able to define a Capture type that can hold low-level types too.
sorear Because you can't. It's against the rules. Functions (I'm talking about functions here, not overloaded sets of functions) have to set an argument type
01:00 BrowserUk joined
sorear It can be [java.lang.Object; or I 01:00
not both
er, java/lang/Object
diakopter oh, I didn't realize you were talking about a single argument.
that wasn't clear 01:01
01:01 lest_away is now known as lestrrat
sorear Oh. Sorry about that. 01:01
masonkramer: Using the dispatcher to implement the dispatcher is going to require some very careful tiptoeing around the circularity saw.
It's probably possible, but I'm not smart enough to do it
diakopter jnthn: I suspect you're right. 01:03
sorear A few years ago, some far-too-clever ML hackers implemented a garbage collector with a built-in type inference engine, which was able to walk up and down the stack and determine, by looking at who passed what arguments where, what was actually a reference. They could pass full-word non-references to polymorphic functions.
jnthn: Exactly
jnthn: Although, if you "define a Capture type that can hold low-level types too", that's just autoboxing by any other name 01:04
diakopter I think my problem is I can't picture in my mind this thing you call "the MMD engine"
sorear Hmm, I'm typing rather quickly. Usually that means I'm attacking someone?
01:05 sahadev joined
diakopter I don't know; does it? :) 01:05
jnthn Could just be higher than usual coffee intake. ;-)
BrowserUk ?
phenny BrowserUk: 12 May 09:28Z <ruoso> tell BrowserUk please take a look at perl6-language history. TimToady sent your comments about the threading model there and I'm posting a reply...
masonkramer sorear: no, you're not - I am learning a lot - it feels like drinking from a firehose, but it's not aggressive
sorear diakopter: It can take many forms. Ours lives in src/pmc/perl6multisub.pmc, in particular the 'invoke' vtable 01:06
01:06 orafu left 01:08 orafu joined, wknight8111 left
diakopter I think it will be interesting to observe p6 code written when primitive types are implemented/available as well as Any. I postulate that there will be a goodly amount of primitive type annotations. 01:09
sorear You really only need them for function arguments 01:11
01:11 eternaleye left
diakopter given the presence/behavior of coercions, the most-used coercions will of course still work 01:11
sorear Everything else can be produced by a straightforward fixed-point analysis
Kleene++
01:11 BrowserUk left
diakopter sorear: do you mean as opposed to just-plain-locals? 01:12
sorear yes
diakopter ij
ok
01:14 justatheory joined
masonkramer fixed point analysis? "everything else"? 01:14
sorear Ok
First, you need a finite model 01:15
diakopter finite model?
sorear of Any
for this, we'll need to reduce Any to a finite set of values
like Any(), {all defined Ints}, {all objects of type ClassHOW}, etc
masonkramer Any is a finite set, ok
keep going :)
sorear now make a big table of all the variables in your function, and the subset of values it can take 01:16
initially, assume that arguments can hold the values in their signatures, and variables nothing
when you get to a statement, add all possibilities
masonkramer <phase space of a function signature S, loaded>
sorear since there are only a finite number of variables and a finite number of values in our model, you can only add a possibility a finite number of times 01:17
which means that the iteration is guaranteed to terminate
diakopter barring overflow?
sorear there's some brilliant work by Kleene showing that the state at the end of the iteration is consistant
so we now have a type signature for every variable 01:18
for best results, transform into SSA first
diakopter: Overflow?
diakopter never mind; I must be a bumbling idiot. Proceed. 01:19
01:27 snarkyboojum left 01:30 pencilk joined 01:33 Oktal joined
sorear oh, one more thing 01:40
in the ideal world, where we just use subsets of Any directly, the calculation is exact
you always get the precise sets of legal values
not subsets of Any, but subsets of Any^{number of variables} 01:41
but if you do it that way, it usually won't halt
since it amounts to running every possible trace of the function
when you use the finite model, you create a halting language based on the language, which you can use for analysis 01:42
the art of fixed point analyzers lies in choosing the right model - fine enough to tell you what you need, but coarse enough to keep runtime under control
01:44 snarkyboojum joined 01:45 sahadev left
masonkramer so...just to make sure I'm on the same page as you - fixed point analysis is a way of inferring type based on iterating through the entire phase M of the function signature + variables of a function, considering all possible values for every variable and argument in the entire phase space, and then inferring a type that can hold all possible values for each function? 01:48
*for each variable 01:51
?
sorear Yes. 01:54
Because the notion of "type" is so broad it actually underpins several different optimizations
nqp: loop { say(2); last; } 01:58
p6eval nqp: OUTPUT«Confused at line 1, near "loop { say"␤current instr.: 'parrot;HLL;Grammar;panic' pc 500 (src/cheats/hll-grammar.pir:197)␤»
sorear nqp: repeat { say(2); last; }
p6eval nqp: OUTPUT«Confused at line 1, near "repeat { s"␤current instr.: 'parrot;HLL;Grammar;panic' pc 500 (src/cheats/hll-grammar.pir:197)␤»
sorear nqp: while 1 { say(2); last; }
p6eval nqp: OUTPUT«2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2␤2􏿽xE2􏿽x90
sorear Help 01:59
masonkramer haha 02:07
sorear: it just seems rather slow (impossibly slow?) to iterate the entire phase space of a function, (with phase_space = set_size_of(Any)^n) in order to do type inference. I think there's something I'm missing up there around there is "choosing the right model...keep runtime under control" - what I don't understand is why this type inference would be happening at runtime 02:13
sorear masonkramer: the run time of the compiler
set_size_of(Any) is infinity
masonkramer but you can model it in a way that is feasibly computable 02:14
sorear the unapproximated algorithm only works in mathland
masonkramer that's what you're trying to tell me
sorear yes
masonkramer neato
02:19 agentzh joined 02:21 snarkyboojum left 02:22 justatheory left 02:23 lestrrat is now known as lest_away
masonkramer So - is compile-time type inference already in Rakudo? my $i = "25" + 10; #what is $i's type? 02:24
02:24 eternaleye joined
sorear masonkramer: Any 02:24
02:26 felipe left
masonkramer But could it theoretically be inferred to be an Int? Since, in the enumeration, it can only contain values that are legal for Int? 02:27
02:29 JimmyZ joined
sorear It could if any of this was implemented in Rakudo 02:31
masonkramer but from what you were saying earlier, there is no real benefit to doing that until MMD optimization 02:32
sorear Yes.
dalek p-rx: 458dbfe | sorear++ | (2 files):
Implement :flat :named (|%hash)
 02:39
02:39 eternaleye left 02:44 sahadev joined
colomon masonkramer: If you have a + with something that's not a Numeric type on both sides, each side is converted to Numeric if possible. 02:49
rakudo: say "25".Numeric
p6eval rakudo ed2409: OUTPUT«25␤» 02:50
colomon rakudo: say "25".Numeric.WHAT
p6eval rakudo ed2409: OUTPUT«Num()␤»
colomon rakudo: say "25" + 10 02:51
p6eval rakudo ed2409: OUTPUT«35␤»
masonkramer colomon: but I'm thinking about this from the PoV of compile-time type checking. $i with an Int type would tend to be more useful for that purpose 02:52
02:53 eternaleye joined
colomon and I'm telling you what Rakudo actually does. 02:56
masonkramer right - thank you 02:57
03:01 plobsing joined 03:22 Schwern joined, bubaflub left
dalek p-rx: f5f582d | sorear++ | (3 files):
Basic loop control (next/last/redo terms)
 03:29
03:37 xinming left 03:38 sahadev left
dalek p-rx: 379f1b2 | sorear++ | src/NQP/Grammar.pm:
Fix a typo (pmichaud++)
 03:41
03:45 patspam left 03:50 Oktal left
diakopter TimToady: did you see this one: lamp.epfl.ch/~phaller/doc/haller07a...sunify.pdf 03:51
03:52 rv2733 joined 03:57 jonrafkind joined 04:00 am0c joined
dalek p-rx: c5482a9 | pmichaud++ | (2 files):
Switch HLL::Actions to use find_codepoint opcode, bump PARROT_REVISION.
 04:03
colomon masonkramer: I keep on thinking about what you could know at compile time with that example. 04:10
On the one hand, if the equation is just "25" + 10, it can not only know the answer is an Int (assuming Str.Numeric can return an Int), it can know the Int is 35. 04:11
If all you actually know is Str + Int, then you can easily say the type is either Numeric or Fail. 04:12
But I'm not sure how much better than that you can get.
04:12 rv2733 left 04:33 Intensity joined 04:38 pencilk left
sorear colomon: there's only one multi that matches that situation, the (Any + Any) one, so the code is inlined to Str.Numeric + Int.Numeric 04:40
Int.Numeric is inlined to Int
Str.Numeric has a type signature which specs a return of Int|Num
so we see Int|Num + Int
which is (Int + Int)|(Num + Int)
= Int|Num
04:41 snarkyboojum joined
eternaleye In S05, under :exhaustive, it says 'say "@()";' - shouldn't this be @[], since it's a zen slice of @/ and not an invocation? 04:50
sorear @/ doesn't exist 04:53
@() is a list of all anonymous captures
05:10 xinming joined 05:24 felipe joined
dalek p-rx: c3b0e83 | sorear++ | src/stage0/ (3 files):
Rebootstrap so REPL stuff can use |% and last
 05:28
05:33 lest_away is now known as lestrrat 05:34 snarkyboojum left 05:38 Schwern left 05:49 [mark] joined
eternaleye diakopter: Fascinating example of parallel development: a similar paper was published the same year at a different conference, but in Haskell. See my message to p6-lang for the link. The main difference I'm seeing is that the Haskell one focused on client-server relationships, rather than actor-model, but unifying the two papers' approaches could prove very interesting. 05:53
06:03 hercynium left 06:04 hercynium joined
eternaleye diakopter: On reading more closely, they actually refer to the one I linked as citation 15 06:06
diakopter: Their 'receive' looks one _heck_ of a lot like our given/CATCH/etc - this paper's a great find! 06:08
And gather/take could probably be implemented trivially on top of their actor { ... } construct 06:10
06:12 pencilk joined 06:13 hercynium left, alester left
eternaleye probably using a send( gatherActor, 'get' ) -> receive { when 'get' { try { $gatherblock.() CATCH { reply($_) } } } } or some such 06:14
06:15 snarkyboojum joined 06:18 aindilis joined 06:19 hercynium joined 06:24 rv2733 joined
diakopter eternaleye: :) 06:27
eternaleye Or maybe more like actor { try { $gatherblock.() CATCH { my $taken = $_; receive { when 'get' { reply( $taken ) } } } } }
diakopter let's see what daniel says 06:28
eternaleye + a resume($taken)
That whole paper is just ridiculously elegent and intuitive, without sacrificing usability 06:29
It even would live the Perlish idea of 'you can modify the underlying system' 06:30
diakopter it received best paper at the conference at which it was presented 06:31
eternaleye The one I linked to was in the /topic back when Pugs was still the leading implementation - I'd say _this_ paper should be in the /topic today
diakopter heh. I assume you meant a ; before the CATCH
eternaleye Ah, yes 06:32
diakopter when viewed through a dark lens, ruoso's proposal(s) aren't all that dissimilar (being actor-ish) 06:33
eternaleye The beauty is that what I entered is very nearly feature-complete as an implementation of gather, all it would need added would be getArg and batch and batchArg variants
diakopter yeah but, someone has to implemenat actor and receive/reply and ... scala 06:34
eternaleye Which are trivial loops and/or maps of reply() and resume() with slice()
Well, the scala bits are almost superfluous. The actor() and receive/reply would be some work, yes, but the parrot hybrid threads GSoC would lay a great deal of the groundwork 06:35
"Unfortunately, on the JVM there is no safe way for library code to find out if a thread is 06:42
blocked." - this is something we could address, since Parrot is still figuring out its threading system
diakopter luckily for me, CLR provides ThreadState of which WaitSleepJoin means blocked 06:47
06:49 k23z__ joined
eternaleye Hm, andThen/orElse would need renamed, which would likely be an unbelievably colorful bikeshed 06:50
(orElse and orelse meaning vastly different things would probably be a Bad Idea) 06:51
Ooh, looking on his website there's a 2009 paper on the subject 07:02
(Philipp Haller, that is)
07:04 clintongormley joined
eternaleye His website is a gold mine: lamp.epfl.ch/~phaller/ 07:05
07:06 snarkyboojum_ joined 07:09 snarkyboojum left, snarkyboojum_ is now known as snarkyboojum 07:17 meppl joined 07:21 iblechbot joined 07:22 jonrafkind left
eternaleye diakopter: His paper on translucent functions is interesting, as it points out (and then fixes) a performance trap in the original paper. 07:22
Hm, maybe the trap could be avoided altogether by reusing MMD semantics rather than given/when semantics... 07:24
No, that wouldn't avoid it, just make it easier to minimize by using a well-optimized building block. The algorithm would still be flawed. 07:27
07:30 snarkyboojum left
eternaleye Hm, translucent functions might have a hard time coexisting with junctions 07:36
07:36 [mark] left 07:41 Su-Shee joined
eternaleye Also, it seems incompatible with when's use of smartmatching rather than just typechecking. Perhaps a shared queue like in the original, with 'virtual' queues for each condition, and virtual queues are populated with references to the object in the shared queue which matches the condition? That also obviates the need for the timestamp, since the messages are time-ordered in the shared queue, but it avoids having to rescan the whole queue on 07:41
react {...} calls. It would need backreferences to remove handled messages from virtual queues other than the matched one though
07:41 Schwern joined
eternaleye Join patterns could be represented by & junctions as the when clause 07:43
moritz_ good morning 08:07
phenny moritz_: 12 May 23:08Z <bubaflub> tell moritz_ ok. lemme know if i should add similar tests for the other twigils
08:22 plobsing left
moritz_ phenny: tell bubaflub sure, would be great to have some kind of systematic test for all of them - see perlcabal.org/syn/S02.html#line_1849 for the complete list; if you have troubles using one of them, feel free too ask 08:23
phenny moritz_: I'll pass that on when bubaflub is around.
moritz_ std: :$<a>
p6eval std 30620: OUTPUT«===SORRY!===␤Bogus statement at /tmp/2XpONBVXfq line 1:␤------> <BOL>⏏:$<a>␤Parse failed␤FAILED 00:01 113m␤»
moritz_ TimToady: shouldn't that construct a colon pair, like other combinations of colon + twigil'ed variable? 08:24
:$~foo
std: :$~foo
p6eval std 30620: OUTPUT«ok 00:01 110m␤»
08:29 Schwern left 08:34 dakkar joined 08:45 fridim_ joined 08:50 envi^home joined
pugssvn r30621 | moritz++ | [S06] fill empty section with a reference to what we have about that topic 08:51
08:58 snarkyboojum joined 09:01 dakkar left 09:02 dakkar joined 09:05 kfo joined 09:15 iblechbot_ joined
sjohnson hello 09:16
09:16 iblechbot left
sorear hi 09:17
09:25 proller left 09:31 am0c left 09:39 pmurias joined 09:40 proller joined
sjohnson sorear: hi 09:40
09:41 masak joined
masak oh hai, #perl6 09:41
sjohnson hji
dalek ok: 4bdbcd4 | moritz++ | src/subs-n-sigs.pod:
[subs] correct syntax in an example
 09:42
ok: 6418b41 | moritz++ | src/subs-n-sigs.pod:
[subs] first shot at introspection
masak rakudo: say ('h' .. 'j')[0, 2, 1]
p6eval rakudo ed2409: OUTPUT«Method 'postcircumfix:<[ ]>' not found for invocant of class 'Range'␤current instr.: '!postcircumfix:<[ ]>' pc 11329 (src/builtins/Role.pir:26)␤»
sjohnson （　｀ー´）
masak bug or feature?
should I be able to index a range?
moritz_ bug, IMHO
masak submits it as such
moritz_ alpha: say ('h' .. 'j')[0, 2, 1]
p6eval alpha 30e0ed: OUTPUT«hji␤»
sjohnson ... damn bugs..
moritz_ pugs: say ('h' .. 'j')[0, 2, 1]
p6eval pugs: OUTPUT«hji␤»
sjohnson sprays raid on rakudo 09:43
moritz_ all sane implementations agree :-)
elf: say ('h' .. 'j')[0, 2, 1]
p6eval elf 30621: OUTPUT«Can't call method "postcircumfix__91_32_93" without a package or object reference at (eval 128) line 3.␤ at ./elf_h line 5881␤»
masak in other news, I dig chromatic's post www.modernperlbooks.com/mt/2010/05/...angry.html
and I feel for Aaron Sherman on p6c. I hope he gets a good answer.
(partly because I want to know it too.) 09:44
09:45 am0c joined, TiMBuS joined
moritz_ rakudo: my regex foo { f.o }; say Grammar.parse('fuo', :rule(&foo)) 09:47
p6eval rakudo ed2409: OUTPUT«fuo␤»
moritz_ that's no the answer you were looking for, right? :-)
sjohnson rakudo: declare something; test the next thing; print what the value is 09:48
p6eval rakudo ed2409: OUTPUT«Could not find sub &something␤current instr.: '_block14' pc 29 (EVAL_1:0)␤»
masak yeah, at least part of it.
it doesn't answer the question asked: how do I call a grammar rule from an rx rule? 09:49
/<foo>/ seems to no longer work, is that intentional?
moritz_ nope; by latest spec it should first search for a lexical &foo, then for a method 09:50
rakudo doesn't do the lexical searching yet
masak hm; makes sense.
I think that it doing so is an R* priority.
moritz_ +1
masak preferably sooner. 09:51
cognominal what is the way to call a method when you have its name as a string? 09:55
moritz_ $obj."$method-name"()
masak parens required to protect against concat p5ism. 09:57
cognominal thx, I thought I tried it. 09:58
apparently noy
*not
09:59 moritz_ sets mode: +oo masak cognominal
moritz_ I've tried to change the subversion post-commit hook to put the first line of the log message into the email subject 10:00
sjohnson $obj.~$method-name; <-- won't this work too? 10:01
moritz_ I seriously doubt it
std: my ($obj, $meth); $obj.$method
p6eval std 30621: OUTPUT«===SORRY!===␤Variable $method is not predeclared at /tmp/IgoFtx8Z4L line 1:␤------> my ($obj, $meth); $obj.$method⏏<EOL>␤Check failed␤FAILED 00:01 113m␤»
moritz_ std: my ($obj, $meth); $obj.$meth
p6eval std 30621: OUTPUT«ok 00:02 112m␤»
moritz_ std: my ($obj, $meth); $obj."$meth"
p6eval std 30621: OUTPUT«===SORRY!===␤Unsupported use of . to concatenate strings or to call a quoted method; in Perl 6 please use ~ to concatenate, or if you meant to call a quoted method, please supply the required parentheses at /tmp/vytlUJK4dr line 1:␤------> my ($obj, $meth);
..$o…
moritz_ std: my ($obj, $meth); $obj."$meth"()
p6eval std 30621: OUTPUT«ok 00:01 112m␤» 10:02
moritz_ std: my ($obj, $meth); $obj.~$meth
p6eval std 30621: OUTPUT«===SORRY!===␤Confused at /tmp/hTvKMPPq95 line 1:␤------> my ($obj, $meth); $obj.⏏~$meth␤ expecting postfix_prefix_meta_operator␤Parse failed␤FAILED 00:01 112m␤»
moritz_ if there are problems with commit hooks, feel free to bug me 10:03
10:07 snarkyboojum left 10:10 snarkyboojum joined 10:11 yinyin left
masak here's someone else who doesn't like Second Systems: www.schockwellenreiter.de/blog/2010...nd-python/ 10:17
moritz_ he doesn't say nice things about Perl 6 either :-) 10:18
Su-Shee masak: he hoping for the same reasons that perl 6 will hopefully be a dead born child. ;)
also, is a famous german blogger idiot. ;) 10:19
he is.
masak he's hoping for a stillborn, yes.
sounds like something a person would say to stir up feelings.
maybe to get angry comments or something. 10:20
moritz_ he has the opinion that since he has spent learning an overly complex system, it's a disadvantage to simplify it
10:20 pencilk left
masak that's basically the argument of some Perl 5 people, though :) 10:20
Su-Shee he's the typical middle aged aging german engineer hoping for no change at all, ignore him. seriously.
masak except that they add that the overly complex system really is simple if you look at it the right way. :P
Su-Shee: consider it done. 10:21
moritz_ was about to write a highly sarcastic blog post and register it there as a trackback
but it's probably not worth the effort
Su-Shee moritz_: depends on how many new readers you want. ;) 10:22
masak also, the Internet tends to treat sarcasm as damage and route around it.
as chromatic wrote, positive blogging is worth so much more than defensive blogging. 10:24
Su-Shee a well the first rule of ruling successfully is to learn to ignore the masses and only listen to the right people. (not necessarily your friends or your peers though) 10:25
10:26 lestrrat is now known as lest_away
moritz_ well, don't ignore the masses, just ignore what they say :-) 10:26
Su-Shee no. ignore the masses. you'll get success by attracting the right group of early adopters. the masses follow _them_ not you. 10:27
10:27 agentzh left
moritz_ never heard of early adaptors in the context of ruling 10:27
Su-Shee moritz_: well you're a physicist and not a political scientist. ;) 10:28
10:31 JimmyZ left
masak rakudo: say try {} 10:32
10:32 PZt left
p6eval rakudo ed2409: OUTPUT«Null PMC access in type()␤current instr.: '_block14' pc 29 (EVAL_1:0)␤» 10:32
masak submits rakudobug
meh, we already had that one :) 10:33
do you know what I think is a source for hope? it's that most of the projects on the list on proto.perl6.org/ didn't exist a year ago. imagine where the Perl 6 ecosystem will be a year from now. 10:35
hm, maybe the 'fresh' medal should link to the project's commit log... 10:37
moritz_ didn't we have a chart of number of proto projects over time?
that would fit nicely onto that page too
maybe at the bottom
masak yes; I believe it's in Tim Bunce's talk.
aye; slide 83 of www.slideshare.net/Tim.Bunce/perl-myths-200909 10:38
moritz_ moritz.faui2k3.org/tmp/proto-projects-list.png 10:41
more up to date
10:42 pmurias left, pmurias joined
masak seems we peaked in May :) 10:44
mathw waves 10:45
moritz_ particles
masak duals
moritz_ I've beautified it a bit 10:48
masak out of 36 hand-picked bugs deserving (IMHO) to be fixed by the Rakudo Star release, 14 have been fixed so far. use.perl.org/~masak/journal/39597 10:49
'There are currently over 400 new/open bugs in RT.' 10:50
masak chuckles
10:55 cxl_ joined 10:59 cxl_ left 11:11 pmurias left 11:17 rv2733 left 11:20 snarkyboojum left, snarkyboojum joined 11:22 masonkramer left, masonkramer joined
colomon just finally unpacked his copy of the Dragon Book. Also Sligo and Michigan tunebooks, Newfoundland songbooks, and my most heavily used math books. Good box! 11:37
moritz_ and I thought colomon knew math already :-)
colomon I know what books to look in, at any rate. :) 11:38
k23z__ colomon, math books, which ? 11:39
*which ones
colomon In particular, Linear Algebra (Friedberg / Insel / Spence) and Vector Calculus (Marsden / Tromba). Plus Numerical Recipes in C. 11:40
moritz_ www.amazon.de/Taschenbuch-Mathemati...-1-catcorr # the bible
colomon "Tashcenbuch"? 11:42
*ch
Su-Shee *HAHA* the bronstein calling a taschenbuch is great :) 11:43
colomon: poket book
pocket
colomon It's 1216 pages long, yes?! 11:44
If that's the table of contents there, it looks like it covers a *lot* of useful stuff.
k23z__ you must be an engineer if I'm not mistaken 11:45
colomon k23z__: I'm not an engineer, but I write software used by engineers. 11:46
jnthn o/ 11:50
11:50 mberends joined
masak jnthn! \o/ 11:50
colomon \o 11:51
jnthn masak!
masak jnthn: I've been constantly reminded, in the past few days, why it is that we elected to call the making of presentation slides 'procrastinating'. :)
jnthn masak: Ты еше в России? :-) 11:53
masak no, no, not yet :)
jnthn doesn't need to be envious yet then ;-)
masak my plane is Sunday.
s/is/leaves on/
jnthn Oh...loads of time to procrastinate!
masak well, um... :/
I accidentally an Esperanto camp in between. :P 11:54
jnthn Way to make your brain asplode. :-)
Juerd Ĉu plena kampado? :P 11:55
mberends \o nice to be online for a few minutes :-)
dalek kudo: 70795ca | (Martin Berends)++ | tools/test_summary.pl:
add a second report example to tools/test_summary.pl
masak Juerd: ne vere. sed mi ne trovis pli bonan tradukon por 'renkontiĝo'.
Juerd O hai mberends
masak mberends: hi!
jnthn mberends: \o/ 11:56
mberends hai Juerd, sorry I forgot to tell you I had to miss Amsterdam.pm :-(
Juerd masak: meeting. :)
masak Juerd: 'meeting' sonas kiel maltro granda okazo, ĉar nia daŭros dum tagoj.
Juerd: kaj enhavas 'meetings'.
Juerd masak: Congress.
masak tio sonas tro serioze :P
k23z__ masak, what language is that ? 11:57
swedish or esperanto ?!
or.. ?
masak k23z__: it's Esperanto.
Juerd Vi estas malkontentigebla :)
masak 100 years older than Perl. :)
Juerd: we're going to a renkontiĝo. :) that's it. 11:58
Juerd Make that 200
masak Juerd: please show me that math.
11:58 am0c left
Juerd Oh, I'm mistaken 11:58
I thought Esperanto was 17xx but it's 18xx 11:59
And I had no idea that it was indeed almost exactly 100 years older :)
masak :)
11:59 M_o_C joined
Juerd I want to live through 2087 and see what language that year will bring. 11:59
jnthn Perl 7? 12:00
Juerd jnthn: Esperanto, Perl 6, Perl 7?
s/6/5/
Maybe a language that unifies speech and programming
(No, not BASIC)
jnthn COBOL?
<ducks>
masak there are uncanny parallels between Perl and Esperanto. both had benovelent dictators. both were 'open source' but centrally controlled in some sense. both 'contain the world' in the Mahler sense.
I guess that would make Perl 6 Ido. :( 12:01
Juerd No
masak NO! :(
Juerd Ido is not Esperanto
Perl 6 is Perl
masak also, Perl 6 wasn't developed mostly by the French. 12:02
Juerd A someone less awkward analogy would be Esperanto : Ido :: Perl : Kurila
s/one/what/
12:02 orafu left
masak but Ido does have a spec, and it does tend slightly more towards sanity and 'logic' than does Esperanto. 12:02
jnthn masak: If it was, we'd have an easy explanation for why it's taking so long though. ;-) 12:03
"La grèves!" ;-)
k23z__ Juerd, is Kurila like actively developed or something, I saw it on CPAN but is it actually used ?
masak jnthn: it's not to late to blame it all on the French, I guess.
cognominal: no offense.
jnthn menat it mostly light-heartedly...but hopes he never has the misfortune to visit France while a strike is taking place again. 12:04
:-)
12:04 orafu joined
masak 'kurila began life as "Perl 7". The name caused so much controversy that it was changed to kurila.' www.perlfoundation.org/perl5/index.cgi?kurila 12:04
who would've guessed? :P
in other news, I've decided to rename Yapsi into Perl 8. 12:05
:P
k23z__ "kurila is an experimental, incompatible fork of Perl 5 by Gerard Goossen. Its intent is to allow experimentation with the internals without the burden of backwards compatibility. kurila radically rewrites and even removing some features."
12:06 bluescreen joined
masak is happy there are people exploring those venues 12:06
12:06 bluescreen is now known as Guest41966
masak it explores the extent of TimToady's statement about something based on insanity still being insane. 12:06
Juerd k23z__: It's being developed but has a user base of 1
mberends gerard++'s talk at the Dutch Perl Workshop was very impressive 12:07
k23z__ Juerd, haha
masak even Perl 6 has a larger user base than that... :)
Juerd It became pythonic in syntax, at least in the sense that it now has semantic identation.
s/ident/indent/
jnthn figures he should deal with some of his work pile
k23z__ masak, is there any estimate for p6's user base ?
masak, the current one that is 12:08
Juerd k23z__: Well, at least 2 :)
Su-Shee now you're modest. it's at least 3.
masak k23z__: I don't know what such an estimate would measure. perhaps Rakudo downloads per month?
Juerd \o/
masak k23z__: or number of querents per day on the #perl6 channel?
but neither of those are very accurate to begin with. 12:09
k23z__ masak, I think unique IP address downloads of Rakudo would be pretty accurate
jnthn number of authors with module sin proto? :-)
(under-estimate but a maybe interesting statistic)
Juerd wants to commit a module sin 12:10
masak jnthn: that's not 'users', that's 'project authors'. 12:11
12:12 am0c joined
masak k23z__: there are some numbers here: github.com/rakudo/rakudo/downloads 12:12
k23z__: but many of us here don't register there because we use the Git repo directly.
jnthn masak: Yeah...it gives a lower bound though I guess. 12:14
But certainly only tells part of the story.
masak also, not all who download go on to become what I'd call 'users'. 12:15
maybe they just like having a 'perl6' binary on their computer :)
I know I do.
12:17 hugme joined
moritz_ hugme: add Juerd to sin 12:17
hugme moritz_: sorry, I don't know anything about project 'sin'
moritz_ bad luck :/
masak nothing that can't be fixed.
jnthn Just gotta find something it can be an acronym for. :-) 12:18
12:19 JimmyZ joined
masak something recursive, maybe. "Sin Isn't Nutritious"? 12:20
mberends loliblogged: blogs.perl.org/users/martin_berends...mance.html 12:22
masak mberends: interesting. mberends++ 12:23
mberends: those 'freezes' have been reported earlier by quietfanatic, when he did some performance-intensive graphics, or even just printing lines to $*OUT. 12:24
mberends yeah, they're fairly random
masak at that time (maybe 3/4 year ago), it created a sort of oscillating, wobbly feel to the execution. 12:25
jnthn Said 'freezes' are probably when GC runs happen.
masak aye.
jnthn mberends++ # will read it in detail later 12:26
12:26 fridim_ left
masak mberends: suggestion: add a link in the last sentence to perlgeek.de/blog-en/perl-6/contribu...ounce.html 12:27
mberends sure, thanks :)
masak moritz_: there's an unclosed <code> in perlgeek.de/blog-en/perl-6/contribu...-test.html
12:27 ruoso joined
mberends masak: link added 12:31
masak \o/
12:34 snarkyboojum left
mberends \o (moving out of street-wifi range), working a bit on zavolaj/mysql too 12:34
12:35 mberends left 12:37 orafu left, orafu joined
masak yay! real DB support in Rakudo would be sweet. 12:37
haven't seen much blogging about the progress of that.
jnthn masak: The mysql example in the zavolaj repo works decently, afaik. 12:38
It just needs for somebody to package up a module nicely.
"just" :-) 12:39
masak that sounds intensely promising.
12:40 xinming left
masak hah! the example involves a table called 'nom'! github.com/jnthn/zavolaj/blob/maste...lclient.p6 12:40
I wonder who wrote that... :P
jnthn Yeah...mberends++ wrote it after we'd been for lunch. :-) 12:41
masak huh! ok, he was my second guess. :) 12:42
12:42 Guest37407 left
jnthn I'm not to blame for ALL lolspeak in example code. :P 12:42
masak true. 12:43
jnthn
.oO( I guess I can't wriggle out of the blame for all the Slovak proclaiming the virtues of beer in the spectests, though. )
masak lol!
that's a tweet in itself: "How many modules do you know whose spectests are full of Slovak proclaiming the virtues of beer?" 12:44
k23z__ when p6 grows will one be able to prove that the code he writes is correct, like those Haskell people do ? 12:46
masak hm, only four hits for 'pivo' in the spectest suite.
k23z__: funny you should mention that. my talk on Monday will partly be about that.
k23z__: want a sneak peek?
k23z__ masak, sure thing
masak prepares a gist 12:47
jnthn masak: Clearly I don't write enough tests. :-)
masak k23z__: gist.github.com/399787
k23z__ I've been reading from Dijkstra's book "A Discipline of programming" but some stuff came up and I had to postpone it ..
masak, reading .. :)
masak k23z__: note the PRE blocks in the loops.
there's no proof-of-concept for it yet, but given things like that, it should be possible for a piece of software to walk the AST of the program and prove that the sub actually sorts the elements. 12:49
jnthn: "dnes je horuci a potrebujem pivo" 12:50
was that during the summer? :)
seems it was. July. 12:51
k23z__ masak: that is very nice
jnthn Presumably. ;-)
k23z__ masak, does the compiler actually evaluate those PRE-conditions ?
masak, how about post-conditions ?
masak k23z__: no, right now there's no compiler that can handle those.
k23z__ masak, do you plan to implement POST conditions also in the language ?
ah I understand 12:52
masak k23z__: I'm hoping Yapsi will be the first to do so.
k23z__: both PRE and POST are in the spec.
S04.
they can also be applied to other things than loops; methods, for example.
if they're put directly in a class block, they count as being applied to all methods in that class. 12:53
so it's a way to uphold invariants. and it has always been a place where I think static analysis could perform wondrous things.
k23z__ masak, I like this feature very much, I hope rakudo gets it also, it would be a big + :) 12:55
12:55 xinming joined
k23z__ masak, are there other concepts from the realm of formal methods that you plan to include ? 12:55
masak k23z__: perhaps there would be, if I had any formal education in this area. but I don't, I just find it exciting. :) 12:56
it feels like a place where pragmatism and theory can meet a bit. 12:57
12:57 araujo left, araujo joined
mathw has been reading about Perlesque, and likes 13:00
13:00 XaeroOne joined
masak Perlesque is The Awesome, and it's intriguing to think that it might end up running STD.pm6 somehow. 13:04
mathw yes 13:06
and that it could maybe also be implemented on the JVM
[Coke] anyone able to build rakudo-latest with parrot-latest? 13:10
moritz_ [Coke]: worked for me a couple of hours ago 13:11
trying again now
[Coke] nopaste.snit.ch/20543 is my failure. 13:12
moritz_ huh.
yes, I get the same 13:13
did anybody touch nqp in parrot?
it's definitively not a rakudo change causing that 13:15
dalek kudo: 7e0aa33 | moritz++ | build/Makefile.in:
[build] missing pod2man should not be fatal
 13:16
moritz_ masak: thanks, markup fixed
13:17 dju left
[Coke] moritz_: npq, no. pct, I think so. 13:17
if time permits, I'll see if I can figure out which recent change did that.
moritz_ it's parrot-nqp that's being executed
and I'm pretty sure it doesn't use PCT under the hood :-)
13:19 dju joined
moritz_ erm 13:27
I meant PGE
PCT *is* used
all those TLA are confusing me :(
masak be glad they're not FLA, then :) 13:28
jnthn FLA isn't a FLA. 13:30
mathw nitpicker
13:30 PZt joined
jnthn namecaller 13:31
masak 'for any given xLA, if xLA is a xLA, then x == 3' :)
moritz_ only if |x| == 1 13:32
FoLA 13:33
masak ooh!
FivLA
moritz_ SixxLA
masak lol
jnthn SevenLA actually works. 13:34
Well
If you can make SEVEN into an acronym... :-)
13:35 dju is now known as dju_, dju_ is now known as dju__, dju__ is now known as dju 13:40 XaeroOne left, BrowserUk joined
BrowserUk ? 13:40
13:41 BrowserUk left
masak ¿ 13:41
13:45 plobsing joined 13:48 M_o_C left
k23z__ masak, if you have time, read Dijkstra's book, he invented this stuff, I think formal methods are superior to test-driven development, but they require a lot of studying 13:52
masak k23z__: thanks for the tip. 13:53
PerlJam k23z__: sure, formal methods may be superior, but they have a much higher cost in terms of time. 13:54
k23z__ PerlJam, yeah but don't you feel very good when you know you wrote a piece of code that can't go wrong, like.. ever ?
PerlJam (yes, this could be a various of wanting something "fast" or "correct" :)
13:54 lest_away is now known as lestrrat
PerlJam s/various/variation/ 13:54
k23z__: I'm happy with "good enough" rather than "perfect" usually 13:55
13:55 JordiGH joined
PerlJam But, it depends on the application. If I were writing some software that controlled a pace-maker or the trajectory of an airplane or something, "perfect" would tug on me a little more :) 13:56
masak I wouldn't mind 'perfect' more often if it was cheap. 14:01
k23z__ PerlJam, now that you mention it, there's a book called "Why programs fail" and there is an example about a software bug that caused the death of a pilot and a passenger on a plane landing in Poland
PerlJam, I'm sure there are many much more disastrous results of software bugs(probably in that book as well) 14:02
I haven't read it all yet tough ...
*though
JordiGH I'm afraid I can't let you land this plane, John.
[Coke] I think notfound fixed the build,t esting...
k23z__ JordiGH, hahaha 14:03
like HAL 3000 , or what was the name ?
JordiGH The machines, they're really out to kill.
[Coke] HAL9000 14:04
(and his sister, SAL)
masak there was something called a Therac-25, used in medicine. it contained a race condition and killed three patients with excess radiation. en.wikipedia.org/wiki/Therac-25
JordiGH Ah, that's a famous example, masak. I remember having to read about that in my engineering ethics class in undergrad. 14:05
masak there's also en.wikipedia.org/wiki/Ariane_5_Flight_501 14:06
PerlJam software errors have destroyed expensive rocket systems for NASA; they have caused massive power outages; and they have killed people. Software is a dangerous thing 14:07
[Coke] build fixed. tests added by NotFound++ to insure that regression doesn't occur again.
masak hardware exception during casting from floating point to int.
PerlJam (I'm also willing to bet that the Toyota sticking-gas-pedal problem is really a software error)
masak [Coke]++
masak read that as 'test added to NotFound++'
JordiGH described by patient Ray Cox as "an intense electric shock". It caused him to scream and run out of the treatment room 14:08
[Coke] masak: notfound++ did all the work. I just say "hey, is this borked?"
masak [Coke]++ # anyway 14:09
[Coke] ... I wonder why, when I let myself type on autopilot, that my grammar goes to hell.
PerlJam [Coke]: your reptilian brain isn't very grammatical.
[Coke] PerlJam: ... If it can type, it can handle grammar. :P 14:10
PerlJam typing is easy, that's just a technical skill (like stacking bricks atop each other); grammar requires thinking about what you're typing (like stacking bricks in the form of a house instead of just a pile of bricks) 14:12
jnthn Grammars overrate is
14:12 mikehh left 14:13 kensanata joined, Helios left
masak agree I not. grammar without, extract exceedingly sentence becomes structure from difficult. 14:14
PerlJam masak: I think you mean "extracting" 14:15
;)
masak are you complaining about my lack of grammar? :)
14:20 Helios joined 14:22 JimmyZ left
PerlJam wonders if Perl 6 is an "emerging language" ... emerginglangs.com/ 14:30
14:31 bubaflub joined
tedv It's clearly a language. How do you define "emerging"? 14:31
masak if it's coming out of something, like a nest or a burrow, it's emerging. 14:32
PerlJam The thing is ... I don't see a Perl 6 person in this list radar.oreilly.com/2010/05/announcin...angua.html 14:34
And looking at the other languages ... Perl 6 should be there. 14:35
moritz_ we just need somebody to present it, no?
PerlJam Someone who's going to OSCON. 14:37
14:37 tedv left 14:42 k23z__ left 14:45 TiMBuS left 14:50 mathw left 14:51 mathw joined 14:52 pmurias joined 14:53 JordiGH left, k23z__ joined 15:01 plobsing left 15:07 Khisanth left, kensanata left, lisppaste3 left 15:08 PerlJam left 15:09 PerlJam joined, clintongormley left 15:10 clintongormley joined, alester joined
diakopter PerlJam: TimToady said Perl [6 or otherwise] shouldn't be there since it already has its own oscon track 15:11
pmurias diakopter: hi 15:12
diakopter hi
pmurias diakopter: how is the class implementation progressing? 15:13
15:19 Su-Shee left
diakopter pmurias: nothing in 3 days; $work 15:21
15:22 Khisanth joined
diakopter jnthn: when a class is declared inside another block, and that block runs again, are the class' methods' updated wrt their outer_lex pointers? 15:23
15:24 mantovani left, Lorn left
jnthn diakopter: For anoymous classes certainly yes. 15:24
I guess for others too 15:25
Well
jnthn ponders
diakopter was pondering this yesterday
jnthn I think the answer is yes, and I think Rakudo probably gets it right.
roles are different though in that sense.
A role body doesn't run until the role gets consumed in some way. 15:26
diakopter oh
jnthn Since they body may depend on parameters.
diakopter what about a parametric class then
jnthn Don't exist.
Only roles are parametric in Perl 6.
diakopter ohh 15:27
hee
pmurias diakopter: did you set svn:eol-style on files in the sprixel repo?
diakopter no; feel free to set it however you like and standardize the eols 15:28
when does the role declaration body run 15:29
jnthn diakopter: When a role with a not-yet-seen set of parameters is needed 15:30
Where needed means "composed" really
diakopter ok 15:31
jnthn Since puns are done out of composition.
diakopter similar to when a .NET generic class is composed at runtime from a runtime-generatd type, I guess.
jnthn So essentially you have a "cache" of captures to roles.
15:31 iblechbot_ left
jnthn A role in Perl 6 is actually always a role factory. 15:31
That can potentially generate many roles.
diakopter ok. 15:32
jnthn rakudo: role Foo[::T] { say T }; class C1 does Foo[Int] { }; class C2 does Foo[Str] { }; class C3 does Foo[Int] { } 15:33
p6eval rakudo 7e0aa3: OUTPUT«Could not find sub &say␤current instr.: '_block52' pc 437 (EVAL_1:177)␤»
jnthn oh, grr
jnthn wonders if that bug goes away with pmichaud++s upcoming lexical/closure fixes.
But anyway it'd print Int\nStr\n
Not Int\nStr\nInt\n 15:34
Since we already manufactured a role for Int
e, for when we pass the parameter Int
PerlJam diakopter: re emerging langs ... really? that seems ... odd. 15:36
15:39 Lorn joined
PerlJam Sure perl has it's own track, but the point of emerginglangs (as I understand it) is to communicate and focus on language design/implementation issues that may be shared across languages. Not having Perl 6 in that camp leads to further isolation I think. I mean, I'd rather have the C# or Clojure or whatever people talking about the cool things they picked up from Perl 6 than have Perl 6 be invisible. 15:39
pmurias is C# an emerging language? 15:40
diakopter PerlJam: besides, all 80 attendance slots for EmergingLangs are full 15:41
PerlJam pmurias: they're on the list ... emerginglangs.com/speakers/
diakopter so I couldn't attend :)
PerlJam diakopter: you could if you were invited as a speaker :) 15:42
diakopter and only 80 attendance slots with 40 speakers ... seems... odd.
moritz_ elite conference?
diakopter ok, 28 speakers.
PerlJam I'm not sure what to make of that either.
the ratio is still off IMHO
but, yes, I guess it's an elite conf of language hackers. 15:43
(ergo someone from Perl 6 should be there! ;)
diakopter no it seems emerginglangs wanted a language-duke-it-out session during oscon, and it started to get a lot of attention on twitter, so oscon brought it under its umbrella and limited its attendance
since after all, if those 80 people attend the TWO FULL DAYS (5 rounds of wrestling per day?) of emerginglangs, they would miss 2/3 of oscon 15:44
I don't know. seems more like a worship service to me 15:45
15:46 XaeroOne joined
diakopter more likely, the 80 attendees would attend only a small part of the 2 days each 15:46
I mean, PyPy isn't a language 15:47
XaeroOne doesn't perl6 have versions like perl5 ie. perl 5.12 etc?
moritz_ perl5.something is not a Perl 6 version
masak XaeroOne: you mean like Perl 6.2, Perl 6.4, etc? 15:48
XaeroOne yeah, no what I mean is are there version like perl 6.10 or soemthing? I cant find anything like that?
right
diakopter XaeroOne: do you mean Perl 5 mode when using the perl6 binaries?
oh
PerlJam XaeroOne: Yes, Perl 6.0.0 is yet to be released. 15:49
XaeroOne :o
masak XaeroOne: no, it doesn't yet. in that sense, we're still working up to 6.0.0.
XaeroOne: ...which won't be a program like Perl 5, but a document. a specification.
XaeroOne when is 6.0.0 going to be released?
moritz_ when it's done
masak Christmas.
but probably not Christmas 2010.
PerlJam XaeroOne: in the mean time, there's Rakudo ... rakudo.org/how-to-get-rakudo
diakopter XaeroOne: there isn't a schedule for finalizing the specification. see perl6.org/specification/ 15:51
15:51 rmrfslash joined
XaeroOne i am just about learning perl5 and I read the perl5 to perl6 and differences between perl5 and perl6, now I don't like to learn perl5 anymore :s 15:52
masak XaeroOne: why not?
it's a quite powerful language.
with lots of tools and modules.
and a huge user base. 15:53
pmurias diakopter: should i add [perlesque] to commit messages in perlesque
XaeroOne yeah, i just fed up of people pointing out whats wrong with perl5
pmurias XaoeroOne: and a mature implementation
15:53 gpw left
XaeroOne what is perlesque? 15:53
15:53 gpw joined
masak XaeroOne: people will always criticise Perl, especially those who don't use it and haven't tried to learn it. 15:54
XaeroOne i like perl though, specially the special characters :)
masak you mean the sigils? $@%& 15:55
diakopter XaeroOne: perlesque is a p6/p5-ish language that tries to stick to the strongly-typed subset of Perl 6, so that it can run an automatically translated version of Larry Wall's standard parser for Perl 6
on the CLR 15:56
XaeroOne i like the ^/\>< also, i hate the __DATA__ thingy, it doesn't look nice at all
diakopter (Mono on Linux/Mac/Windows or .NET on Windows/Silverlight)
XaeroOne maybe it would make more sense to the python people :P 15:57
dalek meta: r271 | pawelmurias++ | trunk/Sprixel (50 files):
[perlesque] set eol-style:native
 15:58
XaeroOne ok so if I write a basic perl6 program will it still run when perl 6.0.0 is finally released?
15:58 am0c left
XaeroOne coz if there are a lot of changes then it would be better to write it in perl5 and then re-write it later in p6? 15:59
moritz_ the effort of re-writing a Perl 6 program to a later version of the Perl 6 spec is larger than rewriting a p5 program to Perl 6 16:00
16:00 rmrfslash left
PerlJam moritz_: s/is/can be/ 16:01
depends on the the specific features used and how complex the program is.
masak XaeroOne: you're looking for some reassurances that things will stay the way they are now. most things probably will, but some things might change drastically. there's no way to tell.
diakopter perlesque: my $sw = System::Diagnostics::Stopwatch.new(); $sw.Start(); loop (my $a=100000000;$a>0;$a-=1) { }; say($sw.Elapsed) 16:02
p6eval perlesque: OUTPUT«00:00:00.3721830␤»
diakopter perlesquel: .
p6eval perlesquel: OUTPUT«00:00:00.3853940␤real 0.44␤user 0.44␤sys 0.00␤»
masak XaeroOne: if you want stability now, Perl 5 is the version for you.
diakopter meh
masak++
PerlJam XaeroOne: also note that many Perl 6 features have been ported to Perl 5 in one way or another.
masak XaeroOne: those of us who like Perl 6 at the present date, almost by definition, have some way to cope with the lack of stability. 16:03
PerlJam XaeroOne: Perl 6 OOP can be done with Moose in Perl 5. Want everything to be an object? "use autobox" in Perl 5.
XaeroOne: want given/when and smart matching? Just make sure you have a newish version of Perl 5. :) 16:04
XaeroOne yeah, adding things like smart match and switch and stuff like that to perl5 is a good idea, over time it will make the transition to 6 easier
moritz_ though the smart matching in perl 5 isn't really comparable to that of Perl 6
masak XaeroOne: the reaction you had -- reading about Perl 6, and then no longer wanting to learn Perl 5 -- is probably one that people can relate to in here. that said, the more I learn about Perl 5, the more I realize what a nice language it is.
moritz_ it just doesn't work well if you don't have user exposed types
XaeroOne but more i read about perl6 the less it make me want to learn perl5 :( 16:05
PerlJam XaeroOne: Perl 5 makes a really good base for learning Perl 6 16:06
XaeroOne perl6 is so awesome, it has the best things of many things
masak XaeroOne: except we're not done building it yet.
XaeroOne: not wanting to scare you off here, just being honest.
you can do some cool things already with Rakudo et al.
but not everything.
XaeroOne python has this 2to3 script that converts python 2.x scripts to 3.x scripts, any possibility that we could have one pl2p6 kinda thing? 16:07
masak yes.
XaeroOne cool
masak current plan is TimToady writing a grammar for Perl 5.
similar to how STD.pm6 is a grammar for Perl 6.
then you could have some tool walk the parse tree and translate somehow. 16:08
there's already some MAD hooks in the Perl 5 compiler to do something like that. but I believe it's bitrotted some.
pmurias having more people work on the grammar for Perl 5 would be great
XaeroOne ok so on sourceforge>parrot-win32 i need to get what to get p6 working: parrotwin32 setup, parrot-rakudo addon and the docs? 16:10
16:10 molaf joined
pmurias diakopter: how hard would it be to make perlesque clean up it's internal state after running a file 16:11
diakopter pmurias: the two .exe it produces?
easy
oh, you mean in memory 16:12
pmurias yes
diakopter maybe an hour work; all the static things just need made instance things
pmurias i experimented with making it run files sequentially from standard input but i spit out some strange errors
diakopter heh
16:13 XaeroOne left
diakopter jnthn: did you see perlesque can use things from System.dll :) like Stopwatch :) 16:13
jnthn diakopter: Yes, noticed. :-) 16:14
diakopter too bad perlesque doesn't have leaf-detection (for routines), so that the int in the example above can be a true int 16:15
in that example, it's being loaded/stored as a field on a lexpad every time
actually 16:16
it could do it.
since perlesque doesn't contain hash indexing (maybe it shouldn't?) as { }, it could just lookahead to make sure there's no { before the next } 16:17
oh nm that's wrong :D :D
commutalism&
16:19 mantovani joined 16:21 rv2733 joined
TimToady MTTLA :P 16:28
LTATLA :)
16:29 kerframil joined
sorear masak: I gave up on kurila when he removed 'tie' and operator overloading without compensation, because "such features confuse people" :/ 16:31
masak sorear: huh. 16:32
sorear it's certainly interesting, but it violates "Perl will remain Perl" 16:34
can't call it Perl 6 if it doesn't give you rope. :)
masak guess that's why he called it Perl 7 :P
nom & 16:35
16:35 masak left 16:36 justatheory joined
[Coke] huh. www.perl9.org/ 16:46
jnthn std: my @baz = @foo >>>>+<<<< @bar; 16:48
p6eval std 30621: OUTPUT«===SORRY!===␤Variable @foo is not predeclared at /tmp/0JNBSI2XV9 line 1:␤------> my @baz = @foo⏏ >>>>+<<<< @bar;␤Missing << or >> at /tmp/0JNBSI2XV9 line 1:␤------> my @baz = @foo >>>>+<<<⏏< @bar;␤ expecting
..infix_circumfix…
TimToady std: 1 >>[>>+<<]<< 2 16:50
p6eval std 30621: OUTPUT«ok 00:01 110m␤»
16:55 cdarroch joined, cdarroch left, cdarroch joined
TimToady std: 1 >>>>[+]<<<< 2 17:00
p6eval std 30621: OUTPUT«ok 00:01 112m␤»
TimToady that also works
problem is, +< is an operator 17:01
17:01 pmurias_ joined, pmurias_ left
TimToady and >>op<< is rope :) 17:01
17:03 dakkar left
TimToady wonders if we should add op like we added ... 17:05
17:09 k23z__ left
TimToady moritz_: technically, the <a> in $<a> is a subscript, and < isn't a twigil. that does not mean we shouldn't force it to dwim though 17:11
TimToady has visions of :$<a><b><c> turning into 'a/b/c' => $<a><b><c> 17:12
or of :$<a/b/c> looking up $/<a><b><c> 17:13
bbl & 17:14
PerlJam has visions of xpath in perl 6 now 17:15
ruoso would strongly suggest a different subscript operator to that, since "/" is quite a regular character to be in a hash key... 17:17
phenny ruoso: 01:09Z <BrowserUk> tell ruoso I've subscribed to p6.lang, but I've no idea how to reply to posts that happened before I subscribed? Can you post a "placeholder" reply to your 3rd revision, so I can reply to that when it arrives--Or would that bring down the ire of the list upon you?
moritz_ it looks like a neat shortcut, but it breaks the symmetry between Match and Hash+Array 17:18
17:22 SmokeMachine joined 17:31 pmurias left 17:36 eternaleye left 17:37 patrickas joined 17:43 jaldhar_ left 17:52 ash___ left 17:53 ShaneC joined 17:54 ShaneC left 18:00 iblechbot joined
diakopter ruoso: did you get a chance to read that other paper Nigel Sandever and & I suggested? 18:01
18:03 iblechbot left 18:10 perlygatekeeper left 18:19 rv2733 left 18:29 stepnem left 18:31 masak joined
masak ahoy! 18:32
18:34 SmokeMachine left
masak [Coke]: that perl9 page is a bit of good-natured fun. the person obviously knows his Perl 6 in order to make the page in the first place. 18:34
jnthn masak: ahoj! nemam pivo. 18:35
masak jnthn: you'll just have to wait until they open again :P
18:36 stepnem joined 18:37 kel joined
[Coke] masak: it's inspiring me to make a perlX graph. 18:39
masak what's a PerlX graph? 18:40
also, it's an interesting mix of sillyness, intentionally bad ideas, and things that Perl 6 already does with no extra syntax or semantics.
[Coke] masak: I'll post the graph here when it's done. 18:41
masak excellent. 18:42
18:45 proller left 18:55 envi^home left 18:56 patrickas left 19:01 dakkar joined 19:14 eternaleye joined 19:35 eternaleye left 19:36 SmokeMachine joined 19:44 \shade\ is now known as toshade, toshade is now known as \shade\
[Coke] masak: bah. google is lying to me. will make it difficult to trust my graph. :P 19:46
19:46 \shade\ left, \shade\ joined
masak gets an idea of what the graph is about :) 19:46
in what way is Google lying to you?
19:47 eternaleye joined 19:50 \shade\ is now known as shade\ 19:52 SmokeMachine left
[Coke] response in browser and programatically are not teh same. 19:52
19:56 thraidh joined
thraidh hi 19:57
i was just trying to write my first perl6-program with rakudo
and i got somewhat confused by the usage of rules throughout the platform 19:58
is it true that rules are slightly different whether i'm using parrot, nqp or rakudo?
19:58 eternaleye left 19:59 shade\ is now known as {shade}
thraidh especially when it comes to expression-matching with infix operators 19:59
[Coke] note that rakudo and nqp are both implemented /on/ parrot (and parrot ain't perl6); and yes, rakudo implements a lot more of the spec. 20:00
thraidh as i understand the docs, with parrot you can use: proto 'infix:+' or so
nqp has somewhat more magic with <EXPR> 20:01
[Coke] thraidh: you can't use that "with parrot", no.
thraidh but with rakudo nothing seems to work...
are the docs outdated then?
[Coke] what docs are you reading? 20:02
thraidh en.wikibooks.org/wiki/Parrot_Virtua...xpressions
for example
[Coke] wonders who wrote that. 20:03
20:03 lisppaste3 joined, ash_ joined
thraidh or this: www.parrotblog.org/2008/03/episode-...dence.html 20:03
[Coke] the blog entry is out of date, yes. 20:04
thraidh then there is squaak 20:05
which comes with parrot 2.3.0
[Coke] I'm afraid the original page doesn't have a lot of context showing where those proto's are meant to be put. they're certainly not valid PIR - they might be intended for nqp-rx. (And there's a typo in them, so they're not going to work anyway.) 20:06
thraidh which uses the techniques described
[Coke] I don't know if squak has been updated to use nqp-rx instead of the old style PGE. 20:07
diakopter [Coke]: looks like whiteknight added the initial version of that pag
thraidh i accepted that "is optable" and "proto 'infix:+'" are not really current anymore
[Coke] thraidh: - are you looking for "how to write perl6" or "how to write a language targeting parrot" ? 20:08
thraidh but then there is abc from the parrot examples which uses nqp-rx
actually i want to write some sort of interpreter in perl6
[Coke] ok. then I would avoid anything specifically targeting parrot as that will likely just confuse the issue.
thraidh yes 20:09
[Coke] and just concentrate on rakudo and the SYN.
thraidh i've recognized that by now
[Coke] best Perl6 docs are probably off of perlcabal.org/syn/ 20:10
thraidh according to S05 there is no <EXPR>
TimToady EXPR is just a method 20:11
masak S05 doesn't say there is no EXPR :)
[Coke] best sample code I can think of is something like github.com/partcl/partcl-nqp , but that's written primarily in nqp-rx, which is going to have a lot of callouts to parrot.
thraidh is it unique to nqp?
TimToady it's just that any method on Cursor or its derivatives can act like a rule
masak no, it's a rule in the Perl 6 grammar.
thraidh or can i use EXPR with rakudo?
masak it's the root of the operator parser rules.
thraidh i failed so far to do this 20:12
TimToady it's a bottom-up parser pretending to be a normal rule
thraidh i figured that much...
TimToady but if you want one in your own grammar, you have to write/steal one
20:12 meppl left
TimToady or derive from a language that has one 20:12
20:12 iblechbot joined
thraidh how can i steal the one from nqp? 20:13
TimToady I don't know if nqp has one or not
masak thraidh: this piece of code might interest you: github.com/masak/gge/blob/master/ex...es/algebra
thraidh: note how it uses &Perl6::Grammar::number on line 28.
you probably want something like that. 20:14
TimToady that is rather non-spec though
masak it worked when I wrote it :( 20:15
TimToady the name of the Perl 6 grammar is unlikely to be Perl6::Grammar, is what I mean
masak ah, right. 20:16
diakopter They call me Jane ... that's not my name.
masak well, at least Rakudo alpha had it that way, so I used it.
thraidh what i'm actually trying to do is writing some code which can understand latex-equations and produce the corresponding error-equation 20:17
TimToady but also there's no such thing as an optable in standard Perl 6 grammars
masak thraidh: sounds cool.
thraidh: are you using tests? I would.
thraidh masak: tests for what? my own code? or other peoples tests to understand how perl6 works? 20:18
diakopter heh
masak thraidh: both are a good idea, but I meant the former.
thraidh: the spectests might actually be helpful in understanding how rules and regexes work. 20:19
thraidh masak: right now i'm still struggling with building a latex-equation parser
masak right, but see, that's what I'm talking about :) 20:20
thraidh masak: i'm not getting it to compile (since i tried to use the nqp-EXPR)
masak in order to struggle less, it's useful to write it in very tiny increments.
that's how I end up writing grammars, anyway.
usually if I take too large a leap, I'm punished by ending up with something that doesn't parse anything. 20:21
thraidh right... but that's step two, isn't it?
masak no, I don't think it is.
thraidh tests are good, as soon as you begin to get any results, whether correct or not 20:22
masak starting with almost nothing, building up more and more of what you want in the grammar... should be step one.
and tests are just a nice way during that step to make sure you don't regress.
thraidh yes
i agree
masak it all comes down to running something like the tests either manually or automatically.
and then you might as well do it automatically. :)
besides, you'll get a medal. 20:23
diakopter does latex have some other formal grammar?
thraidh not really
diakopter oh; it's dynamic
thraidh only tex can parse tex (or latex) 20:24
diakopter like, symbols can have their meaning changed, new commands added; neat!
thraidh but i do not really need to understand all of latex
just very simple things
masak diakopter: sounds familiar... :)
diakopter thraidh: pauillac.inria.fr/~maranget/hevea/ 20:25
masak thraidh: anyway; welcome to #perl6! hope you'll enjoy it here. you seem to have gotten off to a running start. :) 20:27
thraidh thank you 20:28
by the way: my native language is not english, so i may misunderstand things or produce awkward sentences... i apologize for that 20:29
masak seems to work fine so far. what is your native language?
thraidh diakopter: that site seems to be something with a different quality 20:30
masak: i'm from germany
masak cool. moritz_ is from there too.
and Su-Shee. and probably others.
thraidh germany isn't that small ;) 20:31
masak so I've heard. :)
thraidh anyway, back to my original question 20:32
it seems that i either need to implement all the precedence-stuff manually
or steal nqp's <EXPR>
(but i have no clue how)
masak thraidh: I have an operator precedence parser that you can borrow if you want. 20:33
thraidh or you your parser
masak github.com/masak/gge/blob/master/li...OPTable.pm
[Coke] masak: but... isn't this something he can just do in p6? certainly I didn't write my open operator parser for partcl. 20:34
masak it's a port of PGE's optable parser. PGE used to be what parsed the Rakudo source and regexes.
[Coke] s/open/own/
masak [Coke]: well, you *can* do this with grammars, too.
[Coke]: but I maintain that sometimes doing it with an OPTable parser is shorter and more economical.
I haven't had much traction for that idea. there's no optable parser exposed by the spec, as TimToady pointed out. 20:35
thraidh but then you would need to do something like: rule add_expr { <mul_expr> '+' <mul_expr> } rule mul_expr { <term> '*' <term> } ...
masak oh, and optable parsers can sometimes be faster, too. they're usually preferable in expression-like situations. 20:36
thraidh: yes, exactly.
it's a bit verbose, since the only things you care about are the '+' and the '*'.
thraidh i wanted to avoid that...
[Coke] ooh, perl6 segfault (just try to run github.com/partcl/partcl-nqp/blob/m...Grammar.pm with it.) 20:37
masak thraidh: you can take a look at github.com/masak/gge/blob/master/t/03-optable.t to see what it can do. all those tests but one pass.
[Coke] (rakudo, sorry) 20:38
20:41 sahadev1 joined
[Coke] opens trac.parrot.org/parrot/ticket/1634 to track that. 20:42
thraidh hmm... it seems my project might be more difficult than i anticipated 20:43
masak if that's due to any shortcoming with Perl 6, please feel free to suggest improvements. 20:44
ash_ [Coke]: a real segfault? I haven't gotten a segfault in parrot in a long long time...
[Coke] ash_: I got 2 today.
but it's been a while. =-) 20:45
thraidh i'd like nqp's <EXPR> in rakudo ;)
20:45 kel_ joined
[Coke] ash_: feel free to see if you can duplicate TT #1634. (or reduce the size of the .pm needed to generate it.) 20:45
thraidh maybe as a package, since it does probably not belong into the core 20:46
masak thraidh: use EXPR:from<nqp>; # something like that, then 20:47
thraidh masak: will that work? or is that proposed syntax?
masak thraidh: it will work eventually. Perl 6 is meant to have that syntax, yes. 20:48
20:48 kel left, dju left, dju joined
thraidh masak: until then i'll try to integrate your optable 20:49
masak thraidh: let me know how it goes. you'll need the alpha branch of Rakudo, by the way.
masak goes to bed; early morning at the Russian consulate tomorrow
20:50 masak left
thraidh argh... 20:50
nothing is ever easy...
20:52 pmurias joined 20:57 dju left, dju joined 20:58 dju left 20:59 dju joined
[Coke] perl6: grammar a; 21:05
p6eval elf 30621, pugs, rakudo 7e0aa3: ( no output )
[Coke] rakudo: grammar a; say "ok"; 21:06
p6eval rakudo 7e0aa3: OUTPUT«ok␤»
[Coke] wtf.
for me, that's segfaulting rakudo.
21:06 kel_ left 21:08 bubaflub left
thraidh echo 'grammar a; say "ok";' | rakudo/rakudo/perl6 21:08
[Coke]: works fine for me
[Coke] guessing that it's the parrot version that's the issue. 21:09
thraidh 2.3.0-devel here 21:10
is it forbidden to use comments in grammars? 21:16
21:16 molaf left
thraidh (i know it's not...) 21:16
21:16 pmurias left
thraidh i got a segfault when i commented on rule... it's not reproducible... 21:17
21:26 gbacon left 21:27 wknight8111 joined
thraidh how can i print all methods/properties of an object? 21:29
jnthn $obj.^methods
Gives you a list of all the methods 21:30
Or if you want just the ones in the most derived class, $obj.^methods(:local)
21:30 hercynium left
jnthn rakudo: class Foo { method a { }; method b { }; }; for Foo.^methods { .name.say } 21:30
p6eval rakudo 7e0aa3: OUTPUT«a␤b␤first␤minmax␤does␤grep␤join␤values␤can␤Numeric␤elems␤end␤reduce␤Str␤keys␤reverse␤uniq␤isa␤map␤classify␤pairs␤kv␤ACCEPTS␤min␤max␤pick␤BUILDALL␤new␤Bool␤say␤print␤defined␤WALK␤item␤BUILD␤REJECTS␤clone␤WHICH␤perl␤CREATE␤Capture␤PARROT␤bless␤WHENCE␤WHERE␤list␤␤notdef␤»
jnthn rakudo: class Foo { method a { }; method b { }; }; for Foo.^methods(:local) { .name.say }
21:30 hercynium joined
p6eval rakudo 7e0aa3: OUTPUT«a␤b␤» 21:30
21:31 Guest41966 left
thraidh jnthn: thanks 21:31
rakudo: "hi" ~~ /hi/; $/ 21:33
p6eval rakudo 7e0aa3: ( no output )
21:33 dakkar left
thraidh rakudo: "hi" ~~ /hi/; $/.WHAT 21:33
p6eval rakudo 7e0aa3: ( no output )
thraidh rakudo: "hi" ~~ /hi/; say $/.WHAT
p6eval rakudo 7e0aa3: OUTPUT«Regex::Match()␤»
thraidh rakudo: "hi" ~~ /hi/; say $/.^methods 21:34
p6eval rakudo 7e0aa3: OUTPUT«Method 'methods' not found for invocant of class 'P6metaclass'␤current instr.: '!dispatch_.^' pc 408 (src/glue/dispatch.pir:99)␤»
thraidh hrmpf
moritz_ oh hai 21:37
thraidh: in Rakudo, $/ is not yet a Perl 6 object, so many things (includiing introspection) don't work on it 21:38
thraidh oh
moritz_: thanks
moritz_ rakudo: 'a' ~~ /a/; say $/.WHAT 21:39
p6eval rakudo 7e0aa3: OUTPUT«Regex::Match()␤»
moritz_ should just be Match
rakudo: 'a' ~~ /a/; say $/ ~~ Any
p6eval rakudo 7e0aa3: OUTPUT«1␤»
21:40 gbacon joined
thraidh moritz_: i can access named rules with $<name> or $/<name> (i heard) 21:41
21:41 ash_ left
moritz_ yes, that works in Rakudo 21:41
thraidh moritz_: can i get a list of available names?
moritz_ rakudo: 'a' ~~ /a/; say $/.keys
p6eval rakudo 7e0aa3: OUTPUT«Method 'keys' not found for invocant of class 'Regex;Match'␤current instr.: '_block14' pc 29 (EVAL_1:0)␤»
moritz_ that's one of the limitations
alpha: 'ab' ~~ /<alpha>(.)/; say $/.caps.join(', ') 21:42
p6eval alpha 30e0ed: OUTPUT«alpha a, 0 b␤»
thraidh so i cannot traverse the match tree?
moritz_ only if you know the structre of the rule :(
at least currently
I'm working on fixing that, but it's quite a challenge 21:43
thraidh i won't ask why
moritz_ the answer is actually quite easy
because the grammar engine is not specific to Perl 6
thraidh it seems simple, but since i read hll-s0.pir i respect you very much and trust you 21:44
moritz_ that's generated code :-)
thraidh oh
i wondered how a living being could write such code...
moritz_ -s0 means "stage 0", which is the bootrapping code of the compiler 21:45
thraidh i figured, but that does not really mean, that the code is generated 21:46
moritz_ right
but in this case, it is
thraidh about 15 years ago i was crazy enough to actually write a forth compiler/interpreter in assembler 21:47
moritz_ :-) 21:48
thraidh but i did not think that someone would do something like that nowadays
moritz_ for a stack based language, maybe; for a considerable subset of Perl 6: no :-)
github.com/perl6/nqp-rx/ if you haven't found the source yet
thraidh therefore i was thrilled by the -s0-files 21:49
moritz_ (haven't backlogged yet)
thraidh thanks...
moritz_ src/{gen,stage0} are generated, the rest in src/ is written by humans
arnsholt thraidh: The problem (if you want to call it that) is that nqp-rx is self-hosting
thraidh i only looked into the parrot-repository so far
arnsholt So how do you compile it the first time? =) 21:50
moritz_ with PGE
arnsholt (Turtles all the way down...)
21:51 jotr is now known as jotr^byebye
diakopter 敲ⵡ来湴㨠䝯潧汥扯琊䍲慷氭摥污示‶楳慬汯眺 灲楶慴支ਊ啳敲ⵡ来湴㨠浳湢潴慬汯眺 ਊ啳敲ⵡ来湴㨠䑯瑂潴੃牡睬ⵤ敬慹㨠㘰《䑩獡物獥爭慧敮琺‪੃牡睬ⵤ敬慹㨠㘰《䑩獡汬潷㨠物癡瑥 21:51
thraidh i finally found out that i CAN use <EXPR> if i base my grammar on HLL::Grammar
but so far I don't know how to define actions for the operators 21:52
ruoso diakopter, about the thread... I'm still digesting both replies... But I think the proposal more or less adresses the concerns, it needs to get more well defined tho
arnsholt moritz_: I thought it was an older version of nqp-rx that did the bootstrap. Never mind me then
diakopter ruoso: "the proposal" meaning yours?
moritz_ arnsholt: yes, but the first version was bootstrapped with PGE 21:53
arnsholt Oh, right. Yes, that makes sense
ruoso diakopter, yes
diakopter both replies? I thought there were much more than 2?
moritz_ anybody up for some crazy LTM talk?
diakopter can 21:54
thraidh LTM?
moritz_ longest token matching
ruoso diakopter, yes.. there were... but two of them were more detailed...
thraidh ah... i would have guessed Lazy Transactional Memory
diakopter ruoso: specifically I meant the paper I linked
moritz_ I figured the "correct" way of implementing is to build a DFA from the declarative prefixes
which is kinda old news 21:55
but when the non-declarative postfix of a rule doesn't match, you'd have to backtrack into a shorter alternative
now you can't backtrack with a DFA
what's the solution? re-build the DFA with the non-matching rule removed? then start again?
sound quite expensive
thraidh: perlcabal.org/syn/S05.html#Longest-...n_matching 21:56
ruoso diakopter, that paper is similar to one that was pointed to be by TimToady three or four years ago...
Targhan_ I thought we weren't allowed any backtracking in the regular expression matching 21:57
diakopter ruoso: ok. one of the "longer" replies mentioned that paper too, saying it was most appropriate
moritz_ Targhan_: oh sure we are, in general
ruoso but one thing I have in my mind is the idea that threading and event-based-programming would be better if implicit in Perl 6, via feeds, lazy lists and so on
diakopter ruoso: did you read it? the title is similar, but I didn't think the content was similar 21:58
ruoso didn't fully read it, but skimmed through
21:59 hugme left
ruoso diakopter, but I'm still digesting all the replies, and will still take a more careful read on all the quoted papers 21:59
what I mean is that I share the concerns raised, and to the extent I can see, the model address those 22:00
diakopter ruoso: the long message this morning said he was preparing a detailed proposal
22:00 PerlJam left
ruoso yes, I'm waiting for that one too.. 22:00
diakopter afk&
ruoso decommute &
22:01 dalek left, Juerd left, Juerd joined, pmichaud left 22:02 pugssvn left, pmichaud joined, pugssvn joined
thraidh as far as i understand, LTM only considers (more or less) fixed text, correct? 22:04
moritz_ nope; it considers traditional regular expressions
thraidh but it "breaks" as soon as <ws> is found 22:05
moritz_ yes, because the default <ws> contains backtracking control
22:05 PerlJam joined
moritz_ which isn't part of REG 22:06
anyway, bed time; see you soon
thraidh gute nacht 22:07
22:07 ruoso left
thraidh traditional REG is anything you can do with a DFA? 22:09
22:09 hercynium left, hercynium joined 22:11 iblechbot left 22:14 estrabd left 22:39 Limbic_Region joined
thraidh rakudo: say [+] (1,2,3) 22:39
p6eval rakudo 7e0aa3: OUTPUT«6␤»
thraidh rakudo: say [+] [1,2,3]
p6eval rakudo 7e0aa3: OUTPUT«6␤»
thraidh rakudo: my @x=[1,2,3]; say [+] @x
p6eval rakudo 7e0aa3: OUTPUT«1 2 3␤»
thraidh rakudo: my @x=(1,2,3); say [+] @x 22:40
p6eval rakudo 7e0aa3: OUTPUT«6␤»
thraidh rakudo: my @x=[1,2,3]; say [+] @@x
p6eval rakudo 7e0aa3: OUTPUT«Unable to parse postcircumfix:sym<[ ]>, couldn't find final ']' at line 11␤current instr.: 'perl6;Regex;Cursor;FAILGOAL' pc 1696 (ext/nqp-rx/src/stage0/Regex-s0.pir:932)␤»
sjohnson perl6!!
22:40 clintongormley left
thraidh perl6: my @x=[1,2,3]; say [+] @@x 22:41
p6eval rakudo 7e0aa3: OUTPUT«Unable to parse postcircumfix:sym<[ ]>, couldn't find final ']' at line 11␤current instr.: 'perl6;Regex;Cursor;FAILGOAL' pc 1696 (ext/nqp-rx/src/stage0/Regex-s0.pir:932)␤»
..pugs: OUTPUT«*** ␤ Unexpected end of input␤ expecting "::"␤ Variable "@@x" requires predeclaration or explicit package name␤ at /tmp/pZlyAgb7Gt line 1, column 27␤»
..elf 30621: OUTPUT«/home/p6eval/pugs/misc/STD_red/match.rb:117:in `block in to_dump0': undefined method `to_dump0' for nil:NilClass (NoMethodError)␤ from /home/p6eval/pugs/misc/STD_red/match.rb:117:in `map'␤ from /home/p6eval/pugs/misc/STD_red/match.rb:117:in `to_dump0'␤ from
../home/p6eval/pugs/misc/S…
thraidh perl6: my @x=[1,2,3]; say [+] @x
p6eval pugs, rakudo 7e0aa3: OUTPUT«1 2 3␤»
..elf 30621: OUTPUT«/home/p6eval/pugs/misc/STD_red/match.rb:117:in `block in to_dump0': undefined method `to_dump0' for nil:NilClass (NoMethodError)␤ from /home/p6eval/pugs/misc/STD_red/match.rb:117:in `map'␤ from /home/p6eval/pugs/misc/STD_red/match.rb:117:in `to_dump0'␤ from
../home/p6eval/pugs/misc/S…
thraidh perl6: my @x=(1,2,3); my $n=[+] @x; say $n; 22:43
p6eval pugs, rakudo 7e0aa3: OUTPUT«6␤»
..elf 30621: OUTPUT«/home/p6eval/pugs/misc/STD_red/match.rb:117:in `block in to_dump0': undefined method `to_dump0' for nil:NilClass (NoMethodError)␤ from /home/p6eval/pugs/misc/STD_red/match.rb:117:in `map'␤ from /home/p6eval/pugs/misc/STD_red/match.rb:117:in `to_dump0'␤ from
../home/p6eval/pugs/misc/S…
22:50 eternaleye joined 22:58 Psyche^ joined, eternaleye left 23:00 skangas left 23:02 Patterner left, Psyche^ is now known as Patterner 23:10 _buno_ joined 23:11 _buno_ left, snarkyboojum joined, kensanata joined 23:12 eternaleye joined
thraidh perl6: "a" ~~ /a { make "b" }/; say $(); 23:14
p6eval pugs: OUTPUT«Error eval perl5: "if (!$INC{'Pugs/Runtime/Match/HsBridge.pm'}) {␤ unshift @INC, '/home/p6eval/.cabal/share/Pugs-6.2.13.14/blib6/pugs/perl5/lib';␤ eval q[require 'Pugs/Runtime/Match/HsBridge.pm'] or die $@;␤}␤'Pugs::Runtime::Match::HsBridge'␤"␤*** '<HANDLE>' trapped by operat…
..elf 30621: OUTPUT«Undefined subroutine &GLOBAL::circumfix__36_32_41 called at (eval 125) line 4.␤ at ./elf_h line 5881␤»
..rakudo 7e0aa3: OUTPUT«␤»
23:18 kensanata left 23:22 masonkramer left, masonkramer joined 23:23 masonkramer left
lue I find it amazing that the guys of Diaspora* love Ruby because the creator allows you to do something in more than one way, that his way isn't the only way. 23:24
Seem familiar?... 23:25
[and hi]
Tene hi lue
lue
.oO(the new wikipedia logo is nice)
 23:27
thraidh hi lue 23:28
does anyone know why <<"a" ~~ /a { make "b" }/; say $();>> does not return "b"?
23:30 cdarroch left
jnthn thraidh: I think that's syntax from a while back that changed 23:31
say $/.ast is the new way.
thraidh perl6: "a" ~~ /a { make "b" }/; say $/.ast
p6eval elf 30621: OUTPUT«Global symbol "$_47" requires explicit package name at (eval 125) line 4.␤ at ./elf_h line 5881␤»
..pugs: OUTPUT«Error eval perl5: "if (!$INC{'Pugs/Runtime/Match/HsBridge.pm'}) {␤ unshift @INC, '/home/p6eval/.cabal/share/Pugs-6.2.13.14/blib6/pugs/perl5/lib';␤ eval q[require 'Pugs/Runtime/Match/HsBridge.pm'] or die $@;␤}␤'Pugs::Runtime::Match::HsBridge'␤"␤*** '<HANDLE>' trapped by operat…
..rakudo 7e0aa3: OUTPUT«␤»
23:32 rgrau left
thraidh hmm... that should have worked, shouldn't it? 23:33
jnthn I think so. 23:34
rakudo: "a" ~~ /a { make "b" }/; say $/
p6eval rakudo 7e0aa3: OUTPUT«a␤»
Tene rakudo: "a" ~~ /a { make "b" }/; say $/.perl 23:35
p6eval rakudo 7e0aa3: OUTPUT«Method 'perl' not found for invocant of class 'Regex;Match'␤current instr.: '_block14' pc 29 (EVAL_1:0)␤»
jnthn Yeah, certainly shoulda.
nqp: my $m = "a" ~~ /a { make "b" }/; say($m.ast);
p6eval nqp: OUTPUT«Assignment ("=") not supported in NQP, use ":=" instead at line 1, near " \"a\" ~~ /a"␤current instr.: 'parrot;HLL;Grammar;panic' pc 500 (src/cheats/hll-grammar.pir:197)␤»
jnthn nqp: my $m := "a" ~~ /a { make "b" }/; say($m.ast);
p6eval nqp: OUTPUT«␤»
jnthn nqp: my $m := "a" ~~ /a { make "b" }/; say($m); 23:36
p6eval nqp: OUTPUT«a␤»
jnthn ah...
Not Rakudo specific though.
Seems to be in the regex engine.
thraidh rakudo: grammar g { rule TOP { a } } class a { method TOP($/) { make "b" } } say g.parse("a").ast 23:37
p6eval rakudo 7e0aa3: OUTPUT«Confused at line 11, near "grammar g "␤current instr.: 'perl6;HLL;Grammar;panic' pc 501 (ext/nqp-rx/src/stage0/HLL-s0.pir:327)␤»
thraidh rakudo: grammar g { rule TOP { a }; }; class a { method TOP($/) { make "b" }; }; say g.parse("a").ast 23:38
p6eval rakudo 7e0aa3: OUTPUT«␤»
thraidh rakudo: grammar g { rule TOP { a }; }; class a { method TOP($/) { make "b" }; }; say g.parse("a", :actions(a)).ast
p6eval rakudo 7e0aa3: OUTPUT«b␤»
thraidh THAT works
rakudo: grammar g { rule TOP { a { make "c" } }; }; class a { method TOP($/) { make "b" }; }; say g.parse("a").ast 23:39
p6eval rakudo 7e0aa3: OUTPUT«␤»
thraidh that should work, too (i think)
i'll go to bed now... good night 23:48
lue good night 23:49
23:53 thraidh left
lue afk 23:58

Powered by Cro and the Raku® Programming Language.

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

Thank you!