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 GammaRay on 31 December 2005.
00:13 putter joined
putter ping? anyone about? 00:13
00:14 cdpruden joined
scrottie raises an eyebrow 00:15
putter :) If we define both multi f(Num $n){$n+1} and multi f(Num $n){$n+2} in the same file, f(10) will call the first one defined. But if they are defined in separate files and use()ed, the same one gets picked regardless of use() order. I'd very much like a way to get the order-dependence with separate files. Thoughts? 00:20
rent is now known as _brentp 00:21
putter Basically, between S29, PIL2JS, PILn etal, Prelude.pm, Prim.hs, and some PILRun, we have most of the pieces to create a unified p6 defined mostly in p6, with backend specific bits. 00:23
But to pull it together, one needs to be able to override multis.
scrottie Subclass?
Interesting. I should have you tell me about S29, PIL2JS, PILn etal, Preclude.pm, Prim.hs, and PILRun. 00:24
putter good thought, but since the objective is to make, say, Array::splice be something specific, the best we could do with subclasses is to have an "upside-down" tree of implementation choices above Array, and then use a single point macro switch to choose between say ArrayOnJS and ArrayOnHs. 00:26
scrottie I haven't been following Pugs. Guess I should go back and read the blog, so "RTFM" is a valid reply. Granting that, what's PILRun? 00:28
putter or even just JS.pm: macro Array () { "ArrayOnJS" } . hmm... there's something to be said for that... but one really wants selective import with renaming...
scrottie I personally like to live by the rule that if two things should be interchangable, they should present the same interface and inherit from the same abstract root class. 00:30
If you can take shortcuts to achieve that, great. Otherwise, do it the long way. What might start as a lot of reptition gives flexibility in the future as the implemenations diverge. 00:31
putter S29 is docs/AES/S29draft.pod which has signatures and some code. PIL2JS is the JavaScript backend, which has perl5/PIL2JS/lib6/Prelude/JS/ , PILn is src/PIL/Native/Bootstrap/ and some of the other stuff stevan and audreyt have been doing, PILRun is the currently inactive P5 backend, for which 00:34
I have some p6 code which got posted and discussed, but I don't think ever got checked in.
scrottie nods 00:36
thank you. intersting.
putter You know, a class/role-based system might just work. On say the JS backend, Foo is FooOnJS (eg, Array is ArrayOnJS). ArrayOnJS is ArrayGeneric, ArrayJSMods; With roles, the role body would need a bunch of method stubs pointing to Mods, as roles are less tolerant of conflicts than classes. But that should be fine... 00:42
The rationale for ArrayOnJS, rather than simply calling it Array on the JS backend, is... hmm... the thought was it would be more easy to do mutant implementations, eg, JS backend prime, Array is ArrayOnJSprime; ArrayOnJSprime is ArrayOnJS { prime mods... }. This argument is weakened by having a good mop, but... 00:46
00:50 cdpruden joined
putter But no, one would just want a set of rules, sucked in to classes. Unfortunately, to have use a role with method m_variant1 and m_variant2, and the class says multi m() { m_variant2() }, you would also need a companion "blocker" role, with stubs for m_variant1/2, for the sole purpose of 00:50
creating a conflict so m_variant1/2 are _not_ inherited into the class. I don't think we currently have a mechanism of otherwise blocking inheritance. Do able, but a bit ugly. 00:51
00:52 daja77_ joined
scrottie It's fine to create subclasses that implement operations as a no-op. It doesn't hold up to OO modeling to willy nilly add or remove methods (unless you get the P6/SmallTalk behavior of black-holing unknown methods without raising an error). 00:53
putter And it doesnt work so well when you want to touch multiple classes with a single decision. Eg, role do_things_way_1 { class A { m_variant1 } class B { m_variant1 } } It would be *really* nice to have a set algebra over methods, but we currently dont... 00:54
scrottie Planning implementation is NP-complete. Do it badly then improve the implementation steadily. Refactor. Get to work. 00:58
putter re no-op, the challenge is one wants to select and merge sets of methods across multiple classes. currently it seems one can't do that (?). if overriding stayed order dependent in use(), then packages/modules could be used as the "sets". The algebra would be a bit weak, but I think we could live with 00:59
01:00 scook0 joined
putter that, at least for now. Oh, not so weak if use() permitted selective import(). 01:01
so use ArrayGeneric; use ArrayOnJS :limited-memory-version; use ArrayRandomTweaks; 01:02
each one overriding some of the methods defined by the previous.
oh, better use ContainersGeneric; use ContainersOnJS :limited-memory; use ContainerTweaks13; 01:04
scrottie Eh? Select and merge sets of methods? That's an inheritance tree. If you have two JavaScript backends, put the common logic in one class, then subclass it twice, each subclass containing the specifics of the particular JavaScript backend.
putter while one might manage the Array version with subclassing, the container version would seem to require mondo hoop jumpage.
scrottie putter: here's a really good book you might enjoy: _Object Oriented Design Heuristics_ 01:05
stevan putter: I am having trouble following what you mean
do you want to have a p6 Array, which can have multiple "backends"? 01:06
one for JS, for p5, etc etc etc
putter peers blearily at the screen, trying to marshal thoughts... brain.yaml |> grep...
stevan Array should be a role
which will define the implementation as completely as possible
if need be in a circular way
then the runtime will implement the ^Array class and install it into the global scope/package 01:07
but 9/10th off what is Array can be defined in the Array role
every runtime will end up having a slightly tweaked version of the post Metamodel bootstrap 01:08
to address each models particular issues/concerns
however, we /might/ be able to put all this into PIL^N, in which case this is all a moot point since the runtime will only need to implement PIL^N and everything else Just Works 01:09
thats the whole point behind using the DSL, is that it increases portability by decreasing the amount of duplicated code needed
putter ok. so role rArray does rArrayGeneric, rArrayOnJS, rArrayTweaks {}; rArrayGeneric defines most everything, so most everything in the other roles generate conflicts, which have to be explicity resolved in rArray. I suppose a macro could help. So... 01:10
stevan this was the leason learned from yours and flgocks effort with PILRun and iblech's PILJS
no 01:11
putter: role Array {}
it does nothing
(maybe List, but I am not sure)
putter re piln, hmm...
01:11 oylenshpeegul joined
stevan in the JS runtime, the missing bits of the Array role are implemented in the Array class itself 01:11
theorbtwo DSL? 01:12
stevan but again, ideally this all lives in PIL^N and JS runtime need never know about it cause the JS implements PIL^N
theorbtwo We need a dictionary, or a bot.
stevan Domain Specific Language :)
putter: my guess is that Array can be completely defined in PIL^N 01:13
maybe not as efficiently as we would like
putter re "does nothing", huh? role rAx { multi pop{...defined assuming splice...} multi splice {...defined assuming pop...}, role rAJS { multi splice {...defined using inline_js...} }. How do rArray and Array get defined? 01:14
stevan but then the next step would be to move the implementation down to the PIL^N runtime
putter: in the bootstrap
lets take a simpler example
role Eq { method infix:<==> (Eq $left, Eq $right) { !($left != $right) } method infix:<!=> (Eq $left, Eq $right) { !($left == $right) } } 01:16
so == is obviously the inverse of != and vice versa
putter I don't by "Array can be completely defined in PIL^N". I'm using Array as a strawman proxy for the entire Prelude. So that sentence becomes "the entire Prelude could be completely defined in PIL'N". so... yes, but that would be nuts. ;)
stevan this is a complete role, albeit one you should never "just use" since it will loop infinitly
putter: why can't Array be done in PIL^N? have you looked at the code ? 01:17
it is about 60-70% complete, although it currently is waiting on some p6l information
putter: my point is that with a circular prelude, you need only fill in the peices needed to prevent the infinite meta-circular spiral of death 01:18
putter yes, but take a look at perl5/PIL2JS/lib6/Prelude/JS/Array.pm... do you really want to write all of that by hand in PIL'N?
stevan putter: have you looked at the PIL^N version of Arrary? 01:19
putter yes
err, src/PIL/Native/Bootstrap/Container/Array.pil, yes?
stevan yes
ok, so a good portion of Array is implemented, but not all 01:20
putter wonders if we are having a "violent agreement"
stevan so in the Perl 6 prelude, you do this:
^Array.meta.add_method('missing-method', method { ... });
and fill in the parts we are missing
putter: we may be,... however we are speaking different languages/dialects so hard to tell :) 01:21
putter I agree the objective is a "circular prelude", though that term doesnt quite capture it. The question is how to manage the mixin-ing. If method overriding under use() was order sensitive, I'd be all set, and already writing it, sucking bits in from the previously mentioned sources. It wouldnt be ideal, but it would be fine for now. 01:22
stevan putter: thinks less of multis, and more about classes/roles 01:23
putter But it isnt order sensitive, so I dont know how to cleanly manage the assembly of conflicting sets of bits into a Prelude-de-jure.
theorbtwo ...by making it order-sensitive? 01:24
putter :) yeah...
stevan putter: the old prelude was a collection of multi subs, which would have this problem
it you move many of those "built-ins" into classes and roles, the problem becomes much smaller if not elimnated entierly
putter: also understand that with the MOP, we don't have to "finish" the class, it is not a one-shot-all-or-nothing thing 01:26
putter how eliminated? conflicts in roles have to be explicitly resolved. thus the "role rAx { multi pop{...defined assuming splice...} multi splice {...defined assuming pop...}, role rAJS { multi splice {...defined using inline_js...} }." example. Perhaps I dont understand your response of "no does, instead done by bootstrap"
stevan it can be built incrementally
where is this rAJS coming from? 01:27
putter They are both part of the great P6 on everything common code base.
stevan there should be no AJS, it should be either written in PIL^N or Perl6 01:28
the JS runtime will just run the compiled PIL^N
Perl6 compiles to PIL2, then PIL2 is compiled to PIL^N, then PIL^N is run on the VM 01:29
(or possibly PIL^N is compiled to PIR and run on Parrot)
of if someone writes the equivelent of all the PIL^N code in PIR, then PIL2 can be compiled to PIR directly 01:30
role Array { method pop { ... slice ... } method slice { ... pop ... }} can be defined in PIL^N
or Perl 6 01:31
then create the Array class in PIL^N, and then tie the knot (implement pop or slice)
putter the "only p6 or piln", ok, I don't currently buy that, so let me figure out why...
stevan writes a larger example,.. it is hard to communicate these ideas over IRC 01:32
putter basically you're saying the p6 would should be allowed to talk directly to the backend. anything they might ever want to talk about should be pushed into piln. That seems a recipe for a bloated piln... and unlikely to work. no? 01:35
s//not be allowed/
stevan putter: why does that bloat PILN? 01:36
theorbtwo The idea is that PIL^N is an unbloating filter. P6 is a big scary langauage. PIL^N should be a small simple language, easy to write an interpreter for or a compiler from. 01:38
stevan theorbtwo: exactly
theorbtwo Now, from the sample of pil^n I've seen I rather wonder if that's the case, but that's another question.
putter How about this... on backend 1, you want to use technique 1, on backend 2, tech 2. say preextending arrays is essential on 1, but contraindicated on 2. how do you deal? 01:40
no inline_foo vs piln issue. but you want above-piln p6 algo 1 on backend 1, and 2 on 2. 01:41
stevan putter: I dont understand
this is a problem for the PIL^N interpreter 01:42
pasteling "stevan" at 67.186.136.119 pasted "PIL^N & Perl 6" (30 lines, 814B) at sial.org/pbot/15402
putter one possible argument is 1-that's very unlikely, so 2-use mop if/when it comes up.
stevan putter please look at this paste
putter looking... 01:43
stevan the role defintion is circular since &wave_hands uses pop :)
the Array class then does the role Array
and defines the bit to break the cycle
it defines it using the PIL^N method `splice() 01:44
cycle broken
if it is faster in your PIL^N runtime to implement `pop than `splice
then do that and break the cycle there
then in Perl 6, we can add other methods as we like
make sense? 01:45
putter can one do if $variant == 1 { ...add_method... } else { ...add_method... } ?
stevan why?
in which language?
putter the language of your paste
stevan there are two languages there,.. PIL^N and Perl 6 01:46
putter either
stevan why?
where is the win?
putter lol
stevan what defines $variant
I understand the idea of a better algo in runtimeA so better to implement &pop there,.. 01:47
but why should that be in the same file for runtimeB where it is faster to implement &splice
01:47 jroc joined
stevan each runtime might need a runtime specific bootstrap file 01:48
which does this kind of stuff
01:48 luqui joined
stevan but it makes no sense to mix that all in one,.. you will very quickly get a lot of spaghetti 01:48
putter hmm... maybe this is the conceptual divergence: if everything was a simple cycle, and one just needed to define a little tidbit to break the cycle, then things would be simple. if the "simple cycle" model is too simple, and one really needs more general mixins, then... one needs richer control over how methods get defined.
stevan putter: everything in that paste can be spread out over multiple files if you want
luqui paste? 01:49
stevan sial.org/pbot/paste
putter is there a .replace_method to go with the .add_method?
luqui no, I mean the one you are talking about
putter or .remove_method?
stevan putter: no, just add_method again
yes
remove_method exists 01:50
luqui: sorry sial.org/pbot/15402
putter ok, that will do it.
stevan luqui: we are discussing the circular prelude idea
putter: is that all you needed :)
luqui ahh...
putter wish there was a way to just write it in p6 multi f(Num $x) :replace { $x + 13 }. but maybe the mop cruft can be macro'ed away. 01:51
stevan luqui: note that wave_hands uses pop,.. so it is cicular,.. I should have put that in the paste
luqui okay
stevan putter: the multi approach gets really messy after a while
better to organize into classes and roles first,..anda bare multis as a last resort 01:52
luqui Haskell does this with typeclasses
theorbtwo I'm worried by how complex pil^n seems to be.
luqui many times the methods in the class are defined in terms of each other
stevan luqui: exactly,..
luqui I've been caught off guard by that sometimes; it would be nice if it would tell me that there's still a circular definition
stevan theorbtwo: why do you think PILN is complex?
luqui: yes, that can be tricky I think
luqui it's undecidable in general :-( 01:53
you'd have to have some way to declare it
theorbtwo svn.openfoundry.org/pugs/src/PIL/Na.../Array.pil does not make it look like a terribly simple language. 01:54
I was expecting something closer to an assembley language.
luqui theorbtwo, it's very close to lambda calculus, not assembly
putter so basically what using the mop buys you is that when you use() two files, the expressions in the first file are run before the second, whereas the multi definitions in the second (or the first) may preempt the other, independently of use() order.
what a zoo. 01:55
stevan theorbtwo: it is a very simple language,. thats why things look so complex :)
putter: there is no need to bring use() into this 01:56
it is not relevant in this case
the MOP has nothing to do with the order of files loaded either
theorbtwo stevan: How do you put together the generic prelude with the specific bits for your runtime, then?
stevan theorbtwo: right now we have #include in PIL^N, its nothing fancy, but it works 01:57
putter: you need to get away from the multi defintion thing I think
putter the concrete motivation for this discussion is that a good task for this evening is to pour PIL2JS's and S29's and Prelude.pm, and other's p6 "Prelude" code into one big pile, grouped by pure-p6, p6-on-js-specific, etc. further partitioned by say different strategies for structuring the pure-p6. 01:58
and then use magic mixin method foo to pull backends, or at least their preludes, out of it.
stevan putter: the old PIL runtimes were hampered by the old PIL design, I saw you frustrated by it many times 01:59
I think some of that code is useful, but I am not sure all of it will be 02:00
luqui putter, see my recent blog on use.perl. It is not unrelated...
stevan but you are welcome to figure this out,.. I dont want to stop you on this
it will be useful i think
putter stevan points out "with piln, from the prelude's stand point, there is only one backend - the piln vm". I wonder if one will still need a way for p6 and underlying backend to chat, and more generally, whether a one-size fits all p6 prelude will actually fly. will one need per-underlying-backend prelude mods.
stevan putter: if perl 6 has a need to talk to the backend,.. the PILN/PIL2 design is broken 02:01
putter if so, will need some mod mixin mechanism. how much it gets used in part determines how (in)convenient we can get away with it being.
luqui stevan, it will for optimization purposes at least
putter resyncs with stream, backlogging...
luqui it's a good idea to allow communication between different modules 02:02
stevan luqui: well that is the compiler problem (theoretically of course)
luqui even when they're supposed to be independent
yep, theoretically
stevan remember,.. the plan is to first compile Perl 6 to PIL2
luqui oh, before I go further, what is pil^n's role?
stevan at which point there is no more perl 6,.. only PIL2 02:03
then PIL2 gets compiled to PIL^N, at which point there is no more PIL2, only PIL^N
then the VM runs the PIL^N
clean seperation
luqui ahh right
stevan ideal world,.. sure it is
luqui misunderstanding things
putter p6->pil2->piln. piln is a vm.
stevan but this is all for -Ofun
luqui what is the difference between PIL2 and PIL^n?
stevan PIL^N is the language of the VM 02:04
luqui putter, pil^n is still a repr language
you could have multiple different VMs that all run pil^n
stevan luqui: PIL2 has all the stuff Perl 6 needs that PIL^N does not provide ;)
luqui stevan, i.e. nothing
pil^n is lambda calculus, which is turing complete
putter pil2 is desugared p6. the "p6" compiler proper then chews on the pil2, emitting piln. i believe.
luqui as a repr language, it suffices for anything
02:05 revdiablo joined
stevan luqui: I really dunno,.. I havent seen PIL2 yet 02:05
luqui putter, okay, that seems to work for me
it's just another intermediate language to make things easier
stevan luqui: also PIL^N can be compiled down further
putter i wasnt around when the pil2/piln distinction came into existence, so I'm not entirely clear on motivation. one posiblity is that pil2 looks to be more flexibly tied to "what the compiler feels like dealing with today", whereas piln will get less flexible as n vm implementations get created. 02:06
stevan putter: you need to talk to audreyt for details on PIL2 02:07
past what I have already said that is
luqui To me it seems like: perl6 -> pil2 -> piln. pil2 is there to make the arrows simple
because a perl6 -> piln transformer would be complicated
stevan PIL^N should be complete at this point
luqui: exactly 02:08
luqui so suppose that I'm feeling crazy and hubristic and want to write a C backend to PIL^n
where would I start?
stevan the Haskell source ;) 02:09
luqui for example, where is the AST stored?
stevan luqui: what does <: mean in your use.perl post?
luqui subtype
stevan luqui: it is all in src/PIL/Native/*
luqui Dog <: Animal
stevan although the Repr types for the new objects are in PIL/Repr/* but we are not using those pending a response on p6l 02:10
luqui and what would you use instead? 02:11
luqui saw the p6l message, and wasn't quite sure how to respond
or even what kind of answer you would be looking for 02:12
stevan luqui: what is the $repr argument to CREATE for?
that is the crux of the question
luqui oh, about the use.perl post, most of the details were not related to the above discussion
just the "composable modules" bit
stevan is it for alternate implementation of an opaque type, all which do the same thing
luqui I expect that most of that is up to you guys 02:13
er, us guys
stevan luqui: and gals :P
luqui I'm using the default masculine 02:14
putter stevan: so, just to cross-check my understanding. your model is there are conceptually just two files, Prelude.piln, and Prelude.p6, plus a couple of piln-vm-on-platform-x backends, and that's what p6 is made of.
luqui which, as far as I know, is still in English (despite fights to the contrary)
stevan luqui: your offending my feminine side :P
luqui putter, for a proper definition of "file", it seems so
yeah, females like to get offended by that one in the US 02:15
stevan putter: yeah, for the most part
luqui for some reason it doesn't bother spanish females
stevan slaps luqui with his purse
luqui hey, no fair having keys in your purse! Those hurt!
stevan :P 02:16
putter: as I have stated many times before,.. this *might* work, it *might not* work
putter ok, well that's simple. finish yanking the code from S29, add in code from PIL2JS et al, create big Prelude.p6. Bits of which can be commented out as they get/are defined by "Prelude.piln". 02:17
stevan it feels sane in my head,.. but thats not saying much :P
putter ;)
luqui putter, barring modularization
stevan putter: yes, as I said,.. it is a worthwhile pursuit :)
luqui which Larry says he wants (hooray).
putter wonders if there is a "church of the simple"
luqui that is, you have modules of related functions, a certain set of which are imported by default
putter stevan: it's simple, and thus the Right Thing absent clear argument to the contrary. 02:18
luqui occam's razor?
stevan luqui: so, re: $repr,... if you cant answer, then I need Larry too :) 02:20
I dont think it is for us guys/girls to decide
luqui why? 02:21
putter hmm... especially if you have an Aristotelian model of software... "our task is to search out the perfect form of program x. that is naturally the simplest possible form. thus we apply occam..."
luqui to me, that seems like API design rather than language design
stevan luqui: let me explain the issue
luqui and for API design, we make it and double check its sanity with Larry, not ask Larry how to do it
stevan, explain away. maybe I don't understand fully
stevan if $repr is meant to be an storage type for the opaque object type 02:22
then a class which uses p6array will act the same, and be totally compatible with one which uses p6hash
the differences will be hidden by the opauqe type
BUT
if $repr is meant to provide something more akin to blessing different reference types in p5 02:23
then we have issues
luqui by more akin?
luqui goes over stevan's p6l message again 02:24
stevan package MyObjectArray; sub new { bless [] => $class } sub head { pop(@{shift})
package MyObjectHash; use base 'MyObjectArray'; sub new { bless {} => $class }
that wont work
cause they are incompatible repr types
luqui ah right 02:25
stevan :)
luqui back in a moment
stevan dances the opaque repr dance around luqui, every so often smacking him in the head with a lambda 02:26
02:26 robkinyon joined
stevan hey robkinyon 02:26
robkinyon heyla, stevan
luqui doesn't think that's a very cheerful dance
stevan stops dancing and gives the lambda to luqui so he can dance it this time 02:27
02:30 Cryptic_K joined
svnbot6 r8643 | putter++ | S29draft.pod: typo correction 02:30
luqui stevan, I'm stumpted on the $repr issue 02:43
stevan luqui: me too :) 02:44
luqui I don't think Larry has any idea what he's talking about there (he probably has some specific capability in his head, and tried to document a way to carry that out without telling us what the capability was... he does that)
stevan hence my asking p6l :P
luqui: thats what I figured too,.. and why I asked
I can only read and re-read s12 so many times before I hit myself over the head with lambdas 02:45
luqui hit yourself in the monads with a lambda 02:46
robkinyon is the issue just "How do I know the repr in order to interact with it?" 02:47
stevan has decided Perl is a waste of time and is switching to LISP
robkinyon seconds that notion and wonders if $CLIENT will go for it
dduncan took you long enough
robkinyon starts to read up on mod_lisp
stevan robkinyon: no, we can know the $repr, the issue is "is that what we want to do"
robkinyon huh? 02:48
stevan [21:23] stevanpackage MyObjectArray; sub new { bless [] => $class } sub head { pop(@{shift})
[21:23] stevanpackage MyObjectHash; use base 'MyObjectArray'; sub new { bless {} => $class }
robkinyon isn't that just promoting the same problem from p5 into p6? 02:49
stevan so the question is, are different $repr types in CREATE just alternate storage? or more like the p5 example above
robkinyon more importantly, doesn't YAGNI apply? 02:50
stevan robkinyon: yes, but I can't figure out what larry wants exactly,.. so I am stuck
S12 is not clear enough :)
luqui thinks the specific thing Larry had in mind might be interacting with p5 classes
robkinyon more specifically, you mean the interacting with a hashref that happens to be blessed as a hashref and not an object? 02:51
stevan luqui: yes, I think that too,.. which means we have alternate storage with the same low level access API
robkinyon has been thinking about releasing Ruby-style interfaces to Array and Hash by doing just that 02:52
luqui what are Ruby-style interfaces? 02:53
and how do they differ from Java-style interfaces, or how do they differ from mixins?
stevan interfaces in the style of Ruby :P
robkinyon no
stevan I think he means API,.
robkinyon my $array = Array->new;
$array->[3]
$array->flatten;
luqui yeah, what about that? 02:54
robkinyon the ruby API is damn useful
luqui you mean the methods that arrays provide by default?
robkinyon yeah
luqui okay, now I'm not sure what you mean by "releasing" 02:55
robkinyon of course, Perl won't let me do syntactically-sugared blocks like Ruby
release to CPAN
Ruby::Array
Ruby::Hash
Ruby::File
luqui oh oh oh!
got it
robkinyon it's an object AND an array and you're supposed to treat it as such
luqui you're talking about perl 5
robkinyon yeah
luqui now it makes sense
robkinyon but, i think the idea bears on this discussion 02:56
stevan *cough* this is #perl6 :P
luqui uh... not really
robkinyon because it's something that I think may be causing Larry to want alternate $repr
luqui probably not
robkinyon maybe not that specifically, but ...
luqui in perl 6, arrays already have methods
robkinyon yes
luqui everything can have methods
robkinyon because P6 arrays are self-aware
stevan luqui: and what is their $repr :)
luqui opaque 02:57
they're just things
stevan p6opaque?
robkinyon but, one of the methods they have is []
right?
stevan ducks to avoid luqui's wildly waving hands
luqui okay, it seems that I'm arguing that $repr doesn't make sense from an interface POV 02:58
02:58 vel__ joined
luqui it doesn't have any impact whatsoever 02:58
robkinyon i'm curious - does this mean that, in P6, I can define a class that implements both [] and {}?
luqui it's all just methods. That's what I meant by "opaque" and "just things"
robkinyon, you mean like Match? :-)
robkinyon Fair nuff
luqui robkinyon, you can in Perl 5 too
stevan luqui: right, that is ideal,.. but then why does it need to be in the perl6 level? isnt that an implementation detail? 02:59
robkinyon Contextual::Return, yes
luqui or more simply, "overload"
robkinyon The point is, luqui, that why does any $repr exist other than p6opaque?
If, as you say, everything is "just methods"
luqui I think we're trying to figure out what $repr is, rather than why multiple can exist
robkinyon it's a black box 03:00
stevan I think the questions are related
luqui stevan, and yes, I think it's kind of an implementation detail
stevan I can see the reason for multiple ones
luqui I think it's an API question
stevan but its an imple detail :)
luqui which is why I think it's still up to us to decide what it means
robkinyon one sec - i'm confused
Can I create a P5 hash that has no OO semantics in P6?
stevan luqui: how is it an API question,.. please expand
luqui robkinyon, by "OO semantics" you mean? 03:01
stevan he cant call methods on it
robkinyon Or, does the very act of creating a hash in P6 automatically mean that I can call methods on it?
luqui stevan, it's one of those things that is not fundamental about how the language works, but not entirely hidden from the language user's view
robkinyon If EVERY SINGLE hash in p6 has methods, then there is only the p6opaque $repr
luqui kind of a hook into however it is OO is implemented
robkinyon, I don't understand the reasoning for that conditional 03:02
$repr isn't about whether you can call methods on something
robkinyon no, it is
luqui I think it's about what the methods do inside themselves
stevan luqui: yes :)
luqui i.e. it doesn't matter in perl 5 whether something is a blessed hash or a blessed array
you can call methods on them both in the same way 03:03
robkinyon luqui: Where do I get the hashref for the alternate $repr?
luqui it's about how those methods interact with the innards of the object where $repr matters
robkinyon, I think I see what you're saying
but that means that all reprs are "built on" some p6opaque
see, I'm unclear on what exactly a repr is and how it differs from some other kind of object 03:04
(I think that's one of the big issues. The other big issue is everything else.)
stevan luqu: see doc/notes/something-with-repr-in-the-filename.pod
svnbot6 r8644 | putter++ | src/PIL/misc/S29code.pl - Created.
r8644 | putter++ | Extraction/definition of p6 code from/for S29draft.pod.
r8644 | putter++ | Some may be useful for fleshing out the new PILn backend.
r8644 | putter++ | perl5/PIL2JS/lib6/Prelude/JS/ is another such resource.
r8644 | putter++ | To contribute, just scroll down until pod lines start appearing, then continue the conversion to p6 code.
r8643 | putter++ | S29draft.pod: typo correction
luqui hmm..... it's not CREATE's decision
if you're using an array as a repr, the compiler better know that so it can translate $.foo into appropriate accesses into that array 03:05
stevan luqui: no, you tell CREATE which one you want,.. CREATE defaults to p6opaque
robkinyon That's why I've been babbling about the P5 Array impl.
luqui stevan, I know, but I don't think it should be
robkinyon because it all comes down to where the non-p6opaque thingies come from
because, it doesn't make any sense to bless an object into another class
luqui robkinyon, bootstrapping. you have to start somewhere.
robkinyon, it doesn't? 03:06
stevan ah,.. i see what robkinyon is saying
luqui explain
stevan if Array is an object,.. why use it as a $repr for opaque
p6array
it is already an object
luqui yeah, I gathered that 03:07
stevan which itself is a p6opaque
luqui and it's a bootstrapping issue
stevan spins around a few times to make the point
luqui what's Array's repr?
robkinyon p6opaque
stevan you said p6opaque
robkinyon that's the bootstrap
so, why am i using an object as the repr for another object?
luqui hmm.. right
stevan spins some more,.. it makes more sense when you are dizzy
robkinyon LOL 03:08
luqui robkinyon, maybe for alternate metaclass stuffs
robkinyon no
stevan luqui: no, alternate metaclass is handled in the MOP
robkinyon there is no reason because it comes down to "Why am I using an object to represent my object?"
stevan not releated
luqui robkinyon, because it's useful to do that sometimes? 03:09
robkinyon YAGNI
luqui I'm just thinking about the metaclass stuff
robkinyon YAGNI
putter contemplates the potential for an Eliza program to participate in #perl6 oo discussions... ;)
luqui maybe I'm using the wrong word
robkinyon, stop saying that!
robkinyon LOL
stevan tells me that on a daily basis
stevan luqui: metaclass stuff is just subclassing ^Class
robkinyon luqui: The point is that if you cannot come up with a counterexample, then YAGNI applies 03:10
stevan luqui: best thing is to yell YAGNI back at him,.. but other than that there is no cure
robkinyon it's put into the MAYBE_LATER section
luqui robkinyon, but as language designers, we must take a lot of time to try to come up with a counterexample
robkinyon Absolutely
i'm not saying that the idea has no merit
stevan YAGNI!
luqui and by yelling YAGNI, you're not really allowing that time
:-)
robkinyon No, I'm allowing that time to not interfere with moving forward 03:11
luqui robkinyon, but it's specced!
YAGNI doesn't apply unless we get Larry to agree
stevan squints hard at the spec,.. but its still fuzzy
robkinyon and three very intelligent people talking for almost 30min cannot come up with a reason for the spec
luqui we're interpreting the spec, like christians interpreting the bible, or supreme court justices interpreting the constitution
stevan wonders where those people are,.. and why they arent talking to us
robkinyon Unless the four very intelligent people that constitute @Larry can give us a reason, those of us implementing the spec have to say "Provide a failing testcase" 03:12
stevan luqui: and that kind of stuff has never ever ever caused trouble :P
robkinyon and move on
luqui robkinyon, but the discussion is not going in circles!
robkinyon really?
luqui It was moving along nicely until you interrupted it
I even had an insight that I was about to say
robkinyon we have all agreed that, fundamentally, everything is implemented in terms of p6opaque
luqui robkinyon, "agreed"?
no, I said I see your point 03:13
robkinyon robkinyon: so, why am i using an object as the repr for another object?
luqui: hmm.. right3.
luqui robkinyon, are you asking yourself?
robkinyon no - i'm quoting
luqui okay, I guess it's a limitation of IRC
it wasn't a, "oh you're right"
it was a "hmm, good point"
in any case, that doesn't matter 03:14
I see your point, but I don't necessarily agree
stevan wants to hear luqui's insight
luqui so let's say hypothetically that you want to build an object on top of an Array
robkinyon since all the reasons to use an alternate representation are moot in P6 given p6opaque and []/{} being methods, I fail to see what benefit there is
You would build an object having an array
luqui robkinyon, no, I mean translating member accesses to array lookups 03:15
robkinyon ok
translate member access into accessing the array that you contain as your sole memeber
luqui that is, I want the user to write: class Foo { has $.bar; method baz() { return $.bar } }
and I want to translate that internally into array accesses
internally is the key word
you have to think like a module writer for a second 03:16
robkinyon unless you're going to mix this in somehow due to some random serialization requirement, I can't see what the benefit is
luqui robkinyon, let's say it's interfacing with a C program
robkinyon or, you need to implement as an array in order to have sequential memory locations to do some sort of shared mem
that's fine
luqui which has memory layout requirements
robkinyon are you using the boxed or unboxed array?
luqui boxed
stevan Array
uppercase is boxed,.. lower is unboxed 03:17
robkinyon that doesn't conform to the memory layout requirements
luqui shut up
i'm just trying to get the point across
forget that I said Array
pretend I said CStruct
robkinyon ok ... is that implemented in terms of p6opaque?
luqui the point is, I have a great idea that has to do with laying out objects differently than perl 6 does by default
and yes, it is implemented in terms of p6opaque
but that does *not* mean that it is identical to p6opaque 03:18
robkinyon you want to have direct control over how your memory is laid out
now, i can see a benefit
i apologize if I pissed you off.
luqui it's okay... I just felt like you were getting too detailed when I was posing a thought experiment 03:19
anyway...
robkinyon well, it's kinda like the "rebuttal of relational theory" Ovid posted to Perlmonks a few weeks ago
luqui what I was thinking eariler is that the compiler has to know what layout you're using
justatheory robkinyon: Actually, Ovid is in favor of relational theory. That post was about the fact that SQL is not relational. 03:20
03:20 [Keith] joined
luqui either that or member accesses have to be certain kinds of method calls (an idea that I am not entirely opposed to :-) 03:20
stevan luqui: so you are saying that on the p6 level it makes no diff, but on the compiled level it does
robkinyon justatheory: My response was "Pick an example that's actually normalized"
luqui stevan, not exactly
justatheory robkinyon: My response is; Reate Date.
luqui If I want to implement this new object on top of CStruct, I'd like to do that all from perl 03:21
that is, it doesn't have anything to do with the writer of the class or the user of the class
but the implementor of the metaclass or something like that
stevan ducks the waving hands again
robkinyon luqui: May I try? 03:22
stevan metaclasses dont handle attribute access
luqui well you seem to have a pretty strong notion of what a metaclass is
stevan in intstances
luqui and apparently mine differs from yours in there
thus "or something like that"
stevan luqui: a metaclass is an instace of ^Class
luqui I thought that's what a class was
stevan luqui: it is what ^Class is an instance of
robkinyon luqui: what you're saying is that you want to be able to define a CStruct within the framework of class { ... } and have the compiler and the interpreter both grok you
stevan the "meta
the "meta" part is really optional
luqui metaclass is the class of ^Class
robkinyon, as long as I tell them how to grok me 03:23
stevan luqui: and turtles all the way down
03:23 [Keith] left
stevan luqui: until you stop it and make something an instance of itself 03:23
robkinyon Or, going further, you want to be able to hand something a C header file and it create a usable $repr for you
luqui stevan, what is an instance of itself in this metamodel?
robkinyon luqui: Don't even go there
luqui robkinyon, why not? 03:24
robkinyon it's not needed for you to understand the turtles! :-)
stevan luqui: ^Class is an instance of ^Class
stevan thinks robkinyon had a bad turtle experince as a child
luqui okay, so we're ruling out alternate representations
by the current model
robkinyon there is much blowing of brain gaskets when you get stevan talking about how ^Object is an instance of ^Class
no ... you have a good reason for why alternate representations are useful
stevan and ^Class is an instance of ^Class too :)
luqui but in any case, if my ramblings are anywhere near correct, it's an API issue, a language hacking issue, not a language issue 03:25
robkinyon no ... i think it's a language issue
luqui robkinyon, how come?
robkinyon in part because you need your CStruct representation to not be so free-flowing in its memory usage
at some point, it needs to point to the same contiguous section of RAM
luqui I think it's important to *be able* to do it; it's not very important how it's done
stevan robkinyon: thats an implementation issue I think
robkinyon well ... it's a language issue 03:26
luqui it's one of those points where easiness doesn't matter anymore
and that's what I consider an API issue
robkinyon because where do you define, in Perl, the layout of your CStruct?
luqui robkinyon, in somebody's library that knows about CStructs
stevan luqui: we have decided what make sense.. but I am still not confident that is what $Larry meant
robkinyon given that the entire language can be redefined given enough libraries, that's fair
stevan hence the p6l question :)
luqui stevan, but he hasn't responded 03:27
stevan luqui: yes, he is busy at $work,.. dont you read you own notes :P
luqui well, I was about to give two alternatives, one of which was that one, so yes
stevan :)
luqui it's also probably the case that you gave him something to think about 03:28
unfortunately, Larry doesn't think by talking like we do 03:29
so it's easy to get Warnockedf
s/f$//
stevan oh well,..
ok, time for me to get off the computer 03:30
luqui++ robkinyon++ # productive converstations 03:31
putter++ too
ok, night all &
putter good night stevan & 03:38
03:43 vel__ left 03:48 oylenshpeegul left 03:50 broquaint joined 03:59 jiing_mIRC joined
brent is now known as brentp 04:31
04:35 webmind joined 04:39 Amnesiac joined 04:42 PerlJam joined 04:44 broquaint joined 04:50 chip joined 04:57 broquaint joined 05:01 chip joined 05:06 webmind joined 05:18 broquaint joined, chip joined 05:24 webmind joined 05:31 dduncan left, dduncan joined
dduncan oops 05:31
05:39 elmex joined 05:40 justatheory joined 05:49 justatheory joined 06:35 kakos joined
GeJ greetings all 06:36
06:44 penk joined 06:45 PJF left
Supaplex elo GeJ 07:07
07:12 Alias_ joined
Alias_ seen audreyt 07:12
jabbot Alias_: audreyt was seen 13 hours 35 minutes 19 seconds ago
07:28 iblechbot joined 07:30 lisppaste3 joined
gaal $morning 07:34
dduncan greet ings 07:37
Alias_ $response
dduncan $response_to_response_or_next_comment
07:39 kanru joined 07:41 Aankhen`` joined
gaal putter: a few questions on S29code: 07:55
why "package Math::Trig {}" etc.?
how did you choose the precision of constants? 07:56
any thoughts on plugging this module into the tests?
and, I wonder if there isn't a way to avoid the macro use for trig_func. 07:58
08:50 luqui joined
gaal hey luqui. I'm still not sure about how IMPORT is supposed to go. 08:55
luqui by "supposed to go" you mean? 08:56
luqui thought it worked just like in perl 5
gaal p6 provides a default mechanism for (a) exporting symbols and (b) configuring modules
luqui or do you mean the "standard" export mechanisms
Alias_ is use Foo; finally going to mean use Foo (); ? 08:57
gaal if somebody defines an &IMPORT (or, if it's just like perl 5, &import) sub, does the standard one not get called?
luqui Alias_, I doubt it...
Alias_ crap
luqui yeah, that's a little annoying inconsistency 08:58
gaal, probably something like inheritance
gaal luqui: I want to refactor around line 855 of Parser.hs, it's getting a little insane
luqui looks
gaal but I don't know what it's supposed to do
luqui thinks that whenever you say "line 855", refactor is implied 08:59
gaal what goes on is:
(now i mean_
)
836: if use Foo (), nothing.
838: if there's an &import or &unimport as appropriate, call them. 09:00
846 onwards: symbol export.
(of course 846 on needs to be much more finely grained; that's not my immediate concern)
what I don't understand is, what is the responsibility of the implementor of &import? 09:01
does &import need to cause symbols to be exported?
scook0 gaal: "it's getting a little insane/but I don't know what it's supposed to do" -- I've had that feeling about parts of Parser.hs more than a few times :) 09:02
gaal can it prevent the default exporter from happening?
scook0: the problem isn't with parsing now, it's with p6 semantics not being clear yet.
luqui meaning...
should the implementor of IMPORT export symbols himself?
luqui thinks it probably should be able to
the thing is... it's not well specced AFAIK 09:03
gaal luqui: in that case they need a way to supress the default exporter
luqui: I know :)
luqui maybe the default exporter works like Export.pm
except with a nicer interface
gaal like I keep saying, the problem starts with two different things being conflated 09:04
symbol export control and module configuration.
why not separate them? 09:05
&IMPORT and &LOAD
&UNIMPORT and &UNLOAD
luqui Shouldn't we separate them from the "use" interface, too, then?
scook0 one is an import-time hook, and the other is a full exportation override?
luqui actually thinks that would be a splendid idea
gaal I think we should, but I don't know how that should look
luqui if it weren't for those damned sigils... 09:06
gaal scook0: init-time hook, I guess
luqui you could use named arguments for imports and positionals for config
but that probably wasn't a great idea in the first place
scook0 luqui: I was going to suggest the other way around (named/pos) 09:07
luqui ahh, but named makes sense for imports because they're already unordered... but those damn sigils
scook0 since positionals in general tend to represent the 'real data', whereas named/adverbs specify how it is to be used
luqui I think we probably need to go haskellish (again) 09:08
gaal luqui: I'm not seeing why sigils are a problem here?
luqui have some sensible defaults "import default, import nothing"
and then provide keywords (eg. "qualified") for finer-grained control
maybe not keywords though
gaal qualified+aliases is a whole different moose
luqui sigils are a problem because you can't say: use Foo :@bar
well, you can, but it means something different 09:09
scook0 you can if &use is a macro
luqui scook0, yeah, but that's not very pretty
anyway, I think configuration should have the full power of perl 6 call syntax
gaal :'@bar' :-( 09:10
luqui I'm tempted to say "use" and "import"
import Foo <@bar $baz>
use Foo <vars> # config 09:11
gaal in what order do you need to say these, and when do they happen?
luqui but that might be too fine a distinction for the common man
scook0 use Foo("config") <@bar $baz>
gaal then need to happen at compile time
and config prolly has to happen before import.
luqui oh, I was talking about two separate ways to use a module 09:12
gaal ah.
hmm.
luqui it kinda seems like they should be in the same place 09:13
because some modules would do both at once...
gaal Larry thinks there needs to be alot of control over imported scope
luqui thinks so too
gaal so the import syntax needs to be more complex anyway
as noted in S11 now 09:14
luqui that is a very terse description though
gaal luqui: I understand it though, if it doesn't need to deal at all with config. 09:15
There is a hack we could use
parrot GHC's command line +RTS ... -RTS
that's ugly though :)
luqui ...... use/import is looking better and better
what's RTS? 09:16
gaal luqui: only in the cases you only need one.
luqui not necessarily
you can think of import like roles and use like traits
gaal luqui: there are several "namespaces" of arguments in the GHC command line.
luqui import is for modules that are completely well-behaved, use is for modules that do funny business
gaal you can tell the runtime system things, but also some other subsystem
GHC avoids clashes by letting you say
ghc --make +RTS -M400 -RTS +Someothermodule -M48284 -Someothermodule 09:17
and the -Mx args mean different things.
luqui ahh, I see
that is ugly
gaal well, it solves a difficult problem.
Alias_ ugh... the pain from the techie kicking a plug out the wall continues... 09:18
gaal luqui: I don't buy "use is for modules that do funny business"... you want to init a module tersely sometimes
Alias_ Had to derack my dev server AGAIN after the NIC was intermittently failing
So that's a primary IDE controller and a NIC.
Crossing fingers nothing else goes 09:19
luqui we could kill the use Foo () syntax though
replacing it with import Foo;
Alias_ Server Room Protocol: Look twice, step once!
luqui use Foo # import default stuff, do normal stuff
gaal luqui: how do you say "import defaults from Foo" then?
heh
luqui import Foo # like use Foo (), but accepts import configs
enough people use "use Foo ()" that it might be worth it 09:20
it'll take some more mulling though
and probably more arguing
argue away if you have more :-) 09:21
gaal I think there oughta only be one keyword for compiletime module loading
I don't mind if it's renamed to "load" :)
luqui huh? what's wrong with "use"?
gaal nothing :)
just saying, there has to be one 09:22
09:22 Alias_ left
luqui is there a reason you only want one keyword 09:22
?
gaal don't forget there's also "require"
luqui which is runtime...
scook0 luqui: cargo-cult reasons?
gaal it's going to be very confusing
sure it's runtime
09:22 kane_ joined
luqui as if use Foo () isn't confusing enough already 09:22
gaal you're going to want to use the runtime version occasionally, and want it to feel like the compiletime variant 09:23
luqui we get rid of that inconsistency and replace it with a different kind of inconsistency...
you're suggesting that require should call IMPORT?
gaal nono
sec phone sorry
back. I'm fine with require MyModule; MyModule.IMPORT(x); 09:27
luqui I'm not sure what you mean by "feel like the compiletime variant" then 09:28
gaal it'd be great if "x" could be the same as what it would have been had I said "use MyModule x"
that's what I mean.
luqui yeah... I'm not sure that has anything to do with my side of the argument
import Foo x; require MyModule; MyModule.IMPORT(x) 09:29
use Foo x; require MyModule; MyModule.USE(x)
er, s/Foo/MyModule/g
gaal (also, I still think the return value of use/require should be a callable Module object, so (require MyModule).IMPORT(x)
but I think we already agreed on that one
)
luqui that might be nice
gaal my gripe with your idea is that sometimes you want both USE and IMPORT 09:30
luqui and then importing that module would take two lines?
which is ugly and painful?
fair enough 09:31
gaal yeah, not only because order matters with these lines.
luqui isn't sure how order would matter except in the most bizarre of circumstances
i.e. the use handler populating what is to be exported (/me realizes that's not all that bizarre)
gaal any module that does things at compile time would want its config data known early 09:32
luqui depending on what those things are
gaal there's an incidental objection that people coming from mainstream languages would be used to saying "import Foo" early up.
though that's not a very strong argument :)
luqui my biggest gripe against my argument is that we now have two operations that do almost the same thing 09:33
and subtle distinctions are bad for beginners
maybe some special syntax... 09:34
use Foo config, args importing import, args
gaal well, the unclever approach is to say, importation is symbol table stuff, quote it anyway
luqui 'course, use Foo importing <bar> is a little verbose
gaal actually I like special syntax, if you can come up with a golfy one. 09:35
luqui gaal, I'm not quite sure what you mean
I don't see how quoting it helps
gaal luqui: make all importation explicit stringy args, then you only have to reserve the args for importation
luqui gaal, example?
gaal use Foo :velocity<500> :imports<&takeoff &land> 09:36
luqui I assume s:2nd/importation/configuration/
gaal subsume importation under configuration (like in p5)
luqui gaal, but you're really losing on the power of importation there
scook0 I suppose `use Foo <&takeoff>` could be short for `use Foo :imports<&takeoff>`
gaal maybe Foo's USE function can have a trait declaring if it takes responsibility for importation? 09:37
luqui scook0, and require all args to be named otherwise, like .new?
gaal luqui: yes, that's why I don't like it
luqui a powerful importation mechanism kills all "namespace collision" antihuffmanism 09:38
having to worry about namespace collisions should be a programmer's last concern
we do have siglists 09:39
gaal siglists?
scook0 luqui: what do you mean by "namespace collision" -- which namespace(s)?
luqui use Foo :imports:( :my($foo) :our($bar) );
gaal, signature objects, :(...)
scook0, er, name collision 09:40
making names long so that there's a better chance they'll be unique
gaal luqui: where's that specced? interesting
luqui S06 IIRC
scook0 luqui: i.e. two modules exporting the same short name?
luqui scook0, aye
here's the one I'm working on now: two different methods named the same thing 09:41
that's the hardest one
at least with Perl's world-view it's hard. With Haskell's (and C++'s, but they didn't take advantage of it), it's easy
gaal, btw, :() is actually \() 09:42
gaal luqui: in "Closure paramaters"? wow, that doesn't really advertise the feature very loudly :) 09:43
luqui :() is how it's declared
use Foo :imports\( :my<$foo> :our<$bar> )
geez that's a lot of special syntax for a simple thing
gaal use Foo :velocity<500> importing :my<$foo> :our<$bar>; 09:45
luqui but that may be another good way to go: require use to take only named or only positional parameters
in the positional case, just pass it as the sig to :imports\()
gaal, you like the namey syntax?
I agree it's not bad
except for the common case
scook0 the problem is that it needs to be (a) simple in the simple cases, (b) not too complex in the complex cases, and (c) decent scaling between a & b
luqui use List::Util <reduce>
shouldn't be: use List::Util importing <reduce> 09:46
gaal well, what's more common? export control or config control?
luqui I'd say exports
but I can't back that up
gaal I agree SOMETHING needs to win the short form :)
scook0 it matters less if you make the not-quite-so-short form still pretty short 09:47
(assuming you can) 09:48
luqui also, what about module aliasing
gaal use List::Util importing nothing; # <- Alias won't like that...
scook0 I just wish there wasn't conflict between config-adverbs and scope-adverbs
luqui i.e. use Data.Map as Map
s/use/import/
gaal luqui: that's the whole other moose :) 09:49
luqui is beginning to think that, once again, Haskell got it right
09:49 dada joined
luqui have a very small minilanguage for exports 09:49
that supports aliasing and fine-grained import control
gaal luqui: but then, they have no config control at all
luqui gaal, that's not true
gaal and no control over scoping of symbols
luqui: how do you config a module in haskell? 09:50
luqui not Haskell exactly!
haskell's philosophy
which is essentially: use Foo::Bar::Baz config importing <baz quuz> as Baz
a little englishish minilanguage
(except "as" means something different, and it's too short to be put at the end like that) 09:51
gaal that works for me except people who want the simple case (not realizing there were a few different ones) would hate it
09:51 nnunley joined
luqui gaal, but we have to piss off those people if we're diverging from perl5 at all 09:51
if there is any distinction between config and exports, those folks will yell 09:52
because one's going to be longer
gaal there is one say to avoid it...
luqui ponders use/imp to keep them quiet :-p
gaal, what say is that?
:-)
gaal let the module author provide the default one with precedence :)
luqui yeah, uh.... no 09:53
you want to make people feel like they're in control
scook0 luqui: (one's going to be longer) that's why I liked `use Foo <$symbol>` `use Foo :option`
luqui thinks that's one of the most important things to do
scook0 but it interferes with export control
luqui scook0, how come?
you mean powerful export control... 09:54
okay
scook0 luqui: yes, :my() and the like
gaal yeah, quoting the symbol means no powerful excontrol
luqui no, we still have to quote symbols
scook0 unless you can put the excontrol inside the <>
gaal so I think we all agree no relegation to strings, yes?
oh 09:55
hmm
we're going around in circles
scook0 i.e. use Foo < :my($symbol) > # ugly though
luqui scook0, especially from an implementation perspective
gaal scook0: and fragile too :(
(export a "<" symbol...) 09:56
luqui gaal, oh my god!
operators!
gaal or an op
yeah
luqui maybe that's just: use Foo < infix:<+++> > 09:57
(IIRC <> does balancing)
gaal how does that work? oh, I forgot < > is teh smart
scook0 clever <>
gaal foresight++
luqui not quite:
scook0 (I can see that causing pain in obscure cases though :)
luqui use Foo < circumfix:<[: :]> > # uh oh 09:58
gaal yeah, that helps against the common case, and if someone really wants to overload '<', let them sweat
hmm, why doesn't that work?
scook0 use Foo 'circumfix:<[: :]>'
luqui scook0, yeah... if people are importing circumfix operators, presumably they know what they're doing enough to use those quotes 09:59
scook0 use Foo < circumfix:«[: :]» >
luqui scook0, same problem
the problem isn't balancing, it's word splitting
gaal ah, right.
scook0 (I just realised that myself)
gaal use Foo q#circumfix:<[: :]># :( 10:00
luqui but this does bring up an interesting point
you are allowed to write operators infix:{'+'}
or in many various ways
should we really be quoting these symbols?
scook0 it seems to me that qw<> isn't really what we want
we really want a quote-identifier construct
gaal can you do crazy stuff like infix:$op ?
luqui ?eval my $op; BEGIN { $op = "+" }; say infix:{$op}(3,4) 10:01
evalbot!!!
gaal my $op = <- + / *>.pick; infix:$op := { ... } 10:02
luqui: oh yeah, I was wondering what happened to it a few days ago
I gotta go out and do some stuff before the stores close, bbiab
I hope you guys will still be here in ~40 min? 10:03
luqui gaal, and for yours, $op needs to be chosen at compile time
no, I'm going to bed
gaal luqui: right
luqui will continue to think about this though
azuroth does anybody in here know latin?
luqui isn't even sure how to say "no" in latin
gaal luqui: ok. I was hoping to make export a little saner for the release
but oh well, at least I'll have $larry cornered on Feb :-) 10:04
luqui is there a hackathon?
azuroth I'll tell you how to say yes: sic
luqui that's helpful. so then when somebody asks if I know latin, I can say "sic"
and be lying...
gaal azuroth: port your text to perl, then use Lingua::Perligata or whateverthemoose it is
luqui: osdc.org.il
azuroth: "sic" isn't yes, it's "thus" 10:05
luqui: wanna come? :)
luqui definitely
gaal audrey will be there too.
luqui but alas, I cannot
$$ && $school
:-(
azuroth wow, so it is
gaal azuroth: sic transit gloria mundi 10:06
gaal &
azuroth thus it travels the glorious world...?
s/it/he/
luqui geez, pugs is huge 10:07
took forever to check out
evalbot should be on in about 20 minutes
it has to build from scratch
luqui & 10:08
gaal >> showstoppers for release afaik: 1. PGE/parrot integration is broken 2. multisub exports are broken 10:10
azuroth: "thus passes the glory of the world", but look it up, it has interesting cultural moose
& <- for real 10:11
azuroth ahh, okay
you'll be back
10:15 r0nny joined 10:16 r0nny joined
gaal that took less than expected :-) 10:24
I hope Aankhen`` comes back soon, I want to hear what artists talk about when they get together :) 10:28
azuroth gaal: "why did you tell the story, girl?" is "cur fabulam narrabas, puella?", correct? 10:29
gaal azuroth: haven't the foggiest :) 10:30
I do know puella is girl though!
azuroth oh. you tricked me! :-) 10:31
gaal what's the infinitive of narrabas? 10:33
azuroth narrare 10:34
to tell
gaal knows latin like he knew haskell last march
oh, so it is 1st conj.
azuroth heheh. I'm just learning both
gaal so apparently this is indicative active 2nd person sing. 10:35
imperfect
and fabulam is accusative 10:36
azuroth yeah, so probably more like "why were you telling". but I don't know (plu?)perfect or anything yet
gaal i've no idea what pluperfect means :) "why have you been telling"? 10:37
www.math.ohio-state.edu/~econrad/lang/latin.html # latin cheat sheets, err, docs/quickref/
10:37 penk joined, G2 joined
gaal just noticed the domain name for that site, ha. 10:38
azuroth I think pluperfect is when something was done and finished a while ago. and perfect is just when something's been finished?
gaal why have you told? 10:39
okay
this gets harder than monads when you reach the 3rd declension :)
svnbot6 r8645 | Darren_Duncan++ | r1927@Darren-Duncans-Computer: darrenduncan | 2006-01-13 02:40:06 -0800 10:41
r8645 | Darren_Duncan++ | [/ext/Rosetta|/ext/Rosetta-Engine-Native|/misc/Rosetta-Incubator] : incremented versions of all modules that have a pre-rewrite equivalent on CPAN, so these post-rewrite versions are kept higher than a new CPAN retroactive release due tomorrow that uses the pre-rewrite versions
dduncan and with that, there's probably nothing else I need to commit before 6.2.11 is packaged up 10:42
azuroth I'm up to page twelve of the first book. 5% of my way to learning latin! 10:46
11:31 iblechbot joined, bsb joined
theorbtwo backlogged. 11:32
11:47 pdcawley joined 11:50 EdLin joined 11:56 elmex joined 12:16 ilogger2 joined
ingy seen audreyt 12:28
jabbot ingy: audreyt was seen 18 hours 50 minutes 24 seconds ago
12:48 Stoffe left, putter joined
putter hi gaal, let's see... 12:48
re S29code, 12:49
daja77_ pokes rafl
putter re ' why "package Math::Trig {}" etc.?', because it's speced as a package, rather than say as a module. inserting things directly (multi Math::Trig::mumblefrotz (...) {...}) is fine, but at some point you have to say its a package. 12:50
12:50 _ori joined 12:51 _ori left
putter re ' how did you choose the precision of constants?' - I cut and pasted the first line of the example section of their OEIS database entry. ;) Something more principled will be needed at some point. 12:53
rafl pokes back
daja77_: What's up?
daja77_ why is your jabber timing out 12:54
12:54 calanya joined
rafl daja77_: I'm online and talking to someone. 12:54
putter re ' any thoughts on plugging this module into the tests? ', not quite yet. this is still mostly just a grabbag to help with the "now the piln backend passes sanity, what about the rest of the tests?" Prim.hs port phase. 12:55
gaal putter: I meant, why not "package Math::Trig { multi mumblefrotz }"? 12:56
putter: re precision, the danger is with infectuous overprecision
BigFloats are costly etc. 12:57
daja77_ rafl: can you read me there? 12:58
putter My current impression of the Right Thing is to take a finished S29code (half of it is still unmodified pod), things from Prim.hs which can be easily p6-ified, and blobs of PIL2JS/lib6, and generate a big mostly complete p6 "prelude", to complement the small piln prelude.
gaal nods 12:59
putter second batch of questions in a sec...
gaal Is the exporter properly part of Prim or of Parser? 13:00
it doesn't parse things itself; but it does return bits of AST.
hmm, maybe neither? Pugs.Exporter? 13:01
then again, it only exports (ha) one function, which should probably only be called from Parser... 13:02
putter re ' and, I wonder if there isn't a way to avoid the macro use for trig_func.' - sure... you can compute them at runtime (or BEGIN{} compile time... using something like foreach <sin cos ...> -> $op { M::T::$op ::= -> (Num $n --> Num) { ... Primitive::$op... }}. or you can expand them, generating a much less maintainable blob of 10x bigger code. what's wrong with macros? ;) 13:03
13:04 chris2 joined
gaal :-) 13:04
"Syntax hilighting doesn't work with them!" 13:05
putter second batch: re why not putting them in brackets (M::T { multi mumble }), the spec has fully qualified names (multi M::T::mumble), with adjacent lines going into different packages/classes. and S29code mirrors S29draft.pod as closely as possible. 13:06
re overprecision... hmm... how do you keep both 64bit long double folks and 32bit double folks both happy? perhaps we need a precision pragma? 13:07
re 'Is the exporter properly part of Prim or of Parser? it doesn't parse things itself; but it does return bits of AST.'... well, it sounds like an AST-flavor _macro_ to me! ;) 13:08
gaal :D 13:09
putter re syntax highlighting... having anything but p6 generate the text map info looks basically a lost cause. the usual regexp approach... for p6!?!;) 13:11
gaal putter: util/perl6.vim works nicely most of the time. Of course it isn't perfect. 13:12
13:15 kolibrie joined
putter ah, but that's only because I don't yet have control (complete control! lexically scoped! oh, I sooo can't wait:) over the p6 parser rules, nor AST-based macros. perl6.vim?! Ha! ;) 13:16
gaal well, when you have THAT control you obviously will use the perl-enabled version of vim to load alternate syntax files. :) 13:17
putter hmm... re exporter... it won't in time for next release, but once we have pil2/piln, we can have pil2/piln modifying macros! Which would actually be the Right Way to implement the pure-p6 version of the exporter.
theorbtwo The thing to do is to get pugs to produce some sort of dump with full position information, and get perl6.vim to parse that. 13:18
Viola, always-correct coloration information.
gaal theorbtwo: we're not THAT far from that, actually.
theorbtwo I know. 13:19
putter theorbtwo++ # re position dump - agreed
gaal too bad you can't do this: 13:20
# echo ':D say "hi" if 4==4' | ./pugs -
theorbtwo The problem, BTW, with the plan I just layed out is that it won't do anything vaugely useful for things that simply don't compile. 13:21
gaal indeed.
how do clever IDEs like Eclipse's JDT do it? 13:22
putter actually, I'd like to see a refactoring engine rather than just a dumper. so emacs/vim can say "extract the code between lines n and m into a helper function", and Perl6::CodeEngine does the right thing, and sends back the new text and highlighting. IDEish var docs and completing. what else...?
gaal backtrack a lot I suppose...
yes, that is a holy grail of sorts.
putter s//Perl6::CodeEngine::Server/ 13:23
re 'things that dont compile', hmm... maybe picky grammars should come with a companion lax grammar for highlighting?? or...? 13:26
13:27 avar joined
putter I suppose one could fall back to the wizzy grammar-based equivalent of the current "bag of simple heuristics which somewhat match user expectation" approach. 13:28
theorbtwo Some of the time, it's perfectly reasonable to put up a warning somewhere onscreen and keep using the highlight from the last time it did compile. 13:29
putter re IDEs, tolerant parsers and bags of pattern matchers. tend to be happy with very macroy code like emacs/gnu C files (I actually dont know how well Eclipse currently deals). 13:31
s/happy/un$&/ 13:32
re holy grail, hmm, since p6 will embed ruby and python and..., there is a chance to define the standard way to play this game. 13:33
gaal doesn't have a recent CDT to check 13:34
13:38 bsb joined
putter since there is so much in common between refactorers for different languages, and given p6's needs and the perl communities' unusual attitude of "ooo, another language! yes! let's bring it in! new CPAN module!" (rather than the usual, "we can do that better. no? well, why would you want to do it anyway."),... one might create *the* refactoring engine... ;) 13:39
putter daydreams happily...
gaal putter: maybe looking at eclipse is worthwhile, since they did want to be a metaide. 13:40
putter indeed. 13:41
gaal: a question: ok, say once oo and rules are working, folks pump out say javascript, ruby, python, and scheme backends. so suddenly there are new implementations p6ruby, p6python, etc. what are the resulting community social dynamics? the scheme folks have so many implementations the probably wont care. though #perl6 may see some 13:44
gaal resulting community social dynamics? I don't know, world peace maybe? 13:45
putter well, condescending remarks about syntactic complexity etc (I have particular people in mind;). the ruby folks will be happy as long as it's a really conforming implementation. what about the python folks? "we're not perl" has been their mantra for a decade. I don't have a feel for how that will play out. or more importantly, what we can 13:47
do to help it play out nicely.
re world piece, err, peas, err, peace. I'd be ecstatic if we can just start writing "interesting" software systems. what we've got now is still soo mid-1970's. 13:49
gaal I honestly don't know. In any case it's not like the code generated to high-level languages is very readable; this looks to be useful in programmatic slurpage of code, where original language is sorta hidden away anyway 13:50
putter err, huh? 13:51
gaal hmm, you mean consuming code from cpan in other languages? 13:52
sorry my brain is being drained by it being cold and my not having eaten lunch
I'll probably make more sense after I take a break :) 13:53
putter oh, no. use Language::JavaScripdt::OnPerl6; ... eg, misc/JavaScript-FrontEnd/ . actually, I'd like pugs to eventually do file suffixes, so pugs foo.py does the right thing. 13:54
is it a python implementation? is it p6? if you're a pythonista, do you care? 13:55
pugs -MLanguages::AutoFileHandling foo.js bar.py hee.rb -e 'say "wow"' 13:56
gaal how do objects get bridged among languages? 13:57
robkinyon starts waving his hands a lot in gaal's direction 13:58
gaal fwiw there are a number of python implementations available
13:58 Limbic_Region joined
gaal dodges and decides this is a good time for that break 13:59
robkinyon gaal: Actually, there was a discussion last night between stevan and luqui about that topic
well, a closely-related one, at any rate
gaal bbiab :) &
theorbtwo Gah. 14:02
Building pugs is taking approx 3 years on my system.
...and it's not that weak a system, either.
Limbic_Region theorbtwo - how much free memory does it have? 14:03
putter gaal: adapter methods? even in a pure p6 context, I'd like to be able to say "in this lexical scope, .nth($n) on Foo should result in a call to the underlying .nth with $n+1 ("baaaad silly 1-based indexing CPAN module"). or ".foo means .bar", etc.
theorbtwo Woo, done!
putter wants complete control of his namespace. the ability to correct for CPAN modules with brain-dead apis.
integral ghc still seems to need >500MB on AST.Internals and the precomp prelude
theorbtwo Around half a gig. 14:04
449,948 kB of real mem free, says free.
putter non-trivial.
later & 14:05
theorbtwo dresden-ocl.sourceforge.net/aboutproject.html 14:06
Hm, ./pugs -CPIL2-Perl5 -e 'say "this is a test"' gives only one PPos, but -CPIL2 (with no dash) has two MkPos. 14:13
14:16 SamB joined 14:37 peacock joined 14:39 calanya joined 14:53 hcarty joined 14:58 iblechbot joined 15:05 gantrixx joined 15:26 justatheory joined 15:30 Eimi joined, lisppaste3 joined 15:38 rantanplan_ joined
gantrixx do any of you guys ever talk to Larry Wall? 15:50
theorbtwo @Larry, also known as the Cabal, has a weekly conference call, and $Larry posts to the mailing lists fairly often. 15:52
gantrixx As a linguist, I was wondering if he thought english was the best language to base a programming language off of 15:53
I wonder if we will ever have a hindi or mandrin based programming language 15:54
or perhaps it should be esperanto based
theorbtwo (1) I think it's more based on human languages in general then on English in specific.
(2) Ruby?
gantrixx Ruby is none english? 15:55
theorbtwo Ruby's primary author is Japanese.
gantrixx Take Perl for example, all the key words are english based
theorbtwo Oh.
gantrixx if, print, has, return
theorbtwo That's not so much a matter of it being "based on english".
It's a matter of convience. 15:57
gantrixx the key words are based on english
GeJ considering that 99% of all the programmers in the world speak more or less english, english seems a defacto language for syntax keywords in programming languages.
theorbtwo sh does it less -- it's if...fi, case...esac. That's horrid.
wolverian GeJ, where did you get that figure?
gantrixx and yes, even the grammars are based in english
theorbtwo There used to be localized versions of visual basic. They were horrible.
Some microsoftie did a great blog entry on that.
gantrixx that is chicken and the egg though 15:58
theorbtwo Documentation would have to be even more translated. Communities would be even more fragmented.
gantrixx most programmers speak english because they have to
because the majority of the programming technology is based in english
theorbtwo A hardly-insignificant number speak English because it's the most common language in the world.
GeJ wolverian: inhouse statistics... we have polish, slovak, french, german, chili(an?), american and autrian people here, and all of them grok PBP. :) 15:59
gantrixx english is the most spoken language in the world, but it is not the most widely spoken native language
english is the new latin
you used to have to know latin to study any of the sciences
now you need english
theorbtwo You hardly have to speak English at a native level to be able to undertand if a=1 then print "Ja" else print "Nein" 16:00
gantrixx the language du jour basially depends on who the current world power is
si a == 1 entonces escribe "Ja" 16:01
and you hardly have to speak spanish to understand that
wolverian Lingua::Perligata :)
theorbtwo And, indeed, I can follow it just fine, even though I hardly speak Spanish.
wolverian er, s/::/::Romana::/
gantrixx it's completely arbitrary. the question is if you wanted to make it a non-arbitrary design decision would you base it upon multiple languages, the dominant language, or a common language 16:02
if the s in s/this/that/ is based in english (substitute)
theorbtwo gantrixx: It's worth noting that Ruby was written with English keywords, even though it was written in Japan by a Japanese man. 16:03
What does the y in y/-/_/ stand for?
gantrixx I don't know y
I don't think I've ever used y
wolverian it stands for tr ;) 16:04
theorbtwo Everybody uses tr///, because y/// doesn't stand for anything.
(Except when you're golfing -- trying to do a task in as few keystrokes as possible.)
What does the x in " " x $indent stand for? 16:05
gantrixx do it that many times
theorbtwo Eh?
gantrixx print $indent number of spaces
theorbtwo Yes, but what does the "x" stand for?
gantrixx times
theorbtwo Exactly. 16:06
16:06 eric256_ joined
gantrixx x is commonly recognized in mathematics as times 16:06
theorbtwo It's not the language that they come from that's important, it's that the symbols are as fimilar as possible to the target audience.
Most of the time, when writing a new programming language, that audience is programmers.
gantrixx or the fact that we write our code from left to right reflects western languages 16:07
theorbtwo ...who are already fimilar with "if", even if they aren't English.
gantrixx perhaps if the arabs had a programming language it would be right to left
or the chineese top to bottom
Limbic_Region gantrixx - your original question doesn't appear to be the one you are actually asking though
the programming language is not modeled from a single spoken/written language 16:08
theorbtwo Except left-to-right isn't as clear as you think it is.
gantrixx my original question was, in essense, is english the best language for basing a programming language off of
theorbtwo In sin(exp(2)), the sin is evaluated /last/.
Limbic_Region right - but keywords and syntax are two completely different things
you appear to be concerned with keywords and not syntax
theorbtwo ...and in perl you can write for (1..$_) {print "foo"} or print "foo" for 1..$_. 16:09
integral ((constant 2) >>> exp >>> sin) `app` _|_
theorbtwo I'm not aware of any human languages that read bottom to top.
integral: ==> became >>>? 16:10
gantrixx in Perl I also write "my $a;" and not "$a mi"
integral theorbtwo: haskell, but yeah, s/>>>/==>/ and you get perl6 mostly, but you can then drop the lift and the apply
gantrixx so it is based off the english grammar that posessive pronouns come before the noun
although Perl does allow some flexibility with some of it's grammar, it is still based on english grammar and english key words 16:11
theorbtwo Yes, it is.
16:11 penk left
theorbtwo I think that's for the best, given the current state of the world. 16:11
Not only that, but it's a "best" with /huge/ inertia.
Hm, on second thought, it's a state.
16:11 elmex joined
gantrixx you sound like you work for Microsoft 16:12
"we are the best, because we are the biggest"
theorbtwo In some cases, it's true.
eric256_ isn't english generaly the chosen langauge of science in general? (/me could be 100% wrong there but i thought i read that somehwere)
gantrixx OK, so IE is the best because more people use it than Firefox
English is the language choosen because there are the most jobs in english 16:13
eric256_ gantrixx if you want to talk to the most people and the most people speak one langauge, then you speak that language. thats simple logic
theorbtwo If I were writing a language specificly to be taught to 4th graders in Mexico, then I would possibly spec it in spanish, and use spanish grammar and keywords.
If I'm writing a language to be used by programmers, I'll do it in English.
eric256_ you are mixing uses of best. enlglish might be the 'best' because its most common, not becuase its best suited.
gantrixx exaclty, so why don't we program in language based off mandrin chineese?
certainly there are more native mandrin speakers than english speakers 16:14
theorbtwo It might be. OTOH, I doubt you'll find a language better suited -- none of them are purticularly well suited, I'm going to guess.
stevan gantrixx: cause it's hard to get a keyboard with mandrin characters here in the US :)
eric256_ are there more native mandrin programmers though?
gantrixx so my question is, should the language be based off who ever is economically dominant at the time, or should science (and computer science) have a universal language
should we all give up our native language and speak one universal science language 16:15
theorbtwo It does, to a large extent.
gantrixx ?
theorbtwo That language is C.
eric256_ science and programming do have a universal language, it just happens to be the most economically dominant currently ;)
gantrixx and C isn't based off of english?
theorbtwo Not terribly much.
A few keywords.
But that it is is independent. 16:16
gantrixx key words, and grammar too
eric256_ and many of those keywords don't mean the same thing to english speakers as they do to programmers ;)
theorbtwo It's not English that's the standard, it's C.
gantrixx and all the computer technologies all have english names as well
rantanplan_ Hehehe, but you can program Perl in Latin ;)
theorbtwo ...and you overestimate how much C is like English.
For example, English-speakers commonly put if clauses at the end of the senence. 16:17
That's completely illegal in C.
gantrixx I agree there is more mathematical "language" in most programming languages and mathematics tends to be a universal science language
but you can't deny the fact that most programming languages were written by westerners and based off english
theorbtwo If I have seen further then others, it is by standing on the shoulders of giants. 16:18
eric256_ why would we deny it? and why is that a bad thing?
you seem to have a beef not a question
gantrixx no, it's not a beef, it was a question 16:19
I was wondering what a linguist such as Larry Wall thought
theorbtwo I think if you want to completely change all known syntax and keywords, go ahead -- perl 6 will make it easy.
But nobody will use it.
16:19 vel__ joined
gantrixx but others wanted to discount the fact that programming languages are based off english, so I just defended my assertion 16:19
theorbtwo Programmers think in terms of function(arguments), because they've used other programming languages, and because they've used math. 16:20
gantrixx I don't want to. My native language is english
theorbtwo Is it better then (arguments)function? Not really. Is it worse? Nope.
wolverian first_argument.function(more_arguments) # :)
eric256_ (@arguements) ==> function; 16:21
gantrixx and most comments tend to be in english in source code 16:22
wolverian put function applied to @arguments to @results; # yay!
eric256_ i would disagree on that, i've found plenty of code with what i assume was german commenting
theorbtwo gantrixx: That hasn't purticularly been my experince.
eric256_ and some with spanish variable names, now that can be fun to decipher ;)
gantrixx perhaps you guys work with different code bases than I 16:23
I have seen several HOW-TOs in german, but rarely do I see source code comments in anything but english 16:24
how many people here are non-native english speakers? 16:25
wolverian I am.
gantrixx what is your native language? 16:26
theorbtwo Audrey isn't here at the moment, but she is.
wolverian gantrixx, Finnish.
16:32 avar joined
ingy seen audreyt 16:55
jabbot ingy: audreyt was seen 23 hours 17 minutes 21 seconds ago
16:57 Southen_ joined
eric256_ seen eric256 17:04
jabbot eric256_: eric256 was seen 3 days 22 hours 48 minutes 33 seconds ago
eric256_ seen eric256_
jabbot eric256_: eric256_ was seen 5 seconds ago
eric256_ wonders how he alawys ends up with that darn _
Limbic_Region eric256 - likely your client automatically reconnects (or you do) quickly 17:12
if you have lost connection but the IRC server doesn't know it and you reconnect, your client automatically appends the _ as eric256 is still logged in as far as the server is concerned 17:13
eric256_ stupid computers ;)
Supaplex artifical inteligence has a funny aftertaste don't it? :) 17:14
pdcawley_ Mmm... I spy some hard reading ahead, TaPL just arrived... 17:15
Limbic_Region . o O ( TaPL ? ) 17:22
Types and Programming Languages
pdcawley_ grins, bingo! 17:24
eric256_ is that a book then? 17:25
pdcawley_ Yes. 17:26
GeJ a painful one... 17:28
not only it is big, but reading it hurts.
Well, in my case.
eric256_ hey, since p6 supports lazy lists.. could you then define a list as the function that produces it? 17:29
my @fib = iterator { $_[-2] + $_[-1] };
well with better fibonacii code than that (more specificaly code that works. ;)) 17:30
Limbic_Region eric256_ - my guess is that is not something that will (or should) be supported in the core language 17:32
there is just way to many things to have a sane DYMism 17:33
OTOH, I think the core language should allow subclassing of the array class easily enough for you to roll your own and make your own DWYM
eric256_ - have you read my tutorial/article on iterators? 17:34
eric256_ no
in the pugs tree?
i know i've seen code on perlmonks for lazy lists in p5...they had to be treated special....arg 17:39
clkao 8/win 28 17:42
17:48 nothingmuch joined 17:53 G2 joined
Limbic_Region sorry - work 17:54
eric256_ see www.perl.com/pub/a/2005/06/16/iterators.html
fwiw - you can change your nick by typing /nick <new handle> 17:55
ric256 is now known as eric256 17:57
eric256 work is evil ;) 17:58
17:58 Lartuv joined 18:01 broquaint left
tewk Is there a recommended REPL shell for Perl5? 18:11
theorbtwo perl -de1 18:15
kane_ seen audreyt 18:30
jabbot kane_: audreyt was seen 1 days 52 minutes 22 seconds ago
18:30 Ovid joined
Limbic_Region From a reply to Guido Van Rossum's blog "Can a language be functional, declarative, procedural, OO, statically typed, late binding, uncle tom cobley and all and remain easily accessible or shall we admit that there is not one ring to bind them all" 18:30
is it possible that not everyone has heard of Perl6 yet? 18:31
integral Sounds like he has... 18:34
18:49 elmex joined 18:54 Qiang joined 18:59 G2 joined 19:16 hexmode joined 19:19 ycheng_ joined
gantrixx Actually, I don't think that many people know about Perl6 19:46
it would be cool if /. or LugRadio started covering it 19:47
Khisanth no it wouldn't! 20:03
20:04 putter joined
putter Does anyone know off hand if invocants are automatically rw? I'm trying to reconcile Prim.hs and S29draft.pod: 20:04
\\n List pre pairs safe (rw!Array)\ 20:05
multi Array::pairs (@array : MatchTest *@indextests --> Int|(List of Pair) )
Khisanth would a rw invocant mean that calling a method could actually alter the invocant? as in $foo.bar(); # $foo is now a different object from before the call to bar() 20:08
putter ?
wolverian that's how I would read it. what else can it mean? 20:09
putter looks like PIL2JS lacks a rw for other methods (exists delete) Prim has as rw!Array. I'll leave it off and leave a comment.
putter is assembling a composite Array.pm from assorted sources. 20:10
Khisanth wolverian: but that would be EXTREMELY suprising behavior 20:16
althought potentially "useful" in the right situation 20:17
is there an AES for roles? 20:23
wolverian S12?
svnbot6 r8646 | putter++ | docs/AES/S29draft.pod - corrected typos - adverbs missing their colon. 20:25
r8645 | Darren_Duncan++ | r1927@Darren-Duncans-Computer: darrenduncan | 2006-01-13 02:40:06 -0800
r8645 | Darren_Duncan++ | [/ext/Rosetta|/ext/Rosetta-Engine-Native|/misc/Rosetta-Incubator] : incremented versions of all modules that have a pre-rewrite equivalent on CPAN, so these post-rewrite versions are kept higher than a new CPAN retroactive release due tomorrow that uses the pre-rewrite versions
putter err, what does this mean? &Array::pop<Array> := &Array::splice<Array>.assuming(... what's the <Array> thing? 20:27
S29drarft.pod, but pugs doesn't like it. alternate implentation coming up... 20:31
gaal rehi 20:35
20:40 eric256 left 20:41 Qiang joined
putter hi gaal 20:42
gaal heya 20:45
gaal battles indent levels
putter puzzles over Array.pm code 20:47
.[] should be postcircumfix, no? PIL2JS has it as prefix. puzzlement. 20:48
gaal putter: maybe it's comutative, like in c :-) 20:49
putter oh I need to borrow a clue. ok, my understanding is postcircumfix:<[ ]> is array indexing. Prim.hs defines .[] as array indexing (note the "."). PIL2JS defines a prefix:<[.[]]> as array indexing. whaaaa?!? 20:51
gaal: :) 20:52
gaal putter: I think that's so you can chain lookups, as in 20:53
$x = { a => b => c => d => 42 }; say [.{}] $x <a b c d e> 20:54
putter is it just a short form for say [postcircumfix:<{ }>] ..., or is the associativity actually different? 20:56
gaal dunno 20:57
ingy has guido submitted any patches? 20:58
20:58 dduncan joined
gaal ingy: to pugs? I think he declined the invite. 20:58
ingy really?
putter bah. mission creep. I do not need to understand, merely cut, paste, and comment. thanks gaal ;)
ingy bummer
putter maybe once pugs is running python...
gaal putter: what primary benefit do you see for non-p6 frontends to pugs? (as opposed to, say, parrot) 21:00
(not contesting of course; it's just that I hadn't actually considered this clearly till today)
putter the world is full of code. I want to be able to run it without caring what language it is written in. I wish to escape the maze of twisty little passages which is software engineering, where when you need n components, there are always pointless constraints which 21:03
prevent you from getting more than m of them at a time. where m is always strictly less than n. and discovering the constraints requires a backtracking crawl through design space. 21:04
putter pauses for a deep breath ;)
I want our field to not be so profoundly throughly broken. 21:05
That's all. ;)
gaal putter: okay, but there's some overlap between this and parrot's goals 21:07
putter yes 21:10
gaal so specifically why do it in the pugs level?
putter though with a rather lower-level feel to it. custom coding language parsers in "assembly", rather than building wizzy Arrow based grammar mutators and parser tools. 21:11
21:12 elmex joined
gaal ah, if you want arrows pugs is the place to find them :-) 21:13
putter grammar Language::JavaScripdt::Grammar::FireFox is Language::JavaScripdt::Grammar::ECMA3; GAP (algebra system) "yet another language" grammar is speced in BNF variant number three hundred fourty two. and can be converted to rules by BNF::Frobnicate. ... 21:15
gaal ahhh you want it higher up than arrows, you want it in perl 6! 21:16
21:16 luqui joined
gaal that's too high up for me atm but I get glimpses of the appeal 21:17
morning luqui
luqui mornin'
putter use Perl6::PILN::Transformation::CPSConvert; ...
putter ok, I'll stop then... :) 21:18
gaal slinks off to some well charted haskell :-)
putter ooo, don't slink!
luqui dislikes the Cont monad 21:19
gaal unslinks
putter any annoyance in my tone is just years of frustration at a broken field. at being a shoemaker's child with broken crippled torn bloody gangrenous shoeless feet.
gaal putter: oh, I'm not scared off by the tone (or by the purpose)! just.. too wide for my brain,57 things at a time 21:21
putter thank you for the opportunity to vent. and the reminder that it's "not completely obvious" what's needed. the old "my thesis now written seems obvious, so why finish it?" ;) 21:22
luqui what the heck are you guys talking about?
theses, shoemakers, 57 things...
putter why run python on pugs, rather than simply on parrot. 21:23
luqui because it's easier to compile to pil than it is to pasm?
putter yes :) and manage the compilers as well. 21:24
luqui also because python on JS would be cool... 21:25
svnbot6 r8647 | Darren_Duncan++ | r1931@Darren-Duncans-Computer: darrenduncan | 2006-01-13 13:24:05 -0800
r8647 | Darren_Duncan++ | deleted /misc/Rosetta-Incubator, since the remnants it contains (SQLBuilder, SQLParser, Eng/Gen, Emu/DBI) have dubious value
r8646 | putter++ | docs/AES/S29draft.pod - corrected typos - adverbs missing their colon.
luqui ponders writing a JS compiler to pil; then we could run JS on JS! 21:26
gaal luqui: I think putter is working on that :) 21:27
luqui oh, seriously?
luqui reads over S29, never having done so before 21:28
putter misc/JavaScript-FrontEnd/ ! the grammar is as yet untested (only root level rules currently work in pugs). one could flatten it and test! the runtime and compiler is largely an automated transform on the spec. I havent been keeping the checkedin copy in sync. interested?!
;) 21:29
luqui wow
not really.
but wow, cool
In addition, the list of functions imported into C<::*> will be 21:30
subject to change with each release of Perl.
is that a blessed policy?
putter ah well. the general observation is just that its very easy to do language implementations when your target environment capabilities are a strict superset of the language being implemented.
it would be neat if one could do use Perl3.02; ... ;) 21:32
stevan tosses some LISP into the room and runs out the door
sial.org/pbot/15425
luqui yeah, "neat"
stevan really needs a new hobby 21:33
putter lol
gaal notices it's been a while since stevan made a movie 21:34
stevan wonders how got a hold of those video tapes? I thought I hid them better 21:35
putter fun code. I'd actually really like to have CL front and backends.
stevan I am really digging CL right now
luqui I expect that those would not be that hard to implement
at least the frontend
stevan luqui: it would probably be really easy to make PIL^N VM in it too 21:36
putter I have most of a R5RS scheme on p5 I occasionally putter on. "SchemeOn". tiny p5 kernel and a scheme->p5 transliterator written in scheme. and everything else. biggest challenge was gathering together clean scheme-written-in-scheme code fragments. it was quite a zoo. 21:38
CL might be better. hopefully much better. todo list item is to check out SBCL's code base more closely.
stevan suspects putter might need another hobby even moreso than stevan :) 21:39
putter let's do it! well, we need piln for it to run. but the gathering CL code can be done ahead of time. 21:40
putter is engaging in job search avoidance.
:(
clkao smoke still broken 21:41
pugs: *** Can't locate Test.pm in @*INC (@*INC contains: blib6/lib /usr/local/lib/perl6/mach /usr/local/lib/perl6 /usr/local/lib/perl6/site_perl/mach /usr/local/lib/perl6/site_perl /usr/local/lib/perl6/auto/pugs/perl6/lib /usr/local/lib/perl6/site_perl/auto/pugs/perl6/lib .). everywhere
stevan is off to parts-unknown &
putter bye stevan &
gaal clkao: multi subs still not exported
but that error shouldn't happen
clkao why isn't test.pm in lib6 21:42
gaal blib6/lib
should be
clkao there's only blib6/lib/Test
gaal worksforme... 21:43
blib6/lib/Test.pm you mean? that's ok
or do you have blib6/lib/Test/lib ?
blib6/lib/Test/ is for Test::Builder, not what you need 21:44
you ought to have a Test.pm in blib6/lib
clkao no i don't
gaal du -s blib6
I have ~4.6M 21:45
clkao it's mostly empty. only .exist
but there is some hierarchy 21:46
gaal clkao: what OS are you on?
clkao freebsd. same on darwin iirc. er, there's about 3M
but no Test.pm under blib6/lib 21:47
gaal anyway to solve your immediate problem so you can smoke, please export PERL6LIB=ext/Test/lib
21:47 brentp joined
gaal if you have time, look for errors in make, or try a really really clean copy, etc. 21:47
find blib6 | wc -l => 545 here 21:48
clkao too late, made realclean 21:50
i should sleep and leave it running
it seems to do it. make now copies Test.pm
how stupid is it
anyway, thanks gaal 21:51
zzz & 21:52
gaal night :)
putter does map work on an array's values, or its indexes (regardless of whether they .exist or not)? 21:58
luqui good question
given that there's no way to test existence within the body of map, I'd say its values 21:59
but it's tricky
gaal when does an array's index not .exist?
only when it's out of bounds, but then the map doesn't reach it. no?
svnbot6 r8648 | Darren_Duncan++ | r1935@Darren-Duncans-Computer: darrenduncan | 2006-01-13 13:57:10 -0800
r8648 | Darren_Duncan++ | /ext/Rosetta : updated Copying.pod to demote 4 packages out of the defined framework core; only 4 remain
luqui I think it's a bad idea to keep that distinction...
perl -le 'my @a = (1,2,3); delete $a[1]; print exists $a[1]' 22:00
putter gaal: when one .delete($idx)s? I've no idea.
Khisanth but that should still exists 22:01
gaal Khisanth: unless it was the last element, IIRC 22:02
Khisanth yes
22:02 kanru joined
Khisanth so you are modifying the array as you iterate over it? 22:02
gaal it makes some sense to me, but I'm perlverted. 22:03
22:05 f3l joined
putter ha! S29 has an implementation I overlooked (woops). and the verdict is... 22:05
luqui nobody possibly could have thought of this issue before 22:06
plus, for a description of the crazy inter-array existence semantics, see perl 5 (my example)
putter whatever *@array yields... 22:07
sigh
the delete $a[1] example? why is that odd? ;) 22:08
luqui because $a[0] exists, $a[2] exists, but $a[1] doesn't 22:09
I know you know putter, I was responding to gaal and Khisanth
gaal well, it surprised me but I'm not surprised to be surprised. like I said, perlverted. 22:11
svnbot6 r8649 | luqui++ | Minor editing on S29. 22:14
22:17 wilx joined
putter ooo, a S29 edit. putter goes to see... 22:18
leo_ luqui: is above array.delete semantic also confirmed for p6? It looks weird to me - I thought it's like splice (no holes) 22:21
luqui I sure hope not, it looks very weird to me too 22:22
but I think it hasn't been confirmed or denie 22:23
d
leo_ it's absolutely senseless in combination of native types - there isn't a 'no int' - it's kust maybe zero 22:24
just 0 even
luqui Maybe 0, Just 0, but maybe it's Nothing...
hmm... Maybe 0... the type of things that might have type 0 22:25
luqui smacks himself in the monads with a lambda (and is getting rather accustomed to doing that)
leo_ well, native types don't have anything like nothing 22:26
luqui knows
luqui was just being silly...
I'll fire off a message to p6l about array holes
not that we can hope to get a response from $Larry because he's busy
leo_ thanks 22:28
putter thought he had just finished with Array. luqui++ just changed the signatures. :) such is life.
luqui just cosmetic stuff though, like CALLER<$_> to $+_ 22:29
gaal (enviroment vars)++
luqui ditto
putter luqui: what's wrong with the Array::pairs etc signatures? 22:33
luqui the reason I said XXX wrong is because we're not sure what's right yet
but it has to do with contextual multi dispatch 22:34
those probably suffice for descriptive purposes, but certainly not for implementation purposed
s/sed/ses/ 22:35
putter ok. do they run? never mind. I'll have to check the whole file. another day.
22:45 sproingie joined 22:46 Limbic_Region joined
svnbot6 r8650 | putter++ | src/PIL/misc/prelude/Array.pm - Created. The beginnings of a p6 Prelude for piln. Combines code from PIL2JS, Prim.hs, S29draft, etc. Intended to help bootstrap piln. 22:48
r8650 | putter++ | Todo: Finish up Array.pm coverage. Then test it, and write alternate implementations for anything pugs currently chokes on. Then on to List, et al. Err, "Help!". ;)
putter luqui - S29 doesnt actually say whether keys, kv, etc return List or Array. List? 22:53
no, I have that wrong. 22:54
I guess I was just confused. tnx.
gaal I think it's Juerd that likes to point out that you can't return Arrays in Perl, just items or lists. 22:55
s/that/who/
22:57 Supaplex joined
putter ah. though with auto-refifying, it sure looks like you do. ;) 22:59
svnbot6 r8651 | putter++ | misc/prelude/Array.pm - fix some return types.
gaal putter: indeed
good thing at least that we no longer call it "scalar" context! 23:00
putter <chuckle> indeed 23:02
23:08 beppu joined 23:14 binary42 joined 23:15 bpalmer joined 23:19 dduncan left
putter loqui: along the lines of your removing the <Array>'s, should &join<> := &join<Str>.assuming:delimiter(' '); be &join := &join.assuming:delimiter(' '); ? 23:24
loqui: and is there a reason map and reduce use different conventions for their Code arguments? 23:27
duh. sorry. 23:29
luqui: two questions above, misidentified loqui. end of a long week. 23:30
23:30 lisppaste3 joined
luqui re map and reduce: looks like an oversight to me. they should both be like reduce 23:34
putter, re <Array>s, the &foo<Bar> syntax is no longer correct; it is &foo:(Bar)
I think... 23:35
in any case, you shouldn't be currying the variant but the operation as a whole
svnbot6 r8652 | gaal++ | * Pugs.Prim.Yaml - very minor golfage
luqui and IMO, things like join, map, etc. shouldn't be multis in the first place
but that's something I have to argue with @Larry about
putter thanks! And I assume multi Lists::zip should be multi List::zip (singular)? 23:37
svnbot6 r8653 | gaal++ | Pugs.Parser - 23:38
r8653 | gaal++ | * Introduce Pugs.Parser.Export, because 3 is better than 7 when it comes
r8653 | gaal++ | to indent levels.
gaal hmm, *2* is better than 7 even. 23:39
since I can't count, perhaps I should go to sleep. 23:40
putter lol :) 23:41
gaal btw this still segfaults for me: 23:42
?eval say "mmm" if "a" ~~ /a/
(evalbot's dead?) 23:43
putter embedded or external parrot?
gaal anymoose, it's late. see you tomorrow in releng!
zzZ &
putter good night gaal &
svnbot6 r8654 | putter++ | AES/S29draft.pod - additional tweaks. 23:51
putter luqui: should 23:52
multi Array::grep (@values, MatchTest $test --> Lazy )
instead be
multi Array::grep (@values: MatchTest $test --> Lazy )
?
23:59 scook0 joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!