The Perl 6 Summary for the week ending 2005-07-19

Welcome to another Perl 6 Summary, brought to you by the words 'displacement' and 'activity'. So far today I've caught up with everything unread in NetNewsWire, my Flickr groups, every other mailing list I'm subscribed to and completed about 5 Sudoku. And now I'm dragging out this introduction and I don't know why; I enjoy writing these things.

This week in perl6-compiler

Another quiet week on the list. However, you only have to watch the svn commit log and the other stuff on PlanetSix to know that things are still proceeding apace. Last time I looked it seemed that Stevan Little was working on bootstrapping the Perl 5 implementation of the Perl 6 MetaModel so that it's implemented in terms of itself.

Rather mindbogglingly, Pugs is now targeting Javascript as well.

The Current Pugs release is 6.2.8

http://planetsix.perl-foundation.org

Creating threads in BEGIN

Nicholas Clark posted what he described as a 'note to collective self' wondering about how Perl 6 will cope with people creating threads inside BEGIN blocks. According to Luke "it won't". Larry thought that it might be okay to create threads at CHECK time, so long as any spawned threads didn't do any real work later than CHECK time.

http://groups.google.com/groups?threadm=20050714153213.GB73297@plum.flirble.org

Perl 6 Modules

Gav... (I presume the ellipsis is important to someone) wondered what he needed to do to write Perl 6 modules. Nathan Gray pointed him at the porting howto in the pugs distribution.

http://svn.openfoundry.org/pugs/docs/other/porting_howto

http://groups.google.com/groups?threadm=002a01c58944$c9231c30$0200a8c0@madaboutipv6

Is namespace qualification really required?

Phil Crow came across some weirdness with namespace resolution. It seems that you have to explicitly qualify function names in signatures. Autrijus agreed that it was a bug and asked for Phil to write a TODO test. Discussion ensued -- I think it's fixed in SVN now.

http://groups.google.com/groups?threadm=200507151916.j6FJGm18017026@webmail.sunflower.com

Parsing perl 6 grammars

Nathan Gray wondered about the state of Jeff 'Japhy' Pinyan's effort to implement a Perl 6 rules parser. Japhy said that it's been on hold for a while, but that he'd started to work on it again, basing it on his earlier Regexp::Parser module.

http://groups.google.com/groups?threadm=20050719180105.GA30311@vs2.capwiz.com

Meanwhile, in perl6-internals

PMC changes?

Nicholas Clark wondered if the PMC layout is likely to remain stable, or if there might be changes in relation to the generational garbage collector. In particular, he wanted to know if the API would remain stable. Leo thought that there might be changes in the layout, but the API shouldn't change.

http://groups.google.com/groups?threadm=20050713142025.GR73056@plum.flirble.org

ParTcl Accelerator

Will Coleda showed some timings for ParTcl, the Parrot implementation of Tcl and made a few suggestions about how to get things going faster. Patrick and Leo mused on the issues involved.

http://groups.google.com/groups?threadm=643ED742-47AA-4D16-8A8A-BB956F020E90@coleda.com

Partitioning PMCs

Nicholas Clark had some questions about making PMCs and Ponie play well together, with particular reference to using SvFLAGS().

http://groups.google.com/groups?threadm=20050713140849.GP73056@plum.flirble.org

Embedding/extending interface

Nicholas Clark wondered if chromatic was still waiting for confirmation that his automated embedding tools were the Right Thing. Apparently, chromatic is waiting for confirmation, but offered to send his existing patch, if only to force the discussion.

http://groups.google.com/groups?threadm=20050713144138.GT73056@plum.flirble.org

Ponie Questions

Nicholas Clark had a bunch of questions about various subsystems, mostly in relation to Ponie. Leo came good with answers.

http://groups.google.com/groups?threadm=20050713144716.GU73056@plum.flirble.org

http://groups.google.com/groups?threadm=20050713143805.GS73056@plum.flirble.org

http://groups.google.com/groups?threadm=20050713141314.GQ73056@plum.flirble.org

Parrot Project Management

I'm not sure if Will Coleda's suffering culture shock about the way Parrot project management is done, or if we're really not doing it right. The first rule of Parrot/Perl 6 development is that if you really want something then the only way to guarantee that it gets done is to do it yourself. It's certainly worked for me over the years.

http://groups.google.com/groups?threadm=cff4c4c41b42a1f39d7d4fe68f914841@perl.org

Tcl GC issues... solved

Matt Diephouse announced that as of r8617 in svn, the longstanding GC bug that was occasionally tickled by ParTcl has been fixed. There was no rejoicing on list, but at least one summarizer was really pleased to hear it.

http://groups.google.com/groups?threadm=198c87380507140823527d4165@mail.gmail.com

GMC for dummies

Summer of Code intern, Alexandre Buisse, who is working on a new GC system for Parrot pointed us all at an introduction to the Generational Mark and Compact scheme that he's working to implement. He and Leo had a discussion about implications, assumptions and other stuff.

Bob Rogers asked some tricky questions relating to circular structures and timely destruction. Discussion of this continues.

http://groups.google.com/groups?threadm=42D853A9.8060902@gmail.com

http://perso.ens-lyon.fr/alexandre.buisse/divers/gmc_for_dummies.pod

Register Allocation fun

There was a flurry of patches from Curtis Rawls who appears to be working on refactoring and (one hopes) fixing the IMCC register allocator. Way to go Curtis.

Meanwhile in perl6-language

MML dispatch

The ongoing discussion of the right way to dispatch multimethods is still, um, going on. Damian and Luke appear to have a fundamental disagreement about what's the Right Thing. "Manhattan!" "Pure!" "Manhattan!" -- it's not quite that bad, but positions seem to have been entrenched for a while. Elsewhere in the thread, Larry mused on which was more general, classes or roles. Thomas Sandlaß wondered how they stood in relation to types.

Your summarizer wondered how he was ever going to explain all this and punted.

http://groups.google.com/groups?threadm=42D33D33.6090305@conway.org

Method calls on `$self`

My eyes, they burn! At this rate I'm simply going to use $?SELF in all my Perl 6 classes. Larry's latest suggestion seems to please even fewer people than ./method, which is really saying something. As someone who's not a fan of ./, I was slightly surprised to find myself agreeing with Autrijus who reckons you get used to it really quickly.

http://groups.google.com/groups?threadm=20050712105922.GH7720@c4.convolution.nl

The Perl 6 Library system

In response to a question from Autrijus about coderefs in @INC or whatever Perl 6 is going to call it, Larry mused on the possible eventual design of Perl 6's library system. It seemed to me that he was dropping a rather heavy hint to any interested readers who might like to come up with a first cut of Perl 6's library system.

Later he did some thinking aloud about treating strings as arrays or vice versa.

http://groups.google.com/groups?threadm=20050714184914.GB11335@wall.org

http://groups.google.com/groups?threadm=20050713181410.GA5960@wall.org

Method Resolution Order

Stevan "MetaModel" Little cheered Larry's statement that methods, subs, submethods and "anything in between" all live in the same namespace. If you want to give two code like things the same name then you have to explicitly declare them as multi.

Stevan went on to ask a bunch of questions about the semantics of method resolution so as to get the Perl 6 MetaModel working right. Discussion ensued.

http://groups.google.com/groups?threadm=b61699c72578bbd844606f7cae1399b3@iinteractive.com

Type::Class::Haskell does Role

I haven't the faintest idea what Yuval Kogman is talking about. Dammit I need to learn Haskell now. Luckily Autrijus, Luke, David Formosa and Damian did seem to understand it. There was ASCII art and everything. Sadly no Unicode art, but it's only a matter of time.

http://groups.google.com/groups?threadm=20050714233822.GB11028@woobling.org

Optimization pipeline

Yuval Kogman posted an outline of the optimization pipeline idea that he'd brought up in a hackathon. If confess that it looks rather like something that got discussed a few months (years?) ago which Chip had shot down rather convincingly. (I remember this because I was taking pretty much the same position as Yuval, and I really didn't want to be convinced...)

http://groups.google.com/groups?threadm=20050715000522.GC11028@woobling.org

STM semantics, the Transactional role

Yuval Kogman discussed some issues with STM (Something Transactional Memory -- I'm hazy on the 'Something' and far to lazy to Google). A short discussion ensued.

http://groups.google.com/groups?threadm=20050715214414.GF11028@woobling.org

Accessor only virtual attributes

Sam Vilain wondered what would happen if he made an 'accessor' for an attribute that didn't really exist. He wanted to be able to disguise accessor methods as attributes within class and subclass scope (at least, I think that's what he wants). Larry seemed to think he was barking up the wrong tree -- class attributes are only likely to be accessible using the $.whatever form within their declaring class and not any subclasses. Larry's 'got some driving to do' so expect some more thoughts about this in the next summary.

http://groups.google.com/groups?threadm=42DB234C.6050106@vilain.net

Strange interaction between Pairs and named binding

Autrijus noted that, although:

    for [1..10].pairs -> Pair $x { say $x.value }

works,

    for [1..10].pairs ->      $x { say $x.value }

doesn't. Which is somewhat counter intuitive. The problem is, the pair is being treated as a named argument specifier in the second case. After discussion, Autrijus suggested that the best thing might be to specify that the 'Bare' code object (which includes pointy and non-pointy blocks) have 'Any' as a default parameter type -- essentially turning off the special behaviour of Pairs when calling named blocks. I'm all for this myself, but Larry has yet to speak.

http://groups.google.com/groups?threadm=20050718124224.GA84092@aut.dyndns.org

How do subroutines check types?

Ingo Blechschmidt had some questions about specifying types in subroutine definitions. Specifically he wanted to be able to specify that a sub only take instances of a class Foo and its subclasses but not the class Foo (or its subclasses) itself. Thomas Sandlaß thought that what Ingo wanted is the default behaviour and you actually have to do some work to get it to behave any other way.

http://groups.google.com/groups?threadm=loom.20050719T143727-491@post.gmane.org

Referring to package variables in the default namespace

Matthew Hodgson asked for some clarification of how the default package namespace works. Apparently pugs and Synopsis 10 are slightly at odds. Larry had some answers. Matthew probably has some more questions.

http://groups.google.com/groups?threadm=Pine.LNX.4.61.0507191912180.11917@phoenix.arasphere.net

Crikey! That went quickly

Or, for the traditionalists among you

Acknowledgements, adverts, apologies and alliteration

http://dev.perl.org/perl6/ -- Perl 6 Development site

Check out my website, now running on a different blog engine and adorned with some new articles. Will wonders never cease?

http://www.bofh.org.uk/

Vaguely pretty photos by me can be found at:

http://www.flickr.com/photos/pdcawley