This week on Perl 6, week ending 2003-05-04

Welcome my friends, to the show that never ends. Yes, another week, another Perl 6 Summary, chock full of condensed goodness, Leo Tötsch admiration and a smattering information about the design and development of Perl 6 and its target virtual machine, Parrot.

A quiet week this week. Even the hotbed of discussion that is perl6-language saw fewer than 100 messages. However, in accordance with tradition, I'll start with perl6-internals, which saw all of 47 messages this week, none of them from Leon Brocard.

External Data Interfaces draft PDD

Discussion of the External Data Interfaces PDD continues. Hopefully we'll see the first 'real' version soon.

PMC Keys

Alberto Simões asked for a good description of PMC Keys. No answer yet.

http://groups.google.com/groups

Long option processing

Luke Palmer sent a patch to do long option parsing. Again, Warnock's Dilemma applies.

http://groups.google.com/groups

Problem with readline

Will Coleda announced that he was dusting off his TCL project and found that it threw lots of bus errors. He tracked the problem down down to the readline op. Benjamin Goldberg realised that what was happening was that a filedescriptor was being used as a pointer to a FILE datastructure. Which is never going to be good. (This would never happen in a language with typed values. Well, it might, but the error would be caught before the bus error). No fix yet.

http://groups.google.com/groups

Read buffering in PIO

Possibly prompted by Leon Brocard's nudge the week before, Dan Sugalski took another look at Jürgen Bömmels' rejected patch to add read buffering to the Parrot IO subsystem. Apparently there's a problem with a double free introduced by the patch. After some discussion (including a contribution from Melvin Smith, PIO's original author) of how to address the issue, Jürgen submitted another patch.

http://groups.google.com/groups

Excessive memory usage?

Peter Gibbs 'showed off' a short piece of PASM code that managed to use some 60Mb of memory and to allocate 1.5 million headers. He thought this a little excessive. No comment yet.

http://groups.google.com/groups

Extending pop

Klaas-Jan Stol wondered if it would be useful to have a variant of pop which could pop multiple items off the runtime stack. Dan thought it as a good idea and added a bunch more useful tricks involving stack marks and things, and asked for volunteers to implement it. Nobody has explicitly stepped up to that particular plate...

http://groups.google.com/groups

Clint Pierce shows off his 'mad NCI skeelz'

Clinton A Pierce has been playing with NCI on Win32 and has now got his Parrot BASIC calling Win32 functions natively. ``Mua-hahahaha'', as he so eloquently put it. He's now looking for a generic way to allocate a generic memory area in PASM for use as the target of a function, which should allow him to make even more Win32 calls without having to write an adaptor library in C first.

http://groups.google.com/groups

Dan changes the calling conventions again.

Dan released the final set of calling conventions again. He thinks he won't have to do this again. Again. The changes are all in PDD03 (docs/pdds/pdd03_calling_conventions.pod in the CVS version of the Parrot distribution). The big change is that we no longer use the stack at all for passing parameters, we use an overflow array instead. The smaller change is that the PDD has been clarified somewhat.

http://groups.google.com/groups

And that about wraps it up for the internals mailing list. However, Dan has been publishing some handy stuff in a new ``What the heck is ...?'' series on his 'Squawks of the Parrot' website.

http://www.sidhe.org/~dan/blog/archives/000174.html - Walking the stack

http://www.sidhe.org/~dan/blog/archives/000178.html - Coroutines

Meanwhile, over in perl6-language

There was lots more about types. And some new stuff too...

``I thought there was no bool type?''

Towards the end of last week, Smylers queried the 'bool' entries in the latest draft of Michael Lazzaro's Type Conversion Matrix. The thing is, Larry has said that there won't be boolean type. Smylers was not the first person to make this comment, he probably won't be the last. The stock response is along the lines of ``There's no bool type, but there is bool context.'' but according to Synopsis 6, this appears not to be the case anymore.

Then it all got a bit weird. Nobody quite asked ``What is truth anyway?'' but it was touch and go for a while as people discussed what the value of a bool would be in a numeric context. (I had a good deal of sympathy for the view that there should be some kind of warning...). We eventually ended up in a discussion of multistate logic (which, it seems to me is a candidate for 'something you implement in a module' status.), who knows where we'll end up this week.

http://groups.google.com/groups

Headers

Paul Hodges wondered if it would be possible to write something in Perl 6 that worked in a similar way to the C Preprocessor's #include directive, allowing him to push a common set of use statements and other compile time declarations into a header file, which could then be shared by multiple bits of code. Various people suggested more or less tricky options, but Marco Baringer won the 'simplest thing that could possibly work' prize from your summarizer with:


    macro include ($file) {
      join '', <open $file>;
    }

which does the job very straightforwardly.

http://groups.google.com/groups

Type Conversion Matrix (Take 3)

Michael Lazzaro posted his third attempt at a type conversion matrix. This triggered some discussion on the difference between primitive (int, float, etc) and 'full' types (Int, Float). Michael worried that some of the discussions were bloating primitives leading him to wonder what would be the point of using them if that happened.

It's apparent from the discussions here that the distinction between storage and value types enables a good deal of magic, but that scares people too.

http://groups.google.com/groups

Property Inheritance

David Wheeler popped up with something of a head scratcher. He wanted to know whether, when a method is overridden in a subclass, the overridden method inherits the traits (he said properties, but he meant traits I think) of its SUPER method. Luke Palmer thought it probably depends on the trait, but his guess was that traits would be inherited. Which led David to wonder if that meant you could override traits in a child class.

http://groups.google.com/groups

OO-overhead

For some reason, a discussion on structuring OO code in Perl 5 has been running in perl6-language for a while. It's handy for spotting issues, but not exactly on topic. It is to be hoped that Perl 6 will lose a lot of Perl 5's function and method call overhead though.

Chaining postconditionals

Michael Lazzaro asked for the rationale behind disallowing stuff like:


    return if <expr> for <list>

Short answer: Because Larry said so.

Supporting evidence (from the thread that developed anyway):


    foo $_ if baz for @list unless quux.next() while 1;

or:


    if $X {...} if $Y

http://groups.google.com/groups

Coroutine calling convention

Luke Palmer kicked off some discussion of the various coroutine calling conventions that Dan had discussed on his Squawks of the Parrot website. Luke though that coroutines should return iterator objects instead of the yielded value. Which would mean they weren't actually coroutines...

http://groups.google.com/groups

http://www.sidhe.org/~dan/blog/archives/000178.html - Dan talks Coroutines


Acknowledgements, Announcements and Apologies

So, another summary draws to a close on a glorious May afternoon. Here's to the next one. Thanks to those people who sent the proof I asked for in the last summary. No thanks to the gits who outnumbered them by sending spam to the same address.

If you've appreciated this summary, please consider one or more of the following options:

Visit the home of the Perl programming language: Perl.org

Sponsored by

Monthly Archives

Powered by Movable Type 5.13-en