Thank you for a lovely Christmas Present.
Michael
On Dec 24, 2003, at 2:37 AM, The Perl 6 Summarizer wrote:
The Perl 6 Summary for the week ending 20031221
Welcome one and all to the penultimate Perl 6 Summary for 2003. The
nights are long, the air is cold, freezing fog made the journey
home
from watching *The Return of the King* a deeply fraught experience
(though probably not as fraught for us as for the chap who
obviously
didn't see the roundabout in time and went straight over it).
If you're in the southern hemisphere, feel free to ignore the
foregoing
flavour text.
It should come as no surprise to you by now that we start with the
internals list.
Namespaces II
Dan returned from a bout of paying work to finish up the current
understanding of how namespaces are managed in Parrot. A namespace
selector becomes a multidimensional key (using a multidimensional
key
instead of a single long namespace name makes for easier sharing of
namespaces between languages; there's no need to go guessing what
separator the other language uses in its namespace naming because
that's
essentially irrelevant at the Parrot level). Looking up a
particular
variable in a namespace is still a two step process, first get the
namespace, then do the lookup within the namespace. Dan explained
that
this is a win because it should allow for holding onto namespaces
in a
register when dealing with multiple variables.
Leo Tötsch wasn't entirely convinced by the new syntax introduced,
and
proposed a mechanism which used the standard syntax for accessing
multikeyed structures with a proposed Namespace PMC.
http://tinyurl.com/2wtkd
Another keyed ops proposal
Leo continues to make proposals for rethinking keyed ops. Somewhat
surprisingly, Dan didn't entirely reject the latest one involving a
possible new set of key registers.
http://tinyurl.com/2za5t
Leo adds threads
Leo Tötsch checked in a patch to allow Parrot to spawn threads. As
he
admitted, there's a lot missing, but it's a start. He solicited
comments
and further pointers.
Later in the week he asked for some design decisions related to
making
various of the interpreter's internal data structures thread safe.
Elizabeth Mattijsen had several comments to make, based on
experience
with Perl 5's ithread system. In essence her suggestion was that
as many
things as possible should be made copy on write between threads.
(I have
a faint recollection of Nicholas Clark delivering a wonderful
lightning
talk/dance explaining his proposal for making Perl 5 use copy on
write
structures for ithreads. It's worth seeing if you get the
chance...)
http://tinyurl.com/3egyu
http://tinyurl.com/yrjrm
PDD03 and method calls
Leo queried the design of Parrot's calling conventions for calling
a
method and proposed a slightly different convention with the
object put
in P5 rather than P2. Dan thought that the calling conventions
documented was the right way to go, but didn't seem to convince
everyone.
http://tinyurl.com/2npnn
Rolling back "clone"
Leo pointed out that "clone"'s semantics had got altered to take an
uninitialized destination PMC as an argument back when the GC
system was
flaky. He suggested switching back to a version that creates the
new PMC
itself now that GC was working properly. Dan told Leo to go for it.
http://tinyurl.com/35c57
Meanwhile, in perl6-language
Um... sorry about this, but I'm punting on writing the promised
Roles/Traits/Classes summary. I was planning on doing it today, but
things have been hellish; I've spent about 8 hours fighting fires
today
rather than writing the summary. Sorry.
Work on Perl 6's object system continues apace though; things are
looking very good and powerful.
By the way from now on I'll try and stick to Larry's usage of
capitalizing 'Traits' when referring back to the Traits paper, and
keeping it lower case when referring to Perl 6's compile-time
traits.
Runtime CHECK?
Piers Cawley asked if Perl 6 modules that get loaded at runtime
will
have their "CHECK" blocks (or their equivalent) executed. Larry
initially thought not, but Rafael Garcia-Suarez noted that Perl 5
shouldn't be held up as exemplary and that there was definitely a
need
for some kind of special block that would get run at the end of the
compilation phase of any arbitrary compilation unit.
http://tinyurl.com/2ytze
"but true"
Adam D. Lopresto wondered how the recent work on Roles, Properties
and
Traits fit in with already declared stuff like "but true". Larry
confessed that he was still thinking hard about this, but that he
thought "but" wasn't powerful enough yet.
http://tinyurl.com/2v2ef
Acknowledgements, Announcements, Apologies
Cross your fingers
