The Perl 6 Design team met via phone on 29 November 2006. Larry, Damian, Allison, Patrick, Jesse, and chromatic attended. These are the notes.
The Perl 6 Design team met via phone on 29 November 2006. Larry, Damian, Allison, Patrick, Jesse, and chromatic attended. These are the notes.
Patrick:
- things are going well
- keeping busy
- making good progress
- made the new PAST implementation
- seems to be working out really well
- has the right primitives in the right places for the kinds of compilers I've been working on
- Allison's given me tons of great feedback
- I'm slowly incorporating her ideas
- having tons of fun doing all that
- have lots more of the sanity tests implemented for Perl 6
- have some implemented but not checked in
- I'm letting that simmer before I check in improvements
- finding a couple of sanity tests that build on non-standard Perl 6 features
- I'll do some commits to bring them in line with the latest specs
- hope to have that working in the next couple of weeks
- could start running the Pugs test suite then
- might just point people to my #parrotsketch reports (pmichaud #parrotsketch report for 21 November 2006 and pmichaud #parrotsketch report for 28 November 2006)
- spent an hour walking PerlJam through the compiler tools on #parrot
- I logged it; it'll go into the repository as an overview
c:
- think I've basically convinced Allison of the value of a MOP for Parrot
- will work on that a little bit more
- I had a prototype a couple of months ago
- I'll refresh that
Allison:
- Patrick's stuff has kicked forward the design of the compiler tools
- I've played with it in Punie; have lots of feedback on it
- I'm just about ready to commit it
- I have the last lingering little bits of the IO PDD
- it's pretty much just the Parrot IO object API
- not a big piece
- reading a bit on Ruby metaclasses
- we had a really good conversation about Parrot objects after #parrotsketch
- I have a much better grasp on the scope of the problem
- some of the most important questions
Patrick:
- that's an astounding achievement
- it's difficult to do
- it's also very important and necessary
- I'm ecstatic with the conversation that took place
- that's exactly what I wanted to hear
Allison:
- I expect we'll have snippets like that every once in a while
- that seems to be really helpful
- oh, and maybe we'll use roles instead of metaclasses
Larry:
- keeping my fingers lightly on everything that's going on
- thinking a lot about discussions on p6l
- come to a conclusion on bags and sets
- haven't come to a conclusion about some of the questions of closures
- touches on the nature of dynamic versus lexical scoping
- the relationship of those versus STM
- some of those popped up for Parrot lately too
- did a lot of driving last week for some reason
- also converting some of my CJK radical dictionary programs to Perl 6
- fairly happy with those results
- anxiously looking forward to running those on the Parrot backend
Allison:
- it would be interesting to get a list of features that you need
Patrick:
- that would prioritize things
- or turn the code into tests somehow
- I can target those tests first
Larry:
- other than the basic syntax
- the main thing I've used is roles instead of inheritance
- the roles turn out to be somewhat faster than inheritance
- composition time is more direct
- it's a hand optimization
- in theory the compiler ought to perform it for me, but it doesn't
- I'm not performing much fancy stuff
Patrick:
- the big stumbling block for now is the angle bracket as list constructors
- parsing that is a trick
Jesse:
- how much spec change do you end up making as you hack?
Larry:
- not a lot
- we're definitely in the mode where we try to fill in the cracks that we didn't think about
- those implicitly implied by the things that are already there
- the earth-shaking changes tend to be few and far between
- mostly it's just going over it to see if we're saying everything we want to say
Damian:
- still struggling with time issues
- within one feature of having the POD 6 parser released
- likely a week or two away
Patrick:
- I have a question about coordinating test suites
- as I get close to finishing the sanity tests for the Perl 6 compiler, people will clamor for importing the Pugs test suite into the Parrot repo
- should we clarify that somehow?
- any advice on how to proceed?
Allison:
- the answer is "yes, we can just import the test suite"
- Audrey is okay with that
- we need to try to get contributor agreements from people who have committed to the test suite
- we don't have to wait to get those first
- we do need to try to get the names of contributors
- we'll have someone gradually work on that
c:
- can we use
svn:externals to keep them synchronized?
Jesse:
- I'd hate to see them fall out of synchronization
Patrick:
- I'd like to see a single test suite
- I have difficulty seeing how to coordinate working between two repos
Jesse:
- I hear that the Pugs folks are willing to do plenty of work to make that work
Larry:
Allison:
- they do need to ship with stable Parrot releases
Jesse:
- you can do imports or mirroring
Allison:
- you don't want to require everyone to be a committer on the Pugs project to contribute to the tests though
- even though it's trivially easy
- Perl 6 and Parrot are two independent projects
- maybe we need an independent repository for the Perl 6 test suite
- neither Pugs nor Parrot
- make sure that the infrastructure code for both call into the implementation-agnostic test suite repository
Jesse:
- I worry about what that'll mean for commit bits about the independent test suite
- Pugs does have mirroring into svn.perl.org
Patrick:
- maybe I don't understand how it works
- if I'm working in Parrot's Perl 6 and I need to add a test, what do I do?
- do I svn up for Pugs, add a test, commit a test, then svn up for Parrot?
Jesse:
- would you object to always using the tests directly from the Pugs repository until something sane gets worked out?
- it's not a discussion we can make a decision on here now
- it has to involve the Pugs people too
Allison:
- that's fine
- I'd eventually like the test suite to be pure Perl 6
- or at least abstracted from any implementation infrastructure
Jesse:
- seems to be a reasonable goal once it's not a moving target
Allison:
- don't wait for that to happen!
Jesse:
- I don't want to yank the Pugs test suite, make it the official test suite, and then make it hard for them to contribute to it
Allison:
c:
- we have a repository for the specification documents
- why not put the test suite in that repository?
- not everyone can commit to the specs and Synopses, but that repo
Allison:
- how about this?
- Pugs remains for now the primary repo
- all commits to there
- the Perl svn repo gets a mirror of those tests (not the full Pugs mirror)
- we work on the tests to remove any code specific to a particular implementation
- get the tests generic when we can
general agreement
