The Perl 6 design team met by phone on 08 July 2009. Larry, Allison, Patrick, Jerry, and chromatic attended.
Larry:
- sided by several design issues in the past week
- the number of fundamental issues was higher than usual
- refactoring the global and lexical namespaces to fit together with their symbol tables
- Daniel pointed out a problem in my design
- immutable type objects pointed to mutable symbol tables
- that didn't quite work
- had to turn that inside out
- thinking about the exact semantics of
use
- depends on how symbol tables work out
- should have that soon
- thought a lot about the nature of lists
- are they more primitive than captures?
- had some discussion about that
- provisionally called
Parcels
- may be correct to unify them with captures
- a capture might just be a parcel that can deliver the positionals separately from the names
- the actual structure might be the same thing
- Daniel might have a complaint about it
Patrick:
- I totally agree with the change
- I just know it's going to muck up a few internals along the way
Larry:
- we need something to give us the raw arguments in their order
- has to be what parentheses do by default
Patrick:
Larry:
- that part is not negotiable
- working on the trait mods
- Jonathan's been refactoring that
- still have a collision as a result of defining multis with the same name as we define the new syntax
- may have to split those
- the macroish thing which defines new syntax is different from the semantic declaration of the multis called
- haven't specified that, but it seems like the right approach
- encouraging other people to hack on the specs
- encouraged Darren Duncan to swap
:auths and :vers
- he did other cleanup, which was good
- other people have contributed too
- doesn't seem to be out of hand
- clarified value-producing blocks like
BEGIN and START as statement prefixes
- you don't need curlies now
- you can still do the block form, but they're block or statement
- you can use a statement prefix as a statement itself
- the old model of symbol tables in STD used nested hashes
- using a model now where symbol tables are objects which can function as hashes
- recombined symbol tables and stashes
- we can
does Hash usefully and get the same benefit in a single object
- moved back to the notion that the symbol for a symbol table in a package ends in
::
- tried to get away from that, but it didn't work
- it's a post-sigil, now
- the name canonicalizer in
Cursor now leaves the double-colon at the end of prefix packages
- makes it simpler to look up
- fixed a bug related to pad exporting
- chose the wrong pad
- now the declarand (the declared object) remembers the appropriate pad for the rest of the declaration
- if you put
=$fh as the only thing as a line, now it complains about an illegal POD directive
- changed constants to be
our instead of my, last week
- more chance of name collision within packages and lexical scopes
- I fixed the error message for a collision
- aligned a few more things to the specification
- changed MMD with Jonathan
- mandatory, named parameters are part of the constraint on a multi
- where STD uses that, it declares those parameters as mandatory
- the spec claimed that adverbs and fat-arrow plus circumfix are semantically equivalent
- fixed the circumfix semantics there in the parse
Patrick:
- added more operators to the Rakudo setting
- fixed a few PGE bugs
- talked with Allison about Parrot's installation
- mostly talked about various issues on IRC
- lots of planning, brainstorming, and prioritization for Rakudo and Perl 6
- starting to put together plans for how Rakudo 1.0 will look and when it will come out
- not sure I'm the right person to work on Rakudo install refactor
- will post a couple of messages about that to the list to recruit others
- looks like a lot of important refactoring, but I'm not sure I'm the right person to do that
- getting together stuff for OSCON and YAPC::EU
- finishing up other Parrot tasks for the 1.4 release
- especially HLL interop
- posted instructions for other people to do releases of Rakudo
- anyone can do a release; they don't need to wait a month to do it
- anyone who wants to practice could just do one anytime
- GitHub allows that very nicely
- anyone who wants to be a release manager can practice anytime
c:
- working on Rakudo-exposed GC bugs
Allison:
- trying to catch up on tickets and answer questions
- halfway through the ticket queue
- everything needs a quick skim at this point
- have intentionally not made any plans for the next couple of weeks
- will focus on the release (when I'm not working on OSCON)
- can work on the install stuff
- though I think we should do as little as possible to get it working
Patrick:
- I won't try to refactor Parrot's
make install
- Rakudo wants to be able to build from an installed version of Parrot
- may give up on installing from a build copy
- the downloader installs Parrot to a new directory
- the only blocker there is being able to get a fakecutable working
- need to go through and add conditional Makefile generation to Rakudo's generate script
- am I missing anything?
Allison:
- did it ever work from an installed Parrot?
Patrick:
- can't do a fakecutable there; that never worked
- I can build Rakudo
- it can run if you run the bytecode file
- it won't run from the
perl6 binary
- we need something that can run from there
- people expect that
Allison:
- the fakecutable works if you keep the build directory around
- but not from an installed Parrot
Patrick:
- the other problem is the search path of the dynext/ directory
- in a few places
parrot_config reports compiler settings that won't work on some platforms without conditional makefile generation
Allison:
- it'd be good to get a list of what those are
Patrick:
- I'll make that list soon
- part of the reason I've pushed on this so hard is because we need it
- it has to happen before Rakudo 1.0, but I'll spend my development time elsewhere
Allison:
- we'll get it as soon as possible
Jerry:
Patrick:
c:
- if I use a class as a role in a
does statement, is that okay?
Larry:
- with some caveats, yes
- it takes a snapshot of the class
Patrick:
- my goal between now and the end of YAPC::EU is to develop a detailed plan for Rakudo 1.0
- would like that release to happen by next spring
- if it happens sooner, great
- 1.0 is just a marker tag
- beyond a developer release
Larry:
- Rakudo 1.0 doesn't have to implement all of 6.0.0
- it can be a beta of Perl 6, for example
Patrick:
- need to define what things need to be in that release
- not just spec, but ancillary features
- before that release, we need to use something much closer to STD, if not STD itself
- will make a list of what things need to be in there
- I want to get Rakudo to the point where we can say "It's ready for some real uses"
c:
- is there a performance target?
- I'd like to see at least a doubling in speed
- that's not entirely your problem
Patrick:
- that may happen on its own
- I think that's reasonable
- that's the minimum we can get
Jerry:
- how about working with other Parrot HLLs?
Patrick:
- not sure it's a Rakudo or Perl 6 prereq
- maybe Parrot
Allison:
- I wouldn't put it on the Rakudo list
Patrick:
- we need a better module system
Jerry:
- a bunch of good questions will come up
- having a set of release managers in place will make that easier
Patrick:
- July will be the last release I do by myself, assuming release managers pop up
- and I assume they will
- modeling this after the Parrot Development Summit last year
- charting our progress to 1.0 like Parrot did
- my goal is to finalize things at YAPC::EU
- want to publish a document with milestones and timelines after that
Jerry:
- might be nice looking at which percentage of the spec the tests cover
- and what Rakudo's passing tests cover
Patrick:
- talked about this with Jesse
- very good discussion
- will go through each synopsis and review how Rakudo does with each section of the synopses
- break it down, give each heading a percentage
- would be nice if there were automated tools to help with that
Jerry:
Patrick:
- potentially
- I don't want to block on it
- might shuffle that off to someone if there's an easy task
