Perl 6 Design Minutes for 24 February 2010

chromatic on 2010-03-02T05:12:09

The Perl 6 design team met by phone on 24 February 2010. Larry, Allison, Patrick, and chromatic attended.

Larry:

• my work last week was almost entirely responsive to various discussions on irc and p6l, even when it doesn't seem like it
• clarified that LEAVE-style phasers do not trip till after an exception is handled (and not resumed)
• the implementation of take is specifically before unwinding even if implemented with a control exception
• simplified series operator by moving generator function to the left side (any function on right side will now be a limiting conditional)
• a * is no longer required to intuit the series on the left; the absence of generator before the ... operator is sufficient
• first argument on the right of ... is now always a limiter argument
• for convenience and consistency, added a new ...^ form to exclude a literal limiter from the generated series
• unlike ranges, however, there is no leading exclusion ^... or ^...^
• series is a list associative list infix, and each ... pays attention only the portion of the list immediately to its left (plus the limit from the right)
• an "impossible" limit can terminate a monotonic intuited series even if the limit can never match exactly
• variables now default to a type of Any, and must explicitly declare Mu or Junction type to hold junctions
• this is to reduce pressure to duplicate many functions like == with Mu arguments; most of our failure values should be derived from Any in any case
• a Mu result is more indicative of a major malfunction now, and is caught at first assignment to an Any variable
• Instant/Duration types are biased away from Num and towards Rat/FatRat semantics
• Instant is now completely opaque; we no longer pretend to be the same as TAI, numerically speaking
• Instants are now considered a more basic type than epochs, which are just particular named instants
• all culturally aware time can be based on calculations involving instants and durations
• list associative operators now treat non-matching op names as non-associative rather than right-associative, forcing parens
• Whatever semantics now autocurry any prefix, postfix, or infix operator that doesn't explicitly declare that it handles whateverness itself
• WhateverCode objects now take a signature to keep clear how many args are not yet curried
• so *+* is now more like WhateverCode:($x,$y)
• autocurrying is still transitive so multiple ops can curry themselves around a *
• added semilists as Slicel type to go with Parcel
• this allows us to bind @array[1,2,3] differently from @array[1,2,3;4,5,6], for instance
• the Matcher type now excludes Bool arguments to prevent accidental binding to outer $_ when closure is needed
• when and ~~ will now warn of always/never matching on direct use of True or False names as matcher
• STD generalizes \w lookahead to all twigils now
• STD now treats non-matching list associatives as non-associative
• things like 1 min 2 max 3 are now illegal, and require parenthesization for clarity
• STD now treat invocant colon as just a comma variant so it does not fall afoul of the list associativity change
• CORE now recognizes the TrigBase enumeration

Patrick:

• first release of the new branch of Rakudo last week
• passing ~25,000 tests at the release
• thanks to optimizations from chromatic, Jonathan, and Vasily, Rakudo has a lot of speed improvements
• in particular, it can run those tests in under 10 minutes, non-parallel, depending on your hardware
• older releases took 25 minutes and more
• the regex tests will slow things down
• ultimately, we're seeing a big speed improvement over the past releases
• cleaned up lists and slices, now they work pretty well
• worked with Solomon Foster and others to speed up trig operations
• fixed a bug related to lexicals declared in classes
• fixed the long-standing and often recurring problem with curlies ending a line/statement causing the next statement to be a statement modifier
• easy to fix in the new grammar
• that was nice
• made an initial implementation of the sort method
• it's very short, because Parrot provides one
• there are a few bugs in Rakudo there still, but I'll get them
• planning for the Copenhagen hackathon on March 5 - 9
• Jonathan and I have been updating the Rakudo roadmap
• will check that in in the next couple of hours
• so far, every time we review it, we surprise ourselves at how much we've accomplished
• we're meeting all of the top priority goals without making any heroic efforts
• we'll put those goals in as well as timelines
• most of the major tasks from previous roadmaps have happened

Allison:

• working on Python this week
• attended Python VM summit, Python language summit, and PyCon
• Parrot's on good track to support what Python needs
• useful to make community connections
• when I reviewed Pynie, I was surprised to see how close it is to supporting the whole Python syntax
• some of those features are big, like objects
• but we should support them soon
• Debian packages delayed by the absence of a sponsor
• they should go into Debian soon though
• I put in a request for feature-freeze exception for Ubuntu 10.4
• Parrot 2.0 should go in
• haven't made any commits to the PCC branch
• that'll be a top priority for next week

c:

• fixed a Parrot GC bug for last week's Rakudo release
• made some optimizations in Rakudo and Parrot
• helped Jonathan find a few more
• fixed a long-standing math MMD bug
• still working on HLL subclassing; more tricky than you think
• may be some conflicting design goals about vtable overriding and MMD

Allison:

• Patrick, do we need an explicit deprecation for old PGE and NQP?

Patrick:

• I think Will already added one for NQP
• we can add one for PGE if we need
• they don't necessarily have to disappear at the next release
• but no one's planning to maintain them

Allison:

• no reason not to put in the notice now
• we don't have to remove them at the earliest possible date