The Perl 6 design team met by phone on 16 September 2009. Larry, Patrick, Jerry, Nicholas, and chromatic attended.
Larry:
- mostly worked on my open source ecology talk
- backlogging after the return from family trips
- pushing a slideshow of Guy Steele about the implementation of easily parallelized data structures
- talking over that with Daniel Ruoso
- helping Matthew Wilson bootstrap his JS emitter for STD
- not much spec hacking, except a conjecture about the parallel semantics
Jerry:
- released Parrot 1.6.0
- went very smoothly
- really like how boring it was
- due to release Rakudo 21 tomorrow
Patrick:
- mostly a lot of thinking this week
- protoregexes in PGE, for example
- reran the spectest progress file since the last release
- Moritz and others greatly improved the calculation of the test suite size
- previously it looked as if the suite were shrinking, due to
plan lines
- the curve is more accurate since August
- Rakudo now passes over 15,000 tests -- almost 15,500
- substantial increase since August
- most of those tests are of operators on Nums, Rats, and Ints
- lots of transcendental functions
- I claim very little credit other than getting the process started
- turned on HLL mapping for converting Parrot Floats into Rakudo Nums
- in some places, Parrot Float internals leaked out
- passed more tests, didn't seem to take a speed hit in the spec tests
- that wasn't true with Int and String before
- haven't seen direct reports of slowdowns from anyone else
- will be able to focus all of my energy in Rakudo and NQP over the weekend
Nicholas:
- Leon pointed me to a link about Python 3's uptake a year later
- only about 1% of registered Python packages support Python 3
- they originally had a 2to3 script to help migrate forward
- now they also have a 3to2 script to encourage people to write in the new version
- interesting to see how one version is more important than the other
Larry:
- we already have a couple of those already
- it's what
kp6 was
- it's what STD with
gimme5 does
- what I'm trying to bootstrap on
viv is
- the problem is that Perl 5 engine isn't actually capable of supporting the Perl 6 semantics on the bare metal, as it were
- you either do emulation on top of Perl 5
- or use cheaters on the innards of Perl 5 to sneak the required semantics in
- both of those have their downsides
- code that uses one approach is not interoperable with code that uses a different approach
- neither of them are strictly interoperable with standard Perl 5 code
- STD does regexp matching, but it doesn't use P5 regexps to do it
Nicholas:
- it writes out Perl 5 code to do it longhand?
Larry:
- yes
- if you've seen Damian's recent talk on getting the P5 regexp engine to store trees...
- he had to go through contortions to make that work
- it's not extensible in the way that P6 regexps are
- it's only part of the way there
- it probably has other bugs
- STD has that emulation layer on top
- getting that to smoothly stitch Perl 6 code back to Perl 5 as if it were Perl 5....
- it may be possible at the subroutine boundary, but that's about it
Nicholas:
- it might have all of the fun of using cfront, then eventually the C compiler chokes
Larry:
- any time you do multiple passes, you set yourself up for various dislocations
- most of them painful
- even so, STD is written in a subset of Perl 6 amenable to backtranslation
- using that subset of Perl 6 as a better Perl 5 will take you only so far and no further
c:
- fixed bugs
- improved performance
- 1.6 should be faster than 1.4, measurably
Patrick:
- I have trouble measuring that
- but my gut agrees
- we keep doing things in Rakudo that should slow things down
- but it doesn't get slower
- I just can't measure it
- the number of tests per second stays relatively flat
- Jerry, did you try a practice run of the release?
Jerry:
- hope to do that tonight
- started reading the release manager guide
Patrick:
Jerry:
- it's in English, which helps
Patrick:
- it's similar to the Parrot steps
- do a practice release in your own GitHub account
- the steps are the same as if you were doing a real release
