Perl 6 Design Minutes for 17 February 2010
The Perl 6 design team met by phone on 17 February 2010. Larry, Allison, Patrick, and chromatic attended.
Larry:
- much work clarifying relationship of parcels to everything else (
<a b>, assignment, arguments, captures, parameters, signatures,gather/take, and loop returns) - we now list all scope declarators in one spot
- conjectured some ideas on how to handle the allomorphism of literals more dwimmily
- had already specced some of this behavior for literals found inside
qwangles. - literals that exceed a
Rat64's denominator automatically keep the string form around for coercion to other types - clarified that anon declarator allows a name but simply doesn't install it in the symbol table
- respecced the trig functions to use a pragma to imported fast curried functions
- still uses enum second argument for the general case (rakudo is still stuck on slow strings there)
- on iterators, renamed
.getobjto.getargsince arguments are the typical positional/slicey usage - signatures are never bound against parcels anymore, only against captures
- we now use "argument" as a technical term meaning either a real parcel or an object that can be used independent of context as an argument
- anything that would stay discrete when bound to a positional, basically
-
return,take, and loop return objects are also arguments in that sense - they all return either a parcel or anything that can stand on its own as an argument
- STD now adds a shortname alias on adverbialized names, ignores collisions on the shortname for now, which is okay for multis
- STD now complains about longname (adverbialized) collisions
- STD no longer carps about duplicate anonymous routine declarations
- made the undeclared type message the same for parameters as for other declarations
- clarify the error message about anonymous variables
- no longer report a
$)variable error where)is the$*GOAL - add
WHATetc. to list of functions that require an argument
Allison:
- working on two HLL implementations
- one is Pynie, the other is Camle
- nothing to do with Caml or ML
- I've noticed huge improvements in NQP-rx from the previous NQP
- can't say which feature improvements make the most difference, but I'll migrate Pynie pretty soon to take advantage of the new version
- continuing to shepherd Debian and Ubuntu packages
Patrick:
- essentially all I did was unify things
- previously it had been two or three tools
- it's just one
Allison:
- even the syntax seems more regular
Patrick:
- there are more pieces available in NQP-rx
- Rakudo's -ng is now master
- the old master is now -alpha
- we took a big hit on spectests, but they seem to be coming back quickly
- 5000 tests pass on trunk now
- we have 16k or 17k we haven't re-enabled; they make the spectest slower
- Jonathan thinks we may pass 25,000 tests now
- that's great, considering where we were a week ago
- I redid Rakudo's container, value, and assignment module
- previously variables held values directly
- now they contain reference PMCs
- that cleaned up many things
- we use more PMCs, but now we don't clone and copy as much
- we move references around more
- seems closer to how Perl 6 handles things
- was much easier than I expected
- updated the NQP-rx regex engine and built in constant types
- handles Unicode character names
- reclaims plenty of tests
- answered lots of questions for people adding things into Rakudo
- prioritizing other people writing code over writing code
- increases our developer pool; seems to be working well
- new release of Rakudo planned for tomorrow
- don't know how many tests we'll pass, but it should go well
- plan to put in a few things like
sortand grammars over the next week - then I'll review the RT queue to find bugs and (hopefully) closeable bugs
c:
- working on GC tuning
- also working on String PMC tuning
- working on built-in types and their behavior as classes and parent classes
- the multidispatch bugs in particular I hope to solve