Older blog entries for chromatic (starting at number 476)

Perl 6 Design Minutes for 12 May 2010

Syndicated 2010-06-16 01:47:02 from pudge

Perl 6 Design Minutes for 05 May 2010

The Perl 6 design team met by phone on 05 May 2010. Larry, Allison, Patrick, Will, and chromatic attended.

Larry:

• various spec updates, some major
• removed p5=> description because it's not supported in core
• deleted self:sort construct because self isn't a real syntactic category
• explained Perl patterns in terms of PEGs, and spec'ed tiebreaking rules explicitly
• last but not least, finally purveyed the long-threatened revamp of proto to keep routine and method semantics similar
• they all now work much more like the multiple dispatch semantics currently used by STD, where we always call the proto first
• the proto is then always in charge of the actual multiple dispatch; it can of course delegate that
• and the default for a null body corresponds closely to current semantics
• in hacking news, the lexer generator mislaid any alternative that was a bare. pattern, so cursor_fate never called its alternative, oops
• took me a long time to run that one down, because it resulted in a horrendous backtrack causing mysterious misplaced errors
• revamped character class parsing to be more helpful and correct
• STD now check a normal regex bracket's innards for old-school character class, and warns if found
• added a.looks_like_cclass method to Cursor to detect most accidental uses of P5 ranges
• some valid P6 brackets will complain, but the workarounds are easy
• just put whitespace on both ends is one way
• removed a few of these old-school-ish character classes from STD
• changed:tr language to:cc language since character classes share it
• (translation pays more attention to ordering, but the language is the same)
• turned out parsing character classes discovered issues in STD; various character classes needed to backslash # that would otherwise be a comment
• to that end, we now allow \# in character classes instead of misparsing as unspace
• if we find an invalid - in a regex, we now presume we're in an old-school character class and fail with a sorry instead of a panic to give the character class code a shot at it
• STD now uses ~ syntax for regex brackets to set $*GOAL correctly
• cleaned up recursive panic detection; it was possible to get both false positives and negatives before
• STD shouldn't use 'note' to emit a panic inside a suppose because that leaks the message that should be trapped
• STD now suppresses duplicate sorry messages more correctly
• sorry no longer uses panic in a supposition, but dies directly to throw the exception to the suppose's try block
• STD now allows subscripts on regex variables so $x[0] isn't taken as a character class; still needs speccing

Patrick:

• can we make them consistent?

Larry:

• historically S05 has allowed bare arrays to mean interpolation

Patrick:

• we've never had a working implementation of that

Larry:

• a bare @ would be illegal

Patrick:

• it's currently illegal

Larry:

• you'd have to backslash it to match part of an email address
• it's not like the @ alternations are a big deal one way or another
• that'd be a little more consistent
• I forced it to think of the sigil as $ than what it really is

Patrick:

• after seeing how Jonathan et all did interpolation for quoted strings, I thought we should do the same thing in regexes

Larry:

• STD now has a partial fix to prevent leakage of::T from role signatures
• unfortunately, the current fix will lose signatures of file-scoped generic roles
• this probably has to do with not knowing whether we're really going to want a new pad; unfortunately we'd have to look ahead to know that currently
• various other minor tweaks and bug fixes in STD and Cursor

Patrick:

• mostly responding to messages and reports
• should be able to get back to coding full-time and online for the next week
• plan to resolve the list and closure issues with NQP and Rakudo
• will answer other questions and try to keep other people productive
• planning for the Rakudo Star release on June

Allison:

• busy with the last week of classes
• spent most of it writing a little language with PCT
• it was easy to use and easy to swap the stages of PCT
• I remembered what Patrick did with LOLCODE
• also had a discussion of source code control systems
• next week should be more productive
• need to work more closely with Debian packagers to get packages into Debian

Will:

• cleaning out as many deprecations in Parrot as possible
• trying to improve the speed of CodeString after the immutable STRINGs merge
• bundling lots of little concats helps
• hope to merge in an optimization branch for that by the weekend
• want to make that faster or less memory intensive
• may require the use of a new StringBuilder for Parrot
• hopefully will result in a faster Rakudo build

Patrick:

• I've never seen CodeString take a long time
• unless you run into memory problems

• * discussion of the StringBuilder PMC *

c:

• still working on optimizations, particularly CodeString
• looking at more PBC and PBC-building optimizations
• PBC size went down dramatically and startup improved for Rakudo
• should have that much faster for the 2.4 release
• will poke at GC tasks starting next week

Syndicated 2010-05-10 06:07:34 from pudge

Perl 6 Design Minutes for 28 April 2010

Syndicated 2010-05-02 00:27:42 from pudge

Perl 6 Design Minutes for 21 April 2010

Syndicated 2010-04-28 22:20:35 from pudge

Perl 6 Design Minutes for 14 April 2010

The Perl 6 design team met by phone on 14 April 2010. Larry, Allison, Patrick, Will, and chromatic attended.

Larry:

• on p6l, did a bit of bikeshed paint removal with regard to hyphens vs underscores
• S02 now explicitly disallows both whitespace and unspace in top level of an interpolation
• per spec change, when STD is parsing an interpolation inside quotes and looking for a possible postfix, we now presume that a backslash belongs to the quotes and is not an unspace
• in the src/perl6 directory, renamed all Perl 6.pm files to.pm6 to avoid confusion
• this was necessary because the implementation of STD translates Perl 6 back to the corresponding Perl 5
• the ambiguity was causing problems with tools such as NYTProf
• Cursor.pmc now prefers.pm6 over.pm in any particular directory when searching for Perl 6 code
• as usual lately, most of my hacking work was in improving the human interface of the parser
• STD now distinguishes two final messages: "Parse failed" vs "Check failed"
• STD now warns on attempts to smartmatch with True or False
• STD now distinguishes continuable-but-fatal "sorry" messages from immediately fatal "panic" messages
• sorry messages will eventually fail at check time
• changed many of STD's semantic errors to use sorry messages when the parse state is not affected
• modified moritz++'s conflict marker patch to be more like the Clang compiler's behavior
• conflict markers now emit a "sorry" message and continues parsing one side of the conflict
• also fixed a buglet that prevented it from processing the conflict marker if first thing in the file
• while fixing the vws (vertical white space) rule for that, also changed it so that extra lines are now eaten with \V*\v for consistency
• it had be using \N*\v
• gimme5 now supports pointing to both ends of missing goal message
• STD's "Couldn't find final..." messages now use that capability to point to both ends of the error
• standard quotes now also use the ~ compositor to set the goal and get that behavior
• STD will now dwim <<op>> ("Texas hypers") better even if op contains angles
• suppressed confusing backtracking on ~<< that produced a misleading quotewords error
• some other patches
• CORE.setting now recognizes the 'note' function
• gimme5 now translates note to print STDERR
• cleaned up some unneeded locmesses
• Actions.pm now handles prefix metaops without spewing spurious yaml dumps

Allison:

• worked on TT #389
• the actual fix was about two lines
• spent a lot of time fixing tests around it
• didn't like the original two-line fix
• fixed it in IMCC by passing along the:nsentry flag
• NQP-rx still depends on that feature
• I understood from Patrick that NQP-rx doesn't need that feature
• don't want to launch that before the 2.3 release
• worked on a lot of smaller issues
• worked on the Parrot Developer Virtual Summit
• will talk about some process changes more, as there are details to work out
• will work on GC as the next priority
• useful for Rakudo in general and Parrot concurrency

Patrick:

• catching up on mail and tickets
• should get back to coding in the next couple of days

c:

• worked on the immutable strings branch
• need a couple of changes in the Rakudo binder
• now it's time to convince everyone else it's a worthwhile design change
• going to work on bugfixes
• will try to land the constant string cache
• otherwise, added some optimizations

Will:

• worked on Partcl
• fixed a Parrot bug that broke Rakudo
• does Rakudo need TT #389 in 2.3?

Syndicated 2010-04-21 04:09:26 from pudge

Perl 6 Design Minutes for 07 April 2010

Syndicated 2010-04-09 22:07:25 from pudge

Perl 6 Design Minutes for 31 March 2010

The Perl 6 design team met by phone on 31 March 2010. Larry, Allison, Patrick, Will, and chromatic attended.

Larry:

• in SpecLand, made it clear that the brackets in pairs are not related to subscripts, but follow the corresponding fatarrow semantics
• in particular, name extenders are just strings or list of strings, properly indicated by : or:() in most cases (this includes all operator names).
• we now forbid name extension using the:{} keyless adverbial syntax
• we don't need that because name extensions are really only supposed to use values, not closures
• if we really, really, need to supply a closure as part of a name extension, we can put it in parens, as in:({}).
• we can use that notation for supplying a closure as a first argument to a method without requiring a space between the colon and the curly, as in.map:{...}
• people keep writing that and expecting it to work, so I thought it would be good to make it work
• the colon is still require before the curlies, or it's a hash subscript
• we now capitalize the Junction type again because I couldn't get people to stop capitalizing it
• also, the native aspect of junctions is not their most salient difference from normal types
• conjectured an Each type that autothreads lists like junctions, but is serial and lazy, and is used for its values in list context, not boolean context
• in S05, did much cleanup of cursor semantics to reflect what STD and Rakudo actually do these days
• retargeted the <&foo> regex assertion form to explicitly call a routine, just like always calls a method
• a bare <foo> assertion now prefers to call a lexical function if visible, or calls as a method in current grammar if not
• this is a compile-time distinction, not a fallback at run time
• in code hacking, continued debugging of the backtracking transactions I added last week
• gimme5 now sets the correct xact on || alternations
• deleted more of the transactions that are no longer needed when building match results that are no longer hypothetical
• a LazyMap now always passes through the first result regardless of its associated commit transaction state
• that's because the first cursor in a lazy list always represents the current match hypothesis, not a future hypothesis that needs pruning
• STD now parses to the new specs regarding name extensions not including:{}
• now allow colon form of method arguments to omit the space if the next char is a left curly, which is what people seem to expect anyway
• note that this makes the closure the first argument, not the only argument
• STD now gives more useful error messages when user says things like 'if' as a function call (if(...) {...}), or a statement control like 'given' where one isn't expected ($x = given {...})
• STD now properly objects to unrecognized internal regex modifiers such as:has
• improved the message on adverbs with empty angles (:foo<>) to list some better options
• the problem arises when people think that the angles produce a null string, when in fact they produce a Nil list
• other malformed pairs are also better diagnosed, such as:! not followed by an identifier, or pairs with duplicate arguments
• added a new rule that traps all warnings and errors
• STD now uses suppose in place of custom try blocks in diagnosing such things as two terms in a row, or unexpected infixes
• also uses suppose to soften the warning about backtick-less embedded comments by not complaining if the supposed comment eats the whole line anyway
• put in some code to de-dup identical warnings
• STD now includes the signature's return type (after -->) in the check for redundant 'of' types
• did various tiny speed tweaks, fossil removals
• started playing with how to mark sink context and pure operations
• split out Actions.pm from viv so that it can be used by other STD-based AST builders
• this is in preparation for propagating attributes up and down the AST such as sink context and purity
• eventually this will result in "Useless use of" messages where appropriate, not to mention the ability to do constant folding

Allison:

• worked on the compact_pool() function
• split it into a series of smaller functions
• it could use more work, but it's an improvement
• found one possible bug
• worked on some documentation, especially for PMC attributes

Patrick:

• reviewed some patches to add variable handling in regexes
• they need some changes, but the overall concept is good
• reviewed a few other messages
• most of my time is going toward my family
• hope to get more time to be more active in the next couple of days, but I can't promise that yet

Will:

• talked to Jonathan about Rakudo Star priorities
• he's very pleased with the memory fixes
• the next thing on his list is getting good line numbers in reported errors
• closing tickets
• working to get rid of the last recursive Makefile
• may wait until after the new release, when we remove a lot of deprecated things
• practicing my NQP skills by working on Tcl again

c:

• worked on the Rakudo memory problems
• Vasily and I fixed the big memory use problem
• still some performance tuning to do there
• wrote up tasklists for two other important performance pieces
• will work on line numbers after we get performance back

Syndicated 2010-04-05 04:42:21 from pudge

Perl 6 Design Minutes for 24 March 2010

The Perl 6 design team met by phone on 17 March 2010. Larry, Allison, Patrick, Jerry, and chromatic attended.

Larry:

• clarified that nearly all normal operators autothread, including === and eqv
• specced the \| parcel parameter syntax
• documented that R metaoperator does not change associativity
• clarified that trusts traits do not extend to child classes, and moritz++ specced it
• in STD, we now suppress spurious errors from badinfix lookahead (and react more accurately to bogus terms)
• now put the error location pointer before a bad infix, not after
• we no longer assume missing block punctuation is always semi or comma, but keep them as a suggestion
• missing punctuation message now points before any whitespace
• awesomified error message about no unspace in regexes to explain how to quote space or #
• pass single coeff to radcalc to make:16<.BABEFACE> easier to allow
• gives better message on missing ** part of radix literals
• worked around fact that:: doesn't correctly suppress relexing of multi tokens
• scrapped the workaround and did a complete refactor of commit point transactions; no longer uses exceptions to commit
• instead, it walks the current commit chain to the proper commit target to disable choosers that should not choose any more options
• commit chain aliasing and forking to make a cactus stack is now managed by cursors, mostly transparently
• weighed in on the subject of stability domains (or lack thereof) in Rakudo *

Patrick:

• still working on personal issues, but hope to have some resolution by Saturday
• haven't had much time to work on Rakudo, but show up on #perl6 to give advice sometimes
• read Larry's email to the list; it was very helpful

Allison:

• met some interesting people at SxSW doing open source education technology
• reviewing the roadmap
• the GC sounds like the most important thing to work on next
• trying to catch up from having spotty network access lately

Jerry:

• the other Rakudo developers have started weekly planning meetings
• Jonathan has taken the lead
• plenty of contributors are in the meeting and offered to take on new tasks
• Rakudo Star may have a smaller scope, but it'll still come out in Q2
• it's nice to see that the Rakudo community continues even as Patrick has an extended absence
• still some Parrot issues affecting Rakudo
• PaFo hopes to have its 501(c)3 application done by summer

c:

• bugfixes
• minor optimizations
• helped merge the PCC refactor branch
• working on Rakudo memory issues (long analysis follows, partly Parrot GC and partly NQP behavior)

Syndicated 2010-03-31 21:20:11 from pudge

Perl 6 Design Minutes for 17 March 2010

The Perl 6 design team met by phone on 17 March 2010. Larry, Allison, and chromatic attended.

Larry:

• documented which ops don't autoclose with *, including assignment
• conjectured a generalization of the closure-calling context (value-only lists) that subscripts enforce
• this generalization might allow the autoclosing of some of the current exceptions such as 1..*
• added Z to go with X metaop; documented that X and Z desugar to higher-order methods, crosswith and zipwith
• speculate about how to zip/cross dwimmily with non-identical ops; possibly creating a real use case for surreal precedence
• however, for now sticking with conservative approach of requiring parens on differing list infixes
• hacking on viv again
• trying to get that bootstrapped, so I don't have to use gimme5
• unbitrotted viv --p6 so it exactly reproduces STD.pm again
• various developments with viv --p5 toward replacing gimme5
• should make it easier to emit other parsers eventually
• may emit Rakudo code someday
• it's a race to see whether STD can do that before the current Rakudo parser resyncs with STD
• anyone who wants to bootstrap on some other VM might want to use that
• mostly tired of writing in the subset of Perl 6 that gimme5 understands
• mostly hacking on better error messages, as always
• catches use of non-$ hard reference
• STD now read minds of people who forget that ".meth I" is a two-terms-in-a-row error
• now produces good messages on attempts to use y/// or tr/a-z/A-Z/ syntax
• now reports "previous line missing its semicolon" in the unexpected block checker
• ambiguous use of. probably indicates p5-think, not missing method parens
• STD now has in a q-like sublanguage for tr/// string parsing
• implements the MONKEY_TYPING constraint on augment and supersede declarators
• various random cleanups and bugfixes
• added Z metaoperator
• lots of works on regex flags to unify them into a single %*RX structure at parse time
• makes it easier to do all of the lexical scoping in parallel
• can now remap run-time's $?FOO variables to parser's $*FOO dynamic variables
• otherwise, bugfixes, spec cleanup, and test cleanup

Allison:

• working on tickets
• updating the Parrot roadmap to match our Rakudo Star support plan
• working on the mini-language in NQP for a class assignment
• found a new Pynie developer who saw my talk at Pycon
• may be doing a Summer of Code project in it
• answering lots of questions on IRC and helping out with ideas

c:

• working on lots of little bugs for Parrot
• should have the method namespace bug fixed, with help from Andrew
• exploring some optimization possibilities
• should be able to merge the PCC refactor shortly
• Allison, see TT #1511

Allison:

• we need to add a new opcode, something like set_want
• call it to update the CallContext with expected return information

c:

• works a bit like Perl 5 there
• we could use that information for MMD, that'd be interesting

Syndicated 2010-03-31 21:17:44 from pudge

Perl 6 Design Minutes for 10 March 2010

The Perl 6 design team met by phone on 10 March 2010. Larry, Allison, Patrick, Jerry, Will, and chromatic attended.

Larry:

• scrapped @array[%100_000] modular subscript notation in favor of a more general mapping closure
• put back:s file test, removed:z,:T,:B,:M,:A,:C
• clarified that these are defined on IO, not on strings
• deprecated the {*} and #= reduction stub notations in grammars
• attributive parameters now default to is copy binding; but easy for an attribute to override this with is ref
• tried to move operator definitions to CORE; found one approach that doesn't work and abandoned it
• STD now allows _ in numeric variable names like $10_000.
• factored out curlycheck so we can use it on any trailing curly
• postcircumfix:<{ }> now uses curlycheck for consistency
• STD now speculates missing semicolon when two terms in a row are separated by at least one newline
• removed mention of *.notdef in favor of:!defined
• still need to remove it from the spec though
• ambiguously rebound outer lexicals now detected even if ambiguity propagates from an inner scope
• reports more pertinent information in that case so the difficulty can be understood by the user
• various random debugger refactorings
• properly scope dynamic package names for block-oriented packages to include name declaration
• package_def of; packages now eats statementlist itself to stay inside proper scope
• much work on package qualified names
• correctly parse <$x> part of FOO::<$x> as part of variable name
• correctly follow symbolically indirected OUTER:: links
• find_top_pkg no longer cares if name ends in::
• STD now figures out whether initial components lead to package or lexical scope
• no longer scans outer scopes on qualified names
• now handles FOO::<$x> form in check_variable
• no longer checks for @/% mistakes on qualified names

Patrick:

• haven't had much hacking time lately due to personal demands
• should be able to hack again later today and the rest of the week

Allison:

• worked on the PCC refactor
• that went well; the hackathon was good
• it didn't pull in a lot of people, but me dedicating the weekend to it was helpful
• also pulled in a few other people willing to try things out
• we made good progress
• our initial task is over
• we're in the nebulous stage of debugging
• need to review a change in optional return values
• also worked on Ubuntu and Debian packaging
• Parrot 2.0 is in both
• it'll be in April's Lucid Lynx Ubuntu

Jerry:

• Google Summer of Code is starting
• TPF and PaFo are teaming up this year
• we're working on the organization application
• Jonathan Leto is leading things and I'm backing him up
• we're looking for mentors and ideas; see the TPF GSoC wiki page

Will:

• started going through Rakudo's RT queue
• did more Parrot building and cleanup work
• no longer invoking Perl to invoke the C compiler for each build file
• shaved some time off the build
• eliminated one recursive make, leaving two
• then I can remove more things from config

c:

• worked on a bunch of branches
• fixed a couple of bugs
• hope to get more bug fixing time in

Syndicated 2010-03-23 01:37:49 from pudge