Older blog entries for chromatic (starting at number 458)

Perl 5 Design Minutes for 06 January 2010

The Perl 6 design team met by phone on 06 January 2010. Larry, Allison, Patrick, Will, and chromatic attended.

Larry:

• in Spec Land, renamed p{} to qp{} to avoid using up another common single letter
• bare say/print is now just a warning
• Carl Mäsak dug up fossilized restriction on hash literals, which I removed
• since the insides of blocks are now parsed as statements, there is no longer an inconsistency in line-ending curlies
• refined the picking vs grabbing semantics with respect to immutable vs mutable bags and such
• to avoid legacy confusion, renamed break/continue to succeed/proceed
• clarified that an implicit succeed returns the value of the whole when block
• it is not somehow magically inserted around the last statement
• renamed true to so to avoid confusion of the predicate with the True enum value
• at the suggestion of moritz++, split Any up into Any and Cool types
• Cool stands for Convenient OO Loopbacks, or any other acronym you'd like
• the built-in types derived from Cool are the ones that do Perlish dwimmy coercions
• user types still derive from Any by default, so aren't born with gazillions of methods
• conjecturally, also keep "last-resort" multis in Cool package
• responded non-explosively to a potentially explosive rant/twitter
• clarified various things in response
• it is not necessary that all implementations be equally good at everything
• there will be a minimal Perl 5ish grammar alongside STD that any VM can support as a well-behaved subset
• it is also acceptable to support bug-for-bug compatibility with Perl 5
• the language designer is neither omniscient nor omnipotent
• the design process is therefore convergent on the part of all parties involved
• the rate of convergence is an emergent property, and is to be forced
• convergence is deemed to be positive as long as anyone is still working on Perl 6
• the solidification of the spec is also part of the convergence, and depends on proven implementation
• unproven parts of the spec are to be considered implicitly conjectural
• as implementations converge on specs, we can throw out or delay parts of the spec that as yet unproven
• everyone is allowed to panic once.
• on to implementation, found fencepost precedence error inside list prefix
• moved the default initparse method from STD into Curso so other grammars don't have to define it
• added quote modifier:p (aka:path) so we can form the qp{} path literal
• installed better warnings about bare say/print
• generalized the say/print warning to anything a p5er might try that might be in p6 without the defaulting
• STD and CORE now support recent renamings to so, succeed, and proceed
• no longer reports "Bogus statement" when "Missing term" is more accurate
• now catches/\b/ and advises to use an appropriate p6 word boundary assertion instead
• emits better message when an intended reduce is interpreted as composer
• detects most attempts to use postfix after whitespace, and suggests omitting whitespace
• now parses tick-less embedded comment syntax as line-end comment (but still warns for now)

Patrick:

• took the last couple of weeks off
• keeping up with things, but not much writing code
• read the S01 changes with great interest
• glad to see them
• answered some PGE questions for Carl

Larry:

• we've talked about them all along
• they weren't written down in an obvious place

Will:

• working on Parrot
• trying to move as much out of the configure process into a Makefile as possible
• intended to improve the build
• attempting to remove recursive makes
• avoid unnecessary rebuilds
• improve dependency tracking in the build
• probably ready after 2.0

Patrick:

• I'm impressed
• thanks for taking that on; we needed it

Allison:

• working on deprecation notices
• we've talked about a lot of things over the past six months
• not sure they're all in the file appropriately

c:

• working on bugfixes
• also working on deprecations

Patrick:

• I intend to merge the ng branch before the January release
• some people are antsy, but I have a lot of confidence
• we'll probably pass about 70% of the test suite
• it looks like a regression, but we have different features added now
• lazy lists work, for example
• lots of things fudged in the previous version work now

Syndicated 2010-01-23 00:12:46 from pudge

Perl 6 Design Minutes for 16 December 2009

The Perl 6 design team met by phone on 16 December 2009. Allison, Patrick, Jerry, and chromatic attended.

Patrick:

• finished my Hague Grant
• sent the final report to Jesse to wild approval
• drafted a new version of PDD 31 on HLL interop
• write some code to implement part of that in NQP's HLLCompiler
• added various tests
• need to get languages using that now
• Rakudo will use that
• it's the basis for Rakudo's use and import
• if it works for Rakudo, it should follow for other languages which use HLLCmpiler
• had a few comments about missing pieces and corrections
• it's still a draft spec, but we need iteration to finish the spec
• working on little bits of code here and there
• adding features for projects which use NQP
• trying to return to using Rakudo and updating the -ng branch

Allison:

• working on the Pynie refresh this week
• started over with the Python 3 grammar
• have that translated into NQP-rx
• it compiles and can parse a few things
• working my way through the grammar
• no actions set up yet

Patrick:

• there is a.DEBUG rule
• call it on a subrule to turn on tracing from that rule down
• that saves you from having to put in panic statements

Allison:

• is there a good NQP-rx tutorial for actions?

Patrick:

• working on it
• is your work in the Pynie repo?

Allison:

• it's in Mercurial on bitbucket.org under project pynie
• that's what Python 3 uses
• also the Parrot roadmap session went very well

Jerry:

• we still have some actions to take based on that work
• need to convert our priority list into Trac tickets and wiki items
• Parrot's goal for every month until Rakudo * is to support Rakudo * and HLLs in general

Allison:

• that's valuable
• it changes our priorities in the next year
• it moves things between "would be nice" and "necessary"

c:

• worked on the roadmap session
• helping with the Context/CallSignature merge
• will do a dry run of the Rakudo release today

Syndicated 2010-01-22 01:52:43 from pudge

Perl 6 Design Minutes for 09 December 2009

The Perl 6 design team met by phone on 09 December 2009. Larry, Allison, Patrick, Will, and chromatic attended.

Will:

• some work on NQP port of Partcl
• Patrick has been very helpful
• sent a message to the Parrot list about the planning meeting this Sunday
• initiated a community document to discuss those plans

Allison:

• implemented large chunks of obscure C code to perform fast string matching using the FFT
• wondering if that'd be useful in Parrot
• maybe we do our indexing operations by character set in the NFG form
• also does very basic pattern matching by leaving out optional characters
• could be useable in the core tests, where it's tricky to depend on PGE
• this was my final assignment before the Christmas break
• have a month off to work on Parrot stuff then
• I'll show off my assignment when I submit it

Patrick:

• finished my final report for my Hague grant
• haven't quite finished the grant, but left TODO items
• rather than trying to finish everything and then write the report, I'd draft the report and keep notes on what I needed to finish
• need to work on HLL interop
• enable Perl 6 and other Parrot languages to load libraries from other HLLs
• will work on that over the next few days
• had several coversations about optimizations, constants, and inferior runloops
• made minor PAST improvements
• integer constants can automatically promote to num constants without going through a PMC
• updated NQP to make it easier to write custom operator subs, if you're using the operator precedence parser
• implemented the beginnings of smart matching
• not full Perl 6 smart match
• makes sense in the Parrot context
• can match against regexes, tokens, rules, and any types with protoobjects
• code looks more like Perl 6
• not much on Rakudo besides answering questions
• will get back to the Rakudo-ng merge after finishing my grant work
• also worked on Partcl
• updated its regex syntax, particularly for enumerated character classes
• fixed it to handle unquoted, non-word characters in regexes
• previously it only handled barewords as literal matches
• it's closer to the Perl 5 syntax now

Larry:

• didn't like the name PairValSet, renamed it EnumMap
• likewise PairSet is now PairMap, and PairVal is just Enum
• so individual constant pairs now called "enums"
• we distinguish pairs, which have read-write values, from enums, which are constant in the value
• you can now do.enums on hashes and arrays as well as enumerations
• differs from.pairs, which give reference semantics into the values of the original data structure
• .enums gives you a constant snapshot
• David Green suggested renaming Enum.name to Enum.key, and he was right, since they're constant pairs
• trying to be consistent about calling the whole type an "enumeration" and referring to the bits as "enums", even though the keyword is enum
• thought people would rebel at typing the long name
• clarified that the anonymous enum is compile-time evaluated as an anonymous list of constants
• you can always cast to an EnumMap at run time for the other behavior
• simplifying conditional semantics
• STD parser now parses a WHENCE closure as part of the typename, rather than relying on subscript parse
• block escape within a closure within a string used to parse as a normal block by responding to comments outside of the block
• already fixed the embedded block in the regex syntax
• made that usable by strings and regexes now
• blocks in regular code try to figure out if they're at the end of a statement
• look for the trailing curly
• inside a string or regex, there are no statements
• it makes no sense to look for the end of the statement there
• the obsolescence messages were still in the old framework that upsets some Perl 5 people
• changed the wording to "Unsupported use of..."
• #perl6 found a precedence inconsistency in parsing of list prefixes vs list infixes in NG
• turned out to be wrong in STD first, and NG copied it
• I fixed it in STD, Patrick fixed it in NG
• otherwise last week was rather too ADD-ish, so mostly did Q&A on IRC

c:

• fixed some bugs
• made some optimizations
• think I've fixed most constant PMCs in PBC now, which should help NQP and Rakudo

Patrick:

• it'll take a while before Jonathan and I can take advantage of that
• Allison, when you push_eh an ExceptionHandler onto an array in a context, it creates an RPA
• does that hold other things besides an EH?

Allison:

• potentially
• events get stored in the scheduler
• only EHs are scoped to a context
• the old pushmark/popmark stuff to do actions used that same global array
• it may have changed to use the same array
• that's deprecated though
• they won't use that array for long

Patrick:

• I need something to replace pushaction and popaction before they go away
• when we handle LEAVE semantics, we want to avoid generating an exception to leave that scope for caching
• I don't want to generate and rethrow actions to go up the stack
• those ops let me do that without generating exceptions

Allison:

• we do need singleton exception objects for FAIL and RETURN
• no extra information needed
• right now, you can insert anything you want in that array
• the local_branch and local_return uses that array

Patrick:

• it shouldn't
• bsr and ret may have
• I provide my own there

Allison:

• oh right
• I might not have checked in that code

Patrick:

• by the way, NQP doesn't use local_branch or local_return

Syndicated 2010-01-21 04:41:19 from pudge

Announce: Parrot 2.0.0 "Inevitable" Released!

Syndicated 2010-01-21 04:37:49 from pudge

Announce: Rakudo Perl 6 development release #24 ("Seoul")

Syndicated 2009-12-18 08:14:37 from pudge

Perl 6 Design Minutes for 02 December 2009

The Perl 6 design team met by phone on 02 December 2009. Larry, Allison, Patrick, Will, and chromatic attended.

Allison:

• mostly mailing list talk
• finishing up course work before the break

Larry:

• renamed void context to sink context
• appears to have worked
• working out the semantics of higher-level abstract conversions with the lower-level ones
• the method conversions.Num,.Str,.Bool are low-level
• the abstract ones are conversions to roles, like + to Numeric
• the generic role is in charge of which conversion to use
• seems to map better to what people expect
• started to straighten out.true versus.Bool
• renamed is ref to is parcel
• made it synonymous with backslash
• did more work on the Rat semantics to separate out what we originally called Ratio
• now called FatRat
• looked at all of the mentions of context in the spec
• removed overloaded meanings
• dynamic contexts are now call frames
• contextual variables are now dynamic variables
• context means two things now, the old list versus scalar notion and some notion of a thread's context (left those alone)

Allison:

• "call frame" makes a lot of sense
• we may kidnap that

Larry:

• Damian has sent in various spec tweaks, as we gear up to do a Perl 6 book
• kicked the p5=> out of the core into something supplied as a macro by a translator
• ranges now preserve interval semantics when you do math on ranges
• renamed blorst to blast so we can talk about phaser blasts
• worked on the nature of constants
• constant declaration is now in the same category as the subset keyword and othe declarators
• now have an anon declarator
• it confused people to misuse my or our in a situation without a name
• had added an else to go with also, as Damian suggested
• removed both of those in favor of a new metaoperator which guarantees sequential evaluation of the arguments
• the S metaoperator
• after deciding that undef is a lousy concept in Perl 6 -- it turns out to mean five or six different things -- we decided to untangle all of them
• temporarily renamed Object to U, for the most generic undefined
• then revised, per Damian's suggestion, to Mu
• that represents the most generic undefined value and the type from which all other classes derive
• one use of undef was in smart matching
• added a.notdef method for the convenience of testing definedness
• possible to write longer versions which work with smart match
• a convenience method felt nicer
• used heavily in the test suite
• can test something directly with the method without having to use smartmatch
• worked on the Rats of unusual precision (not ROUSes)
• some confusion as to when things get limited to 64-bit denominators
• clarified that only the user-visible results use that limit
• intermediate calculations performed by an operator which can and must exceed that limit temporarily
• added a scaling option to the round function
• easy to round to a particular rational value
• after lots of carping about how ill-defined enums are, I rewrote that spec
• they're no longer roles
• the but and does infix operators still can intuit an attribute property mixin
• that's a function of operator DWIMmery though
• enums, to the first approximation, are simply collections of constants
• they also supply a method,.mapping, which returns a hashlike mapping
• that takes care of most of the special cases
• STD now forces the symbols it matches into strings
• lots of hacking to track spec changes, such as renames
• complex number literals now parse the initial plus or minus as part of the complex
• prevents tragedy in negating the wrong part of the number
• need to nail down pure versus impure ideas to move that into constant folding
• infix operators couldn't accept assignment operator forms for baseops with square bracket or hyperoperator forms; fixed that
• worked on error messages, especially the Borg forms
• removed some redundant rules and the term lexer decreased in size by two-thirds
• added a warning about the use of undef; gives a long list of things you might mean instead
• if you use ** in a range in a regex, but you mess it up where it parses.. as matching anything, you now get a malformed range warning
• added a few missing things to the CORE setting, such as range iterator
• now trims pi and e to fit in to a Rat64
• tweaked viv to support code production of infix operators
• don't have to grovel around to find the operator

Patrick:

• not a lot of direct Rakudo coding
• working in the ng branch
• have eval working; it doesn't poke in the middle of the compiler guts
• also provides a cleaner interface that other HLLs can reuse from PCT
• added regular expression matching
• still needs some work
• added array push and unshift
• fixed a few bugs
• otherwise working on some design issues, such as lazy lists
• answering questions for the branch
• lots of syntax improvements in NQP
• improved error messages
• adding a few features
• added:c($x) (continue) and:p($x) (position) options for regex matching in NQP and ng
• those didn't exist in PGE or the previous Rakudo
• cleaned up the abstract match handling (.ast) to match the new spec
• reviewed Stephen Weeks's work to add try and catch blocks, and it was fine
• updated a few things in NQP to make it easier to write built-in operators
• enables the Perl 6 operator syntax directly
• also working on Partcl
• handles quite a bit already, but not enough
• keep plugging away
• the parsing works closely to what the language spec requires
• I like that
• throughout the implementation, we've focused on doing most of it in NQP itself, not PIR
• seems to work out fine
• adding lots of different commands
• that's fed back into improvements for NQP
• working out very well
• started a regex syntax for Partcl which matches what Tcl requires
• previously used the P5 syntax, which isn't an exact match
• an example for other people of how to do it
• mostly doing Parrot fixes to support the other tasks
• a few small updates to PAST
• improving code generation
• enabling features that other languages need
• bugfix for handling self attribute access in method calls
• adding methods and interfaces for the compiler objects
• will refocus on Rakudo's ng branch
• lots of work fleshing out HLL interop interfaces
• necessary to handle module imports
• also a roadmap item and the last piece on my Hague grant
• posted an interim report recently
• that report marks the halfway point, but I'm much further than that
• should finish the grant this week

Will:

• have a version of Partcl mostly written in NQP
• putting some energy into a discussion about reexamining our roadmap priorities
• been doing that incrementally, but never had a specific meeting just to talk about that
• trying to focus some energy on the list

c:

• fixed a bug or two
• profiled and optimized a bit

Allison:

• how long would it take to migrate Pynie to the new NQP?

Patrick:

• shouldn't take too long at all

Allison:

• have a month off between classes starting in a week
• might work on that

Patrick:

• glad to help wherever I can
• might be an area where you run into a problem and I need to write documentation

Will:

• you can look at the NQP version of Partcl to start

Allison:

• examples have been most helpful in the past

Will:

• Stephen's also converting Squaak and abc to NQP
• might be helpful to look at those too

Syndicated 2009-12-09 23:57:09 from pudge

Perl 6 Design Minutes for 18 November 2009

The Perl 6 design team met by phone on 18 November 2009. Larry, Allison, Patrick, Jerry, Will, and chromatic attended.

Patrick:

• adding new features to NQP
• checked in a copy to the Parrot repository
• added the ability to create empty lists and hashes
• added floating point constants
• added array flattening into positional arguments
• added statement modifiers
• S05 changed with regard to subrule aliasing; updated NQP for that
• so glad I created a new regex engine for this
• that would have been difficult to do in PGE while maintaining backward compatibility
• moved expression parsing closer to Perl 6
• cleaned up literal handling
• added the ability to have contextual variables look in global namespace and dynamic caller stack
• simplified ability to set variables in other namespaces
• continuing to bring Rakudo-ng branch up to feature parity with master
• we'll release the Rakudo master tomorrow
• mostly the same as October, but it uses the new Parrot calling conventions
• spent my time working on the Rakudo-ng branch
• we now have pointy blocks
• updated to S05 interpretation
• can parse numbers with decimal points
• they produce Rats, not Floats, but I'm waiting for the spec to settle there
• needs some cleaning anyway
• statement modifiers now work
• implemented Nil, after extensive discussion with Larry
• works really well as an empty Parcel
• implemented the.map method for lists
• this version is lazy and can handle lazy lists
• Jonathan and I updated dotty method operators to work correctly
• still more to do there
• worked on getting exported methods into a place where they're globally visible
• currently they export themselves into the global namespace
• works out pretty well
• wrote a milestone report; completed the first half of my Hague grant milestones
• should finish the rest of the grant by the end of the year
• Jonathan's doing his amazing work on getting Rakudo pieces back in place
• concentrating on method calls and class handling
• we'll work on that through the next week
• we'll pass a lot of the spectest suite in the branch by then
• it's easy to get things to work
• very few blockers in the new branch
• most of the slowdowns are places where we were cheating but don't want to cheat anymore
• have to do some work to figure out what the right approach is
• started working on a new version of Partcl based on NQP
• it's in the early stages, but it's worth trying to help Will there
• so far it looks very promising

Allison:

• working on tickets and small tasks this week
• have no RT tickets left in my name
• there are no RT tickets left at all
• Will did a lot of fantastic work closing, rejecting, and migrating them
• next week, I may work more on Pynie

Larry:

• a bit under the weather this week
• clarified the Nil semantics of empty blocks and missing branches
• straightened up the Nil semantics for Patrick, as to what it actually is
• the spec allowed a few of the phaser blocks to be statements instead
• we opened that up to be almost all of them
• they're now blorst -- block-or statements
• Damian suggested we find some way of renaming those to blast, so we can have phaser blasts
• unfortunately, they're block-or, not block-and
• but maybe we can force that anyway
• hankered for several months on this next one
• when you alias a rule in the grammar, it does not suppress the original capture
• people who wanted to give multiple aliases to an existing rule had to repeat it
• that made sense in the original rule capture conception
• after we added the dot form to suppress capture, it became easy to suppress a capture and alias a the same time
• foo=bar in an assertion gives you both foo and bar in a capture
• you say foo=.bar to suppress the capture
• removed:panic last week
• turned out to please the crowd
• decided to execute:keepall this week
• it's not very powerful, and it duplicates the semantics of.caps and.chunks
• thinking a lot about Rats, as Patrick mentioned
• design-wise, they consisted of a numerator and denominator of the same type
• frequently you want a larger numerator than a denominator
• settled on that as a normal thing
• the default Rat type, even though it still has a 64-bit denominator, is just an Int on top
• added a FatRat, a Rational with arbitrary precision on both numerator and denominator
• easy to get in trouble that way, performance-wise
• as Moritz suggested, Rat types are instantiations of the Rational role with the two types in question
• Damian pointed out that STASH was the only type in all caps
• changed that to Stash for consistency
• Damian also pointed out that the CONTEXT pseudo-package was confusing
• a singular noun implies one thing, even though it searches upward the dynamic stack
• it's now DYNAMIC
• individual frames are now Context; we distinguished them
• worked on Rat normalization
• they do not have negatives in the denominator
• conversion to String now looks to see if the denominator is factorable into only twos and fives
• if so, converts to straight decimal output
• if not, it converts to a Numeric form in the String case
• in the.perl case, it converts to the slash form
• if it can convert conveniently into a decimal form, it does so directly
• unbroke the check yada (bitrotted)
• thus STD now allows forward-declaration of classes and subs
• rearranged integer capture for Patrick, but he said he didn't need it
• improved consistency of literal rules
• changed to the new capturing/aliasing notation
• fixed up the new blorsts
• added a new Parcel type to CORE
• thinking about the literal formats
• a bad problem turned up with negated complex literals
• they had a different meaning than if they weren't literals, in terms of negating the imaginary part
• prototyped adding the pluses and minuses into the literal parts
• thinking of going the other way
• there is no literal form of complexes; always formed by constant folding
• with recent changes to the semantics of angle brackets, a single literal inside angle brackets would recognize the literal forms we're getting away from
• we could use angle brackets to emphasize the literalness of these forms
• some performance ramifications to putting signs in literal rules
• interactions with different sets of rules to apply to user input is questionable
• that could run things you don't want to run in a tainting fashion
• may be wise to separate literal parsing from the grammar's notion of literals
• beyond that, working out the semantics of numification and stringification
• when we want a basic version of those (give me this type) and otherwise (give me something that does this role)
• the + operators probably do abstract numification, where it picks the type you want
• if you say.Num, you get a Num
• if you say.Rat, you get a Rat
• Moritz pointed out inconsistent specs with regard to stringification and type objects
• we expected them to stringify to the name, with parentheses, but other places expected them to stringify to nothing
• if the internal uses of.str are low-level, inside say or print, we can make that one print the actual name
• say Type.WHAT will give you the name so you can see it
• most of the places we do abstract strings, we want to stringify
• if ~ does the concatenation, we can have it check for undef, return null string, and give a warning
• that maps to the use cases pretty well
• most coercion operations probably want to use the abstract coercion
• maybe
• also probably removing is ref in favor of backslash
• haven't specced that yet
• want to play with the importation syntax
• tag set notation is clunky
• thinking about how to track foldability
• what can you consider a literal, from a grammar point of view
• otherwise, bug fixes

Will:

• talked with Patrick about converting Partcl to NQP-rx
• answering questions about that
• hopefully Patrick will do all of the work
• pushing people to close old tickets from RT
• we can finally use Trac for everything now
• now we have 600 Trac tickets to close

c:

• fixed some bugs
• working on medium term planning

Jerry:

• are you releasing NQP-rx regularly?

Patrick:

• I planned to make a new release a week before the Parrot release
• didn't work out this way this week
• it's evolving too quickly
• but that's still the plan

Jerry:

• will Parrot get a new NQP release in the Parrot releases?
• sounds like yes, unless a language on Parrot needs a critical bugfix between the merge and the release

Patrick:

• right
• NQP will be very stable in the next month or two
• already has most of the features I ever want to add
• still has a couple of regex features I need to add
• expect it to be stable before Parrot 2.0
• could be wrong, as people jump to use it
• but that was the point, and we'll see

Jerry:

• blast could be BLocks, Alternatively STatements

Patrick:

• blast is far less ugly than blorst

Syndicated 2009-12-04 01:20:53 from pudge

Perl 6 Design Minutes for 11 November 2009

The Perl 6 design team met by phone on 11 November 2009. Larry, Allison, Patrick, Jerry, and chromatic attended.

Larry:

• named all of the funny-looking blocks "phasers"
• met with wide-spread approval, or at least humor
• no change as to the language itself besides nomenclature
• nice to have a handle to think of them as one thing
• decided there's no name for the concept used by our lexically-scoped control exceptions
• those are now lexotic
• something that'll unwind the stack to a lexically-determined destination
• noticed the old:panic modifier in S05
• decided it was impossible to implement correctly
• very little demand for it
• couldn't think of a good use for it anyway
• I removed it
• the name interpolator,::(), had some strange lookup rules
• now the runtime lookup in that interpolation is identical to what it would be if the name had been there at compile time
• clarified that empty blocks and non-matching conditionals return Nil, as suspected
• documented that clearly now
• STD now uses something like Patrick's O() assertion form instead of the wrong coercion semantics
• Patrick also noticed that if you declare a module with a block, it did not disallow semicolon declaration
• fixed that
• found three places in the test suite where people had used semicolon syntax as forward declarations
• fixed those too
• decided that the STD grammar class needed to split into two
• one is the mainline Perl 6 class
• the other is the base class from which everything else derives
• you don't want all of the rules of the mainline code in that one
• started to split those
• probably could use more reformatting for readability
• thinking about how to do a real parallel NFA
• reading up on the new Go language

Allison:

• another small tasks week
• reviewing patches and answering design questions
• have a long list of things to do
• put that list on the Parrot wiki, which has helped
• Julian submitted a patch for one of those things
• would like to get some time to work on Pynie
• have three massive assignments due at the end of next week, so may not be able to get back to Pynie after that

Patrick:

• the weeks are all blending together so well....
• Jonathan noticed that on GitHub you can get a list of commits on a project's page
• Rakudo commits yesterday exceeded a page
• we're making excellent progress on Rakudo with the new grammar
• wrote a post on use.perl.org about that last night
• we're on track with our plan we published a couple of weeks ago
• made the sanity tests pass sufficiently that we can run the test suite
• now it's just a matter of re-enabling features to pass the test suite
• most of those features are written in Perl 6
• either because of the core Setting or because we went to rewrite them in Perl 6
• should get major swaths of features working then
• we'll see how that goes over the next couple of weeks
• expect that we'll land the new branch sometime in December
• we'll pass as many tests then as we did in the October release
• the new version fixes many features we couldn't implement in the previous version
• lazy lists and constant handling, for example
• this week, I'll bring NQP into the Parrot repo
• lots of people want to use it more easily
• having Parrot compile and install it helps
• will continue adding features to Rakudo
• will guide others to add new features to Rakudo
• it's at the point where lots of other people can do that
• would still like to write a tutorial on how to write a compiler with NQP
• it's probably worth waiting a week on Pynie to take advantage of that
• figured out why NQP-rx code generation was slow
• UTF-8 bit us again
• took care of that
• compilation time from actions.pm went from 108 seconds to 15 on Jonathan's machine
• about a factor of ten speedup in compilation time

Allison:

• some parts of Pynie need replacing

Patrick:

• it's worth it if you want to use protoregexes or the new operator precedence parser
• they'll give you speed too
• will use Punie (Perl 1) as the example
• it has a lot of the features you want in a language

c:

• fixed a few bugs
• talking design
• talking development process and strategy

Patrick:

• on the semicolon form of module declarators
• the only model I can think of which works cleanly is that you can put anything you want prior to that declarator
• once you write one, you can't write another one for the rest of the file
• that's how I read it
• originally, the semicolon form had to come first
• that caused problems with importing things to declare
• if you allow anything other than module declaration, that's a special case
• it's easier to say that you get one semicolon form per file, and that's it

Larry:

• as the errors show in the test suite, people will use the semicolon form out of habit
• whatever they mean by it will be wrong
• they use it as a forward declaration in the tests
• that's neither of these interpretations

Patrick:

• maybe the semicolon form is wrong
• maybe we need a brace form which lets you slurp the rest of the file

Jerry:

• some kind of EOF heredoc?

Larry:

• if we do use it, I want it close to the top of the file
• as soon as you put large blocks of things in front of it, you get into trouble
• it's not obvious which package it is then
• maybe there's some way of separating the name of the file from the declaration of its dependencies
• let me think about that more

Patrick:

• it's not pressing or urgent
• Rakudo will continue to do what it does

Larry:

• that hurts the tests, which assumed some other behavior
• neither STD nor Rakudo caught these errors until I fixed STD
• I like how STD handles this behavior better now

Syndicated 2009-11-16 21:16:39 from pudge

Perl 6 Design Minutes for 04 November 2009

The Perl 6 design team met by phone on 04 November 2009. Allison, Patrick, Jerry, Will, and chromatic attended.

Larry:

• answered lots of questions on #perl6
• speculated about a new test primitive using $a [op] $b to capture pertinent info on falseness
• many minor spec clarifications this week in response to Jonathan and Patrick
• closures in strings and regexes now documented to create a lexical scope
• $_ now defaults to being is ref rather than is rw
• in regexes, tightened up reversed character class ranges, indirect rule failure semantics
• deprecated the CANDO section in packages, since we don't look in packages for either methods or subs
• in hacking, continue to check and fix tests when others check in parsefails
• STD now gives better messages on bogus statements and backslashes
• implemented the $_ is ref semantics (to the extent STD cares, which is not very)
• tightened up error message for precedence failure within a function's arguments
• catch names with null identifiers
• .caps now always returns key/value pairs in STD instead of naked values
• now gives a better message on an unexpected !!
• .panic now protects itself from infinite recursion on probe for two terms
• various other minor bug fixes

Allison:

• asked for a list of things to work on
• worked through a list of RTs
• wrote a migration guide for the PCC changes
• spending a lot of time on my studies
• quantum computation is a fairly dense subject
• should catch up soon, will spend more balanced time on Parrot and school

Jerry:

• answered some S19 questions this week for Patrick
• didn't answer some others; don't have answers yet
• working on those

Patrick:

• busy week for me
• (refer to my latest journal post)
• we're redoing Rakudo's grammar using the new tools
• the Rakudo grammar is a generation closer to the STD grammar
• we're getting rid of cheats in the existing code as we get closer to STD
• Rakudo now supports lazy lists
• it also has true metaoperators
• previously, Rakudo had separate subs for every metaoperator
• now it detects when you use a metaoperator and registers the code to generate that metaoperator at load time
• we don't have to build metaoperators at load time
• you only pay for the ones you use
• fixed the problem of assigning to a constant; no longer permitted
• subroutines now stored with & sigil
• operator subs stored under their true Perl 6 names, not the PGE-faked versions
• most of the sanity suite runs
• Test.pm compiles, but doesn't run
• it needs END blocks; that'll be trivial to add
• Jonathan has class/method declarations, multisubs, type checking all working in Rakudo
• we need to finish getting the core types in place
• also need to parse subroutine names for operator subs
• that should get us to compile the Setting again
• we expect to pass the spectest suite to the same level as the previous version by the end of November
• could take longer than that, but based on our experience in the past week, I'm optimistic
• helped Geoffrey Broadwell and others switch plumage to the new NQP
• they're eager to use the new features of the new NQP
• working on several other tasks
• fixing some bugs for Jonathan
• going to write a blog post about writing a new compiler with NQP
• probably will use Punie as the example
• it's easy to do; already done once
• wrote a compiler for a stack-based language during an IRC conversation last night in about five minutes
• surprised myself by how easy it is

Will:

• working on migrating away from Parrot's RT
• we switched to Trac a year ago
• it's long past time we switched
• we're down to 95/96 tickets now
• encouraging other people to help, so I don't have to do them all
• eagerly awaiting Patrick's help to convert ParTcl over to the new NQP
• still three or four Parrot segfaults blocking some effort in Tcl
• they all have tickets

c:

• fixed bugs

Jerry:

• in reimplementing Rakudo with the new NQP, are you pulling out the old cheats?
• are you putting in new ones?

Patrick:

• we're starting the internals of the grammar from NQP
• the Setting, written in Perl 6, stays the same

Allison:

• is that because the data structure returned from NQP is different?
• the actions work on parse tree data structures, and have those changed from PGE?

Patrick:

• they're significantly different
• it works on protoregexes
• previously, the operator precedence parser was separate
• it used to be part of PGE
• it's more extensible now
• we based NQP's grammar on the Perl 6 grammar
• Rakudo's grammar started before that
• the action methods have changed a lot
• in answer to Jerry's question, we're refusing to put the old cheats back in
• the new cheats go in their own directory called cheats/
• that documents them
• we know we'll eventually have to take them out
• we acknowledge that previous cheats were blockers to Rakudo's progress
• now we come to the point where we ask if we should do the quick cheat or do things the right way
• so far, we've decided to do the things the right way
• hence assigning to a constant no longer works
• we have real lazy lists too

Syndicated 2009-11-07 00:36:09 from pudge

Perl 6 Design Minutes for 28 October 2009

The Perl 6 design team met by phone on 28 October 2009. Larry, Allison, Patrick, and chromatic attended.

Larry:

• unified nouns and terms in the grammar into a single category
• they're really the same thing
• did more spec work
• worked on global linkage between separate compilation units
• kicked the GLOBAL:: package out of core
• the linker has to pull all of those things together
• different units have different ideas what belongs in global namespaces
• never specified that I wanted decimal literals to show up as rationals
• documented that literals stay exact as long as they can
• there was no specification if you match an undefined variable inside a regex
• now it warns and fails to match, rather than matching ''

Patrick:

• if you were looking up out of an aggregate?
• I'd expect to get a message about an undeclared variable

Larry:

• trying not to propogate magical undefinedness without more help from the user

Patrick:

• does it throw the standard "use of undefined value" warning?

Larry:

• it fails to continue to match
• lots of places Parrot and Rakudo throw exceptions I think ought not to be exceptions
• that's been getting better over time
• people continually try to use angle brackets for lists of numbers
• we ought to make them more DWIMmy in terms of elements, but not the type of the whole list
• Carl pointed me at a rant of Guido's about why Python didn't take rationals from abc
• I decided that the default Rat type should not be infinite precision
• we can make one available
• by default, it maxes out precision at something outrageous, like 64 bits for numerator and denominator
• something that'll be easy to calculate and detect overflow on a modern processor
• won't try to go beyond that, which'll trend to overflow
• decided to break the symmetry with integers
• integers don't naturally expand geometrically in terms of accuracy required
• clarified the semantics of private accessors
• they're very primitive, though we're mulling changes to the syntax
• currently feels a little kludgy
• the AST entry in a Match object only gets set explicitly to an AST tree node
• no longer defaults to the string
• the $() form will default to the string, if there's no AST
• you can still get the AST/string semantics that way
• now you can test.ast directly as a boolean to see if you have one
• holding forth on #perl6 and answering questions

Patrick:

• watched as others took care of the Rakudo release
• I like to list that as an accomplishment
• worked on the new implementation of NQP and the regex engine
• it's going extremely well
• it passes all but one test from the Parrot test suite for NQP
• I haven't bothered to make it parse POD comments right yet
• worked on native support of grammars and regexes last night
• worked on having NQP self-host today, compiling its own grammar and action methods
• there's one runtime issue with initialization
• then I have to fix up the makeful
• then it'll be entirely self hosted and won't need Parrot's NQP or PGE to build or run
• very impressed with the new grammar and protoregex features
• makes it really easy to put this stuff together
• frighteningly easy

Larry:

• I am, after all, lazy

Patrick:

• if only you had come up with this idea four years ago!

c:

• he's too lazy for that

Patrick:

• the code reads naturally
• where it doesn't read naturally, I haven't put in features yet
• has contextual operators, defined-or, direct access to PIR ops as if they were function calls
• you don't have to box and unbox within a PIR string
• that gets handled automatically now
• NQP is really convenient for writing PIR subroutines

c:

• the generated code is ~60% smaller

Patrick:

• it's much smaller
• if you combine PIR instructions, NQP-rx knows the types
• you don't have to box and unbox between them
• an integer result that goes into an op that takes a string doesn't go through a PMC for the conversion
• fell out naturally from the PAST implementation
• had a long IRC discussion about fetch and vivify opcodes
• fetch now exists
• don't expect a huge performance improvement, but I do expect our generated code to look nicer
• can replace eight or ten opcodes with a single opcode
• highly reusable in many new situations
• people will admire the efficiency of the generated code
• started a file with various questions to record for Larry
• hopefully that works out well

Larry:

• I have to be careful about updating the file versus the spec

Patrick:

• I'll point it out to you
• reviewed the Perl 6 book, the work of Jonathan, Moritz, and Carl
• very pleased with the progress they're making
• plan to contribute in the next couple of weeks
• Jonathan updated Rakudo to work with Parrot's new PCC on trunk
• there's a bit of a slowdown as a result
• mostly from parsing
• that's mostly PGE
• not too surprised about that
• won't worry about that, as we'll use a new parser soon
• we'll optimize the new parser
• need to get that a little bit further
• then I'll put out a public call for people to profile the NQP tests
• they're relatively small, compared to Rakudo
• they exercise a lot of the parser
• good place to find bottlenecks and improvements
• our next step is to convert Rakudo to its new grammar
• had a breakthrough last night
• obvious place to start: NQP's grammar
• it's heavily based on STD as well
• I'll copy that over to a branch in the Rakudo tree and we'll build it up from there

Allison:

• mainly working on PCC merge and cleanup
• a few fixes
• added a function for Will that I'd planned to add for 2.0 anyway
• released a new version of Pod::Simple, with lots of help from David Wheeler
• that makes me look forward to Perl 6 a lot
• the old format for parsing POD needs to die
• working on Pynie again
• asked for suggestions as to what to work on yesterday
• cleaning up old code as well

c:

• added fetch
• will wait to see how that works out for NQP before adding vivify
• did a lot of profiling, especially with Vasily
• Parrot trunk is as fast as or faster than it was before the PCC merge
• otherwise, fixing bugs

Allison:

• is it possible with the new NQP/PGE combination to use them as lex and yacc?
• in the language tarball, you can ship only PIR code, not NQP code

Patrick:

• the short answer is "yes"
• that's how NQP bootstraps itself
• keeps a PIR copy in the repo
• builds that copy into PBC files, which should be complete by themselves
• then it uses them to compile the source twice through to get source updates
• you could take those PIR files
• they stand alone
• compile them into PBC files and use that as a library

Allison:

• I want to ship the PIR files
• can ship Debian/Ubuntu packages of Pynie that depend only on Parrot

Patrick:

• I consider the PIR files the shippable artifacts
• I can set up Parrot so that those files are part of the Parrot repo
• periodically, with a discussion of periodicity, Parrot puts a copy of NQP files in the repo

Allison:

• do we need it in the tarball?

Patrick:

• that makes it easier for Pynie and other languages

Allison:

• I'd make NQP a build dependency of Pynie in the packages
• the tarball for Pynie only includes the PIR version of Pynie's grammar

Patrick:

• Parrot isn't completely standalone for this
• you can use NQP to build the PIR files for a parser and translator
• running the parser/translator does not depend on NQP.pbc
• the parser and translator will still depend on Regex.pbc and HLL.pbc, which are effectively the new versions of PGE.pbc and PCT/HLLCompiler.pbc

Allison:

• we need that PIR library with Parrot

Patrick:

• there are four components
• they're separate now, but they'll be there
• the regex engine will be a standalone Regex.pbc
• there's a new version of HLLCompiler
• those are the two biggies
• there's a PAST compiler which compiles Perl 6 regular expressions
• only necessary if you want to compile Perl 6 regular expressions
• then there's NQP itself
• I think everyone will want to have three of those four lying around
• every language (hopefully) will use them
• why not include them?
• NQP doesn't need to be in the Parrot runtime
• it can be in parrot-dev

Allison:

• my wariness comes from trying to maintain Pod::Simple outside of the Perl 5 core
• it's painful trying to maintain that on two different development cycles at the same time

Patrick:

• that's similar to what we have now with Rakudo
• Parrot and Rakudo have different cycles

Allison:

• I was hoping that moving NQP out of the Parrot repository helped alleviate that problem
• NQP could have its own release cycle
• it's an external module that people depend on and load when they need or want it

Patrick:

• I don't expect it to evolve much after January
• it'll be very, very stable
• the point isn't to make NQP powerful or add lots of extensions
• it's supposed to be simple

Allison:

• are Perl 6 grammars still changing?

Patrick:

• not much, in my impression
• I suspect they might evolve toward what NQP provide
• we have several months to try different approaches

Allison:

• what's in the repo right now will be there until 2.0
• I was toying with the idea of using CPython's parser and parse its AST
• still think we're better off using PGE in whatever form it ends up with
• looking for ways to make sure that we can still ship Debian packages right away
• even if it takes a while for NQP as a separate package to make it in

Patrick:

• I'll create a patch
• or check in an example with the four files
• I'll leave the existing NQP and PGE alone in the repo
• they'll remain at least through 2.0 and the forseeable future
• I'll set up Parrot to make it easiest for HLL developers
• if not necessarily packagers
• you can change that if you want

Syndicated 2009-11-02 00:43:41 from pudge