|Version 35 (modified by ravi@…, 9 years ago) (diff)|
This page captures the current status of the Haskell' effort. At the moment, this means capturing the current status of the "in" proposals identified by StrawPoll-2.
Since the Haskell Workshop last year, the Haskell community has documented (on this wiki) over 70 proposals for changes to Haskell 98. In March of this year, two subcommittees were established to focus on concurrency and the class system - two difficult areas that are important to the success of Haskell'. The committee has also used StrawPolls to filter and discuss the universe of proposals that has been gathered. Based on the most recent straw poll, 12 proposals (listed in the table at the bottom of the page) have been identified that are expected to get into Haskell' (over 2/3 of the committee in favor). An additional 19 proposals seem likely to get into Haskell' (based on slightly weaker criteria). Members of the committee conflict on 9 of the remaining proposals, and their status will be determined during the writing process. The remaining proposals are not expected to be part of Haskell'.
The concurrency support in Haskell' will be based on the Control.Concurrent library, with minor modifications. There will be a thread-safe interface for mutable state, suitable for use in library code that does not otherwise use concurrency (though what it will be based on is an open issue). There will also be independent FFI annotations for specifying whether foreign calls are concurrent (with other Haskell threads) and reentrant. Bound threads will not be required by the concurrency standard, but they will be an allowed extension with a specified meaning. Open issues include whether standard will require preemptive concurrency, the syntax of the new FFI annotations and the memory-model semantics of IORefs.
The work on the class system has focused on resolving the MultiParamTypeClassesDilemma. The core of the dilemma is that multi-parameter typeclasses are a popular extension that is strongly desired for Haskell'. However, many important uses of MPTCs (like the monad transformer library) require additional extensions to resolve ambiguities in their typechecking. Historically, FunctionalDependencies have been used for this purpose, but they are very tricky to implement and are also difficult to specify in a way that guarantees the termination of typechecking. AssociatedTypes are a possible replacement, but our current implementation experience with them is very limited. The subcommittee has explored several ways to resolve this dilemma (including restricted FDs, fast-tracked ATs and FDs as a "blessed" extension), but, so far, no consensus has emerged. Our current plan is to focus on writing other parts of the Haskell' standard, in the hopes that additional implementation experience with ATs will clarify the situation.
Thus far, libraries have been an underemphasized portion of the Haskell' effort (only 7 of the 70+ proposals have significant library content). However, there is a consensus that a revised standard library is an important part of the Haskell' effort. The current plan is to focus on starting to write the language portions of the standard first, since we have a substantial amount of work to do there which requires focused attention. After that effort is well underway, a companion library effort will begin. Several members of the committee have volunteered for this effort and additional volunteers will be needed.
"definitely-in" Proposal Status
|Description||Ticket||Writers||Draft report text||Decisions made||Currently open issues|
|add some kind of Concurrency||#74||IJ, SM||- basic concurrency interfaces||- cooperative or preemptive|
|- concurrent and reentrant foreign calls||- foreign call annotation syntax|
|- bound threads as a specified extension||- semantics of IORefs|
|add ForeignFunctionInterface||#35||MC, SM||- additional C99 types and casts||- transparent marshalling of newtypes|
|- add signed and unsigned char casts|
|add multi-parameter type classes||#49||MS||- in assuming there is a suitable mechanism for type-indexed types||- MultiParamTypeClassesDilemma|
|add RankNTypes or Rank2Types||#60||AL||- RankN or Rank2?|
|add PolymorphicComponents||#57||AL||- dependency on Rank2Types or RankNTypes|
|add ExistentialQuantification (existential components)||#26||AL, MS, SJT|
|add HierarchicalModules||#24||IJ, BH||None|
|add EmptyDataDeclarations||#25||BH, HN||- dependency on KindAnnotations|
|fix comment syntax grammar||#42||SM||None|
|add PatternGuards||#56||RN, DS||None|
|add InfixTypeConstructors||#78||BH, AL||None|
- IJ Isaac Jones (co-editor)
- JL John Launchbury (co-editor) <john at galois.com>
- MC Manuel M T Chakravarty <chak at cse.unsw.edu.au>
- JG John Goerzen <jgoerzen at complete.org>
- BH Bastiaan Heeren <bastiaan at cs.uu.nl>
- AL Andres Loeh <loeh at iai.uni-bonn.de>
- SM Simon Marlow <simonmar at microsoft.com>
- JM John Meacham <john at repetae.net>This
- RN Ravi Nanavati <ravi at bluespec.com>
- HN Henrik Nilsson <nhn at cs.nott.ac.uk>
- RP Ross Paterson <ross at soi.city.ac.uk>
- SPJ Simon Peyton-Jones <simonpj at microsoft.com>
- DS Don Stewart <dons at cse.unsw.edu.au>
- MS Martin Sulzmann ≠
- AT Audrey Tang <autrijus at gmail.com>
- SJT Simon J. Thompson <S.J.Thompson at kent.ac.uk>
- MW Malcolm Wallace <Malcolm.Wallace at cs.york.ac.uk>
- SW Stephanie Weirich <sweirich at cis.upenn.edu>