environment access (Was: The fate of S, K and I)

[Pascal Serrarens]

Graham,

matthews@math.uga.edu wrote:
 
> Pascal Serrarens wrote:
> > not very flexible. The new Clean 2.0 will have a more flexible module system
> > were you can create functions in a local scope.
> 
> Is there any documentation for what is coming in Clean 2.0? Specifically
> will Clean 2.0 have first class modules?

There is no documentation for Clean 2.0, as far as I know. It is still in
development and features might change. I am not working on Clean 2.0,
therefore I can't aswer specific quetions.
 
> Also there was a recent discussion between Pascal Serrarens and someone
> else on the list comparing Clean I/O to Haskell style monadic I/O (and
> the "do" notation). In this discussion something was mentioned along the
> lines that Clean's I/O is the way it is so that it can support "multiple
> environments". The suggestion was that since Haskell I/O did not support
> multiple environments it was somehow "unsafe". I have quoted all these
> things because I am not sure what is being talked about here. I was
> wondering if anyone (hopefully Pascal?) could post an article/example/
> reference outlining the issues.

First, I want to state that it is not true that supporting multiple
environments is needed for safety. It was the reference-style used in Haskell
wich caused the problem.

I do not know a paper which compares reference style and multiple environemts
with respect to safety directly. But I can give you some references to papers
which discuss monads and environments. First on the Haskell monadic system,
this includes the reference passing style and suggestions for concurrency.

@inproceedings{LaunchburyPeytonJones1994a,
 author      = "Launchbury, J. and {Peyton Jones}, S.L.",
 title       = "Lazy Functional State Threads",
 address     = "Orlando",
 publisher   = "ACM Press",
 booktitle   = "Programming Languages Design and Implementation",
 year        = "1994",
 scope       = "impmonads",
 abstractURL = "http://www.cse.ogi.edu/~jl/biblio.html",
 documentURL = "http://www.cse.ogi.edu/~jl/Papers/stateThreads.ps"
}



Phil wadler proposed a system which did not use references, but a separate
environments system.



@inproceedings{Wadler1992a,
 author       = "Wadler, P.",
 title        = "The essence of functional programming",
 booktitle    = "Annual ACM Symp. on Principles of Prog. Languages",
 year         = "1992"
}



The problem with separate environments and monads is that it is hard to mix
multiple monads. Several papers have been written trying to find a elegant and
flexible way:


@article{Kagawa1997a,
 author       = "Kagawa, K.",
 title        = "Compositional References for Stateful Functional Programming",
 journal      = "1997 ACM SIGPLAN International Conference on Functional
Programming (ICFP'97)",
 address      = "Amsterdam, The Netherlands",
 month        = "June",
 year         = "1997",
 pages        = "217--226"
}

@inproceedings{KingWadler1992a,
 author       = "King, D.J. and Wadler, P.",
 title        = "Combining Monads",
 booktitle    = "Functional Programming, Glasgow 1992",
 editors      = "Launchbury, J. and Sansom, P.M.",
 publisher    = "Spring Verlag",
 series       = "Workshops in Computing",
 year         = "1993",
 note         = "to appear"
}

@inproceedings{HolyerSpiliopoulou1998a,
 author       = "Holyer, I. and Spiliopoulou E.",
 title        = "Concurrent Monadic Interfacing",
 pages        = "253--270"
 booktitle    = "Implementation of Functional Languages, 10th International
Workshop, Draft Proceedings",
 year         = "1998",
 editor       = "Hammond, K. and Davie, T. and Clack, C.",
 address      = "London, UK"
}



For Clean's approach to multiple environments, take a look at Peter Achten's work:



@article{AchtenPlasmijer1995a,
 author      = "Achten, P.M. and Plasmeijer, M.J.",
 title       = "The Ins and Outs of {Clean I/O}",
 journal     = "Journal of Functional Programming",
 year        = "1995",
 volume      = "5",
 number      = "1",
 pages       = "81--110",
 publisher   = "Cambridge University Press",
 scope       = "graphics",
 abstractURL = "ftp://ftp.cs.kun.nl/pub/Clean/papers/ins_outs.abs",
 documentURL = "ftp://ftp.cs.kun.nl/pub/Clean/papers/ins_outs.ps.Z",
 documentSize= "131.8 kbytes",
 keywords    = "GUI I/O system of {Clean}"
}

@phdthesis{Achten1996a,
 author      = "Achten, P.M.",
 title       = "Interactive Functional Programs - Models, Methods and Implementation",
 school      = "University of Nijmegen",
 year        = "1996",
 month       = "February"
}

The example file_func2 in my previous mailing on this subject is a good
example of what could go wrong, I think.



-- 
Pascal Serrarens
Concurrent Clean developer
University of Nijmegen, NL