Re: A demodulizer for Haskell?

[Marcin 'Qrczak' Kowalczyk]

Tue, 20 Jun 2000 17:28:25 -0400 (EDT), Patricia Johann [EMAIL PROTECTED] 
pisze:

 For our research, it would be helpful to have a "demodulizer" for Haskell
 --- that is, a program that can translate a Haskell program consisting of
 several modules into a single module.

I think an opposite tool would be useful :-)

-- 
 __("  Marcin Kowalczyk * [EMAIL PROTECTED] http://qrczak.ids.net.pl/
 \__/GCS/M d- s+:-- a23 C+++$ UL++$ P+++ L++$ E-
  ^^W++ N+++ o? K? w(---) O? M- V? PS-- PE++ Y? PGP+ t
QRCZAK5? X- R tv-- b+++ DI D- G+ e h! r--%++ y-

RE: RE: A demodulizer for Haskell?

[Simon Peyton-Jones]

| Are you suggesting that some combination 
| of GHC flags will allow us to
| achieve demodulization?  

No, it won't.  But GHC does actually do a lot
of what you want. There is a lot of tricky stuff
associated with establishing the correct name space,
and GHC does that.  You could suck in a Haskell
module and print it out fully qualified:

module Baz(h,t) where
  import Wug(h)
  t = h 7 + 4

would become

  Baz_t = Prelude.(+) (Wug_h 7) 4

To show that the name-space issue is not trivial,
check out:

module Foo where
  import Baz(h)
  f = h 4

would become

  Foo_f = Wug_h 4   -- Note Wug_h not Baz_h

Other complications in doing this in a standalone
tool include dealing with qualified import, hiding,
'as' clauses, and so on.

It would be fairly easy to modify GHC to 
spit out the fully qualified Haskell for one 
module.  You could then just concatenate all the
output.

Andy's comments about the prelude are on target.
You would almost certainly have to treat that specially.

Simon


| 

RE: A demodulizer for Haskell?

[Erik Meijer]

Have you looked at GHC?

Erik

==
For our research, it would be helpful to have a "demodulizer" for Haskell
--- that is, a program that can translate a Haskell program consisting of
several modules into a single module. Ideally, this tool would also
perform "tree shaking", removing functions that are not accessible from a
specified root module. In particular, we would like to avoid including the
entire prelude in every program.

Stephen Weeks has written this sort of tool --- the defunctorizer --- for
Standard ML of New Jersey. The defunctorizer has been very useful for
adapting SML benchmarks for whole-program compilers.

Is anyone aware of a similar tool for Haskell?

Patricia Johann
Franklyn Turbak

Re: A demodulizer for Haskell?

[Andy Gill]

Patricia Johann wrote:
 
 For our research, it would be helpful to have a "demodulizer" for Haskell
 --- that is, a program that can translate a Haskell program consisting of
 several modules into a single module. Ideally, this tool would also
 perform "tree shaking", removing functions that are not accessible from a
 specified root module. In particular, we would like to avoid including the
 entire prelude in every program.
 
 Stephen Weeks has written this sort of tool --- the defunctorizer --- for
 Standard ML of New Jersey. The defunctorizer has been very useful for
 adapting SML benchmarks for whole-program compilers.
 
 Is anyone aware of a similar tool for Haskell?

I like this idea, perhaps combined with a tool that does the
type checking and type class removal, letting small Haskell like
compilers have a go at larger Haskell programs.

One problem with such a tool is what is Prelude and what is magic/builtin?
The Prelude more defines an interface that should be followed than
an implementation. Take, for example, Arrays. They are deeply wired
into our compilers. The GHC prelude is actually many modules that are
of increasing levels of complexity and Haskell understanding. 
Prelude.hi is just the import portal we all see when compiling!

One interesting question is would GHC, using such a tool,
see similar speedups to GRIN?

Andy Gill