[Haskell-cafe] Re: New slogan... (A long speculation)

[ok]

On 9 Oct 2007, at 9:10 am, jerzy.karczmarczuk at info.unicaen.fr wrote:
> * Scheme is very different from what we practice (C++, Fortran,  
> etc., you
> know the song...) It may slow down the *adaptation* of students. They
> *will need* all that imperative stuff you hate. But, as a first  
> language,
> the FLs condition the young minds in a way we do not appreciate.

The only empirical evidence I'm aware of about this comes from
Monash University, Caulfield campus, in Melbourne, Australia.
(A former polytechnic amalgamated with a university.)
If memory serves me correctly, students there used to do 3 semesters
of COBOL.  Rob Hagan changed this to 1 semester of Scheme followed by
1 semester of COBOL.  (It is about 10 years since I last saw him, so
my memory is getting fuzzy.  I'm sure about 1 semester of Scheme.)
The interesting thing is that with *less* time spent on COBOL, his
students mastered *more* COBOL.  So in the one case that I have any
personal knowledge of, learning Scheme first *speeded up* the
adaptation of students.

Three things may be of particular interest here.
1.  Scheme has set! and vector-set!, so it is closer to the imperative
     world than Haskell; not least in I/O.
2.  Rob Hagan put a lot of effort into providing a single-stepper on
     PCs that showed exactly what a Scheme program was doing.  Here
     again, insight gained into execution of strict functional code
     can be expected to carry over to insight into the execution of
     C, or Fortran, or Java.  Haskell would not help here.
3.  Rob Hagan encouraged (no, in some cases *required*) students to
     include test cases in their source files and had a simple test
     harness for using them.  Incredibly easy to use.  There *is* a
     Haskell equivalent of this, namely QuickCheck.
>

> There is no lazy untyped language.

Yes there is.  S.  There is an open source version of S called R.
S(R) is dynamically typed.  Function arguments are *always* passed
unevaluated.  It looks imperative, but if you ignore S4 objects it's
basically functional.  An assignment like
	a[i] <- x
is defined to have the meaning
	a <- "[<-"(a, i, x)
and if you have previously done b <- a, b will not be changed.

S has been enormously successful in its application area (statistics
programming).