Aaron Johnson wrote:

> I can't see any reason to explore a more academic and little used language
> other than being purely interested in languages. At one point in my life I
> had the time for this, but no more---i just like getting things done, and
> that's why I'll stick with Python!

I suggest that you read up on the Blub Paradox [0]:

    http://c2.com/cgi/wiki?BlubParadox

The reason to explore these langauges to to get the programmer
productivity of Python (maybe even better), with a runtime speed
approaching that of C and C++ and programming language features
not available ouside Haskell/Ocaml and a few others.
 
Knowing Python well and Haskell not so well, I  am investing time in
Haskell because I think that within 6 months I will be as productive
in Haskell as I am in any other langauge I know, but that my Haskell
code will run many times faster than Python and will result in more
reliable programs than I could write in other languages.

A couple of points to back this up.

Firstly, these so called 'academic' langauges [1] like Haskell and
Ocaml have a least one feature that Python/C/C++/Java/etc doesn't,
variant types and pattern matching:

    http://www.mega-nerd.com/erikd/Blog/CodeHacking/Ocaml/variant_types.html
    http://www.mega-nerd.com/erikd/Blog/CodeHacking/Ocaml/variant_types_code.html

The variant types and pattern matching make coding some problems
like computer algebra systems and compilers so much easier and less
error prone than they would be in langauges without these features.

Secondly, languages like Ocaml and Haskell have the ability to have
the compiler prove that the code obeys certain properties. For instance
a function to turn a list of Ints into a list of strings might have
a function signature of:

    stringize_ints :: [Int] -> [String]

This function signature specifies that the function has a single
input parameter, a list of Ints and outputs a list of Strings.
If the implemention of that function doesn't meet the specification
then that is a comile time error. If you call that function with
something other than a list of Ints, then thats a compile time
error. While you may think you know this form langauges like C,
C++ and Java, Ocaml and Haskell take the idea of compile time
guarantees much further and is a less obtrusive way than C/C++/Java.

While I agree this may not be important for code that you write 
for your own use, I believe it is important if you are writing
code for other people to use.

At work I write code in Bash, Python, C, C++, Ocaml and Haskell
for embedded kiosk type devices. Of this code, only code written
in C, C++,  Ocaml and Haskell (apart from very small bash scripts)
ever gets run on the the kiosk devices because these languages
provide compile time guarantees of correctness of a larger subset
of functionality than code written in Bash or Python. Similarly we
are moving from C and C++ to Ocaml and Haskell because these later
languages have  greater guarantees of correctness than C and C++.

Depending on these compile time guarantees is not by any means a
subsitute for testing, but it does mean I don't have to test for
errors the compiler can find.

Finally, just because Python is the current hot language that
everyone wants to use, it wasn't always that way. Before Python
became really popular people were using Perl or Java and before
that C++ and before that C and before that Algol or something.

This history (and the  Blub Paradox) suggests that while Python may
be highly popular now it too will one day be superceded by something
else. Having used C, C++, Java, Perl, Python, Ocaml and Haskell among
others, I have a feeling that the next big language will be 
more like Haskell than any other langauge that is available now.
My hope is that it is a langauge with dependant types, a feature
missing from Haskell and Ocaml.

Erik

[0] I know there is some controversy over the Blub Paradox, that is
    why I linked to the C2 wiki instead of the original Paul Graham
    article.

[1] The best known book about Haskell is called "Real World Haskell"
    http://www.realworldhaskell.org/
-- 
----------------------------------------------------------------------
Erik de Castro Lopo
http://www.mega-nerd.com/