Re: Programming skills required by mathematicians?

On Jun 26, 8:03 pm, rAgAv <ragav.pa...@[EMAIL PROTECTED]
> wrote:
> > You did, actually: the existence of other languages that are better
> > at those things.
>
> What might those be?
>
> List them...

As a prolog, let me just quote Stroustrup himself:

"Language comparisons are rarely meaningful and even less often fair.
A good comparison of major programming languages requires more
effort than most people are willing to spend, experience in a wide
range
of application areas, a rigid maintenance of a detached and impartial
point of view, and a sense of fairness."

Anything I can manage to cram into a few paragraphs of usenet post
will be woefully incomplete. That said, while I admire Stroustrup
and also Stepanov, and believe their work should be studied, I
consider C++ to have some serious shortcomings. There are
few projects I would use it on, and none of them are applications.

The big, obvious gripe with C++ is that it is so closely wedded
to its C heritage. This certainly enabled market acceptance,
but it also puts a real damper on innovation. C is a low
level programming language, even by the standards of its day.
By maintaining compatibility, C++ is necessarily also riddled
with low-level features. These low level features make C++ a
good choice if one is writing low level code; device drivers
for example. These same features make C++ a bad choice for
pretty much anything that isn't low level.

Grandparent poster mentioned the combination of performance
and the ability to manage complexity. Rather than try to compare
specifically against some weighted average of attributes, I'll instead
look at those things individually. For performance, C++ is certainly
good, but if performance is your top criterion, then you certainly
cannot pick C++, but must rather use some language designed
for high-performance and having the best tuned high performance
numeric libraries. I am speaking of course of Fortran. Some might
be appalled that I say so, but regardless, there are problem
domains in which C++ cannot compete with Fortran on
performance. Note that I did not say that Fortran is a
good language.

As far as getting performance on the cheap, the easiest way to
go is with the language that is, overall, almost at the level of C++,
and which will sometimes meet and exceed it, and which also has
an extraordinarily broad set of available libraries, and is, today,
more commercially popular that C++, and that is Java. The less
educated are now screaming the "Java is slow" mantra that dates
from 1997, when it was true. A lot of people reject the idea
out of hand, but the fact is the static compilation model that is
the usual one with C++ is unable to take advantage of a class of
optimizations that a HotSpot style VM can, and the reverse is not
true.

For ability to manage complexity, I'd give C++ a poor grade; almost
any language one can name does at least as good; even Java's
impoverished module system is better. People who only know
a few languages in the Algol family and some OOPLs tend to think
that OOP is a great tool for managing complexity. It's better
than nothing, but it's a poor substitute for a decent module
system, especially a parameterizable module system. It is
astoni****ng just how many design patterns (esp. "Dependency
Injection" and its many variations) are simply efforts to make up
for the lack of a decent module system. Also see Peter Norvig's
analysis on the how many design patterns are simpler or even
go away completely in functional languages.

For a good industrial-grade PL with actual modern language
design behind it, I'd point anyone at OCaml.

For the OP's question, namely for someone intending a math
education, it strikes me that there are only two good answers,
SML and Haskell, and I'm astonished that no one has mentioned
either one yet. (Although whoever it was who mentioned CASs
was obviously right on the mark, and these are more directly
relevant than the two I just named. But CASs aren't what we
usually think of as PLs.)

Anyone interested in math ought to have a good grounding in
logic, and anyone interested in logic ought to be at least a bit
familiar with Prolog, or possibly instead some more modern
logic language. (Mercury?)

Anyone interested in set theory ought to be at least a bit
familiar with SQL. It has the reputation of being only for
boring accounting-type jobs, but that reputation is entirely
bogus; SQL has some amazingly advanced features not
found anywhere else, and is enough of a completely different
take on computing that it can be quite broadening. Oh,
and more job postings mention SQL than any other
language, the last several times I've checked; more
than C++, more than Java.

And anyone interested in distributed computing ought to
read about Erlang, at least.

In general, a lot of the languages comparisons or language
wars I see are between some absolutely minuscule set of
languages, none of which have anything in them the least
bit interesting from a PL research point of view. The
conversation is usually limited to discussing C++, Java,
Ruby, Python, and occasionally Perl. These languages
represent just about the tiniest wedge imaginable of the
design space. They are all fundamentally Algol descendants,
all imperative, and have little or nothing in the way of a
formal semantics, virtually nothing in the way of language
features that were invented more recently than 1968.


Marshall