OO misconceptions

[Steve Holden]

"François Pinard" <pinard at iro.umontreal.ca> wrote in ...
> [James_Althoff at i2.com]
>
> > Plus the fact that Perl sucks.  (Just kidding.  (Well, not really.
> > (Well, kinda, sorta))). ;-)
>
> Perl was immensely useful to me.  Than Python came, which allows me to
> write bigger projects, allowing them to be much more maintainable.  I'm
very
> grateful to Perl, believe me.  But I would resent being forced into using
> Perl, now that I have something which I enjoy even more that I enjoyed
Perl.
>
I did my share of Perl in a commercial web environment, and some of it was
great. Far less great was the bunch of grot masquerading as CGI scripts
produced by programmers with only six months experience whose first language
was self-taught Perl. Can't blame the language for that, though, I suppose.

I then gave Perl up when I changed jobs, and becuase I was programming for
pleasure for a while I went back to Icon. Having been involved with Snobol
in a previous incarnation I really liked Icon. A very well structured
language design, with most features orthogonal, and generators built in to
expression evaluation. Not the fastest language, like Python it ended up
interpreting bytecodes. Very similar assignment semantics to Python too,
which helped enormously when I finally decided I'd like something more
object-oriented.

Interestingly, Griswold's team *did* produce a compiler for Icon - it
generated C, which you then compiled in the stamdard way. The C source it
produced was long and tortuous, and from experience your programs would be
larger by a factor of three or more when compiled, and the speedup was often
less than 20% for the text-processing programs I wrote for production.

> I'm still dreaming about an automatic Perl to Python converter.  Despite
many
> consider this is not possible, I stick with my naive impression we could
> do something about it.  It could revolve around three parts: analysing
Perl
> into a tree, applying some tree-transformational engine using a
specialised
> language (I wrote a few such in my life), and producing final Python text.
>
Suppose a Perl-to-Python conversion were possible (and I'm not suggesting it
isn't): would it be the kind of Python you'd want to use to teach the
language? I suspect not, any more than I would have used the compiled Icon
to teach C. Given that Perl is ubiquitous, would there really be any point?

I like to poke fun at Perl as well as the next person, but I'd rather do
something new in Python and keep the trusty old Perl scripts around,
performing day in and day out, than translate them into Python just to avoid
using Perl. When I was using it every day it was a concise language, but the
further behind I left it the harder it got to read.

> Would you believe, this is the first step which looks most difficult to
me.
> Parsing Perl might not be that difficult after all, but its tokenizer is
> extremely hairy, and rather discouraging.  If I could swallow that snake,
> the rest would not necessarily be easy, but at least, it would be fun! :-)
>
Ah, well, if you'd do it for fun I have no objection at all. If computing
ever stops being fun I'll stop doing it.

regards
 Steve
--
http://www.holdenweb.com/