11 Aug
2008
11 Aug
'08
9:01 a.m.
On 11 Aug 2008 at 11:29, Ray Arachelian wrote:
Of course there'd be a lot of manual labor
involved in bringing things
up to standards as computer translated code is usually fuglier than the
original.
In particular, COBOL translated to Java would be very very fugly
indeed--the languages are miles apart in terms of language-specific
features (e.g. Does Java have subroutines that can be both open (i.e.
fall through) and closed (i.e. as the object of a PERFORM--or worse,
as part of a PERFORM...THRU)?). You'd have an easier time
translating RPG to Java. I shudder to think of what trying to work
on the result would look like--or even guaranteeing that it's
computationaly equivalent.
How about automated translation of GPSS/360 to SNOBOL? Or PL/I to
Prolog?
Has anyone had good results with automated translation of any
programming language to any very different one? To me, the results
of automated translation mostly look like Babelfish output and
require one to spend as much time cleaning up the output as they
would have spent rewriting it from scratch. Both require a detailed
understanding of what the code does.
Cheers,
Chuck