>
> Mark> Similar issues with Mysql. It's faster,
>
> But it doesn't matter *how* fast you get the *wrong* answer. :)
>
I thought one of the first rules of software engineering was "First make it
right and only then make it fast!"
Granted, most of my experience has more to do with number crunching and
scientific computing, but for the past 27 years, I always first ensured the
code was provably correct before worrying about optimization. And this
always involved a test suite that applied every possible mathematical test
with randomly generated test data. For example, my code for matrix algebra
problems generally came with test code and driver that would exercise the
code with tens of millions of random matrices and vectors. One failure,
even in millions of test cases, means the code is somehow broken. I have
seen, and fixed, such cases before. One of my problems with database
development is how to construct analogously strong test cases in order to
prove the code correct. And it seems to get harder as the database gets
larger. :-( In other words, I know how to do this with algebra and
calculus, and C++/Java/FORTRAN programming, but not with sets and SQL
programming.
Cheers,
Ted
