[mercury-users] Mercury applications, AI, backtracking

[Randall Helzerman]

> On 14-Sep-1998, Don Smith <dsmith at cs.waikato.ac.nz> wrote:
> > Hi,
> > 
> > I am wondering: is it reasonable to expect Mercury to succeed as a
> > _mainstream_ programming language when the vast majority of
> > conventional applications would seem to involve little or no
> > backtracking? (But see below.)


Instead of asking

        "Do a lot of mainstream applications require backtracking?"

I think the question should be:

        "Can the declarative approach be used in mainstream programming?"

The promise of logic programming is that we specify what we want, not
how to get it.  There are many mainstream applications which could benefit
from this; I'll just mention two:


1.  Parsing/file filters/etc.

    Prolog's ability to just specify a grammar and automatically have a
    parser for it is wonderful here.  The current champ in this area
    is perl, which is great if you are just parsing a regular language,
    but if you are parsing a non-regular language, perl is no better than
    any other language--you still have to hand-write your own recursive-
    decent parser.  I see real potential for a declarative scripting
    language. Here at Intel, we've got about 400 people whose full-time
    job is to maintain perl scripts which manage data communication
    between various CAD tools.  If we had a scripting language which
    made a programmer even 2% more productive, we could redeploy
    8 programmers and save over US$1 million per year.


2.  Databases

    There is no application which is more mainstream than databases, and
    databases should be the killer-ap of logic programming. Every
    company needs a database, and its just not that large of conceptual
    leap for a sql programmer to become a prolog programmer.  And mercury's
    type system already is rich enough to encompass all of the recent
    work in object/relational databases.