[Sigia-l] The Horror of Access Database

[Andrew McNaughton]

On Thu, 4 Jul 2002, Listera wrote:

> > The point was that an RDBMS doesn't use inverted indexes, unlike a
> > full-text system. And SQL isn't designed for doing these types of
> > queries.
>
> Yes and no. A database, in my book anyway, ought to be a datastore and not
> much else. The *application* logic belongs in the application server and the
> language with which you get at the data shouldn't make that much of a
> difference. In that sense, there's nothing that prevents SQL or any database
> (think storage) from handling full-text indexing.

Many database engines do have built in inverted text indexes these days.
Most are rather inferior to dedicated text indexing software, but they are
suitable for many purposes.

Ignoring questions of what a database should be, or what SQL is, there is
a real dilemma as to how the boolean logic model should mesh with
relevancy ranking.

Fortunately, most applications the number of items being searched is
small enough that boolean logic and a well planned metadata schema are
sufficient.

> > This makes it difficult to build a search interface to a relational
> > database so it functions like a full-text web search engine.
>
> With sets, it's a piece of cake, as I have previously explained here. You do
> have to go through a layer of abstraction and conversion.

The notion of sets you have proposed earlier is what is typically referred
to as an inverted index.  The performance claims you've made earlier are
questionable.

Andrew