Re: Why is there a tsquery data type?

Gregory Stark wrote:
> "Bruce Momjian"  writes:
> 
> > Gregory Stark wrote:
> >> "Bruce Momjian"  writes:
> >> 
> >> > There is no question things would be clearer with only one text search
> >> > data type.  The only value I can see to having a tsquery data type is
> >> > that you can store a tsquery value in a column, but why would that be
> >> > much better than just storing it in a TEXT field?
> >> 
> >> When you try storing a tsquery in a column does it alert you if you have an
> >> invalid syntax at that point? Storing it as text would mean not finding out
> >> until you try to use the query.
> >
> > Yes it does check syntax:
> >
> > 	test=> select 'lkjadsf kjfdsa'::tsquery;
> > 	ERROR:  syntax error in tsearch query: "lkjadsf kjfdsa"
> >
> > A larger question is how many people store queries in the database to
> > make it worth the complexity.
> 
> So would this still happen if you didn't have a tsquery type? Or would it
> throw the error only when the actual matching operator executed and tried to
> parse the text?

Well, if you didn't have a tsquery data type then you would get the
error when the TEXT was cast to tsquery for the search itself.

> >> Is converting a text query into the internal format faster or less memory
> >> intensive than converting text into the internal representation? When you run
> >> a query like "WHERE '...' @@ col" if there wasn't a tsquery data type then
> >> '...' would have to be parsed over and over again for each row.
> >
> > No, internally the TEXT string would be converted to something the
> > system could deal with for that query, which is probably what 99% of all
> > queries are going to do anyway by calling to_tsquery().
> 
> How would that happen if there wasn't a tsquery type?

There is an internal C structure which holds the tsquery information. 
My guess is that we would internally have something like tsquery but it
wouldn't be user-visible perhaps.  Right now I am a little confused
about how to do this and keep the data-type-independent nature of the
backend.  You are right that we might have to end up re-parsing the TEXT
field every time it hits the @@ operator, which is a pain.

--  Bruce Momjian            http://momjian.us EnterpriseDB                               http://www.enterprisedb.com
 + If your life is a hard drive, Christ can be your backup. +