Thread

1. Re: Updateable views...

   Eric D Nielsen <nielsene@mit.edu> — 2003-03-05T15:39:02Z

   > Gavin Sherry <swm@linuxworld.com.au> writes:
   > > I haven't had time to look into it further, but it occurs to me that
   > > handling views which rely on joins would be far from trivial.
   > 
   > Views containing joins would not be updatable; problem solved. 
   
   I see how that is what the spec says, but aren't the majority of joins that
   people use/want to update a join of some type?  I thought that SQL99 allowed
   updating view created by joins.
   
   In either case is this a place where "exceeding" the spec would be a good 
   thing or a bad thing?
   
   Eric
   
   

2. Re: Updateable views...

   Tom Lane <tgl@sss.pgh.pa.us> — 2003-03-05T15:47:44Z

   Eric D Nielsen <nielsene@MIT.EDU> writes:
   > In either case is this a place where "exceeding" the spec would be a good 
   > thing or a bad thing?
   
   Unless there is an obvious definition of what updating a join means
   (obvious not only to the implementor, but to the user) I think this
   is dangerous territory.
   
   			regards, tom lane
   
   

3. Re: Updateable views...

   Rod Taylor <rbt@rbt.ca> — 2003-03-05T16:12:01Z

   On Wed, 2003-03-05 at 10:39, Eric D Nielsen wrote:
   > > Gavin Sherry <swm@linuxworld.com.au> writes:
   > > > I haven't had time to look into it further, but it occurs to me that
   > > > handling views which rely on joins would be far from trivial.
   > > 
   > > Views containing joins would not be updatable; problem solved. 
   > 
   > I see how that is what the spec says, but aren't the majority of joins that
   > people use/want to update a join of some type?  I thought that SQL99 allowed
   > updating view created by joins.
   > 
   > In either case is this a place where "exceeding" the spec would be a good 
   > thing or a bad thing?
   
   Lets try to meet the spec first, then debate about whether extending it
   is a good or bad thing :)
   
   -- 
   Rod Taylor <rbt@rbt.ca>
   
   PGP Key: http://www.rbt.ca/rbtpub.asc
   

4. Re: Updateable views...

   Curt Sampson <cjs@cynic.net> — 2003-03-05T16:29:45Z

   On Wed, 5 Mar 2003, Eric D Nielsen wrote:
   
   > I see how that is what the spec says, but aren't the majority of joins that
   > people use/want to update a join of some type?  I thought that SQL99 allowed
   > updating view created by joins.
   >
   > In either case is this a place where "exceeding" the spec would be a good
   > thing or a bad thing?
   
   This is a case where exceeding the spec would be a very good thing.
   And there has been lots of research on how to make more complex views
   updatable; see Date's _Introduction to Database Systems, 7th Edition_
   for an in-depth discussion of this.
   
   That said, one step at a time is always good, and even having just the
   very simplest views updatable would be a very nice thing.
   
   cjs
   -- 
   Curt Sampson  <cjs@cynic.net>   +81 90 7737 2974   http://www.netbsd.org
       Don't you know, in this new Dark Age, we're all light.  --XTC
   
   

5. Re: Updateable views...

   Greg Stark <gsstark@mit.edu> — 2003-03-05T17:00:10Z

   Tom Lane <tgl@sss.pgh.pa.us> writes:
   
   > Eric D Nielsen <nielsene@MIT.EDU> writes:
   > > In either case is this a place where "exceeding" the spec would be a good 
   > > thing or a bad thing?
   > 
   > Unless there is an obvious definition of what updating a join means
   > (obvious not only to the implementor, but to the user) I think this
   > is dangerous territory.
   
   Joins are a *BIG* part of the reason people want updateable views. In every
   single case that I updated a view it was a join. Just being able to update
   subsets of tables or restricted sets of columns is really a fairly trivial use
   of a powerful feature.
   
   In Oracle the constraint is fairly straightforward (at least to describe): 
   for each column you're updating the primary key of the table it came from has
   to be present in the view.
   
   
   -- 
   greg