Hi,

I've got a customer that has recently upgraded to Ingres 9.3 from
6.4/03.

The following statement now gives the customer a database error:

update case set txt = 'bob' where case.ref = 1;

The error states that there is a syntax error on '.'

Removing the 'case' table correlation name from the where clause allows
it to run without fail.

I've a slight problem in that I cannot change the source, so any
suggestions along the lines of putting quotes around the correlation
name won't help.

Is anyone aware of a 'switch' at the database installation level that
would cause the parser(?) to recognise this as a table correlation name
and not a 'case' statement (which I think is what is happening).

Thanks

Adrian


--
Adrian Williamson

On Feb 2, 2011, at 9:44 AM, Ingres Forums wrote:

Nope.

If you can't change the source, you're out of luck.

Karl

Hi Karl,

As I feared, thanks for responding.

Adrian


--
Adrian Williamson

In the past I have been able to "rescue" similar situations using a
binary editor, substituting strings. Haven't tried it for a long, long
time though...


--
ipauland

Hi Adrian,

Do you mind me asking why the customer has gone to 9.3? This isn't a
supported release and was only ever intended as a release that would give
early exposure to new features for SI's and ISV's. Patches are not available
and upgrading to 10 will almost certainly require an unload, reinstall and
reload.

Peter Gale

On 2 February 2011 14:44, Ingres Forums <
info-ingres (AT) kettleriverconsulting (DOT) com> wrote:



--
Peter Gale
pgale61 (AT) gmail (DOT) com

If you're really desperate... you could try using a hex editor to edit
the ingres iidbms process executable (iimerge?) and replace instances of
'case' with 'pase' or some such. You'd effectively be changing the
reserved word itself so your original query would be unaffected. (Or
perhaps you can hex edit your executable to change case to another name
and change the DB schema accordingly?)

It might not work but might be worth a quick try for just amusement's
sake. :-)

Or you could even recompile your own custom version of Ingres from the
open source with 'case' changed to something else...

Patching the DBMS sounds like a seriously bad idea, I think. My rescue
idea was only with respect to the client app and would not have any
effect on anything else.


--
ipauland

Peter,

I did this quite awhile ago when few people realised this. Ingres
eventually pointed this out after a week of dealing with my issue.
Maybe it still isn't well documented? But like you, I was wondering....

At 11:12 AM 2/2/2011, you wrote:

On 02/02/2011 22:02, Ingres Forums wrote:
"hacking" the client exe is the most sensible course of action compared
with the DBMS executable (but still a little crazy!).

Hi,

Yes I had a couple of goes at global replace on the W4GL image file,
but it seems that W4GL is very sensative to change. (or I wasn't trying
hard enough with the binary editors)

We're on 9.3 Peter, to give us access to the II_WINSOCK2_NODELAY
parameter, which works wonders for the network perfomance.

The database is of a mangelable size so a move to 10 once everyone else
has found the bugs would be an activity measured in tens of minutes.

Thanks for your input, the binary edit of the parser to distract it
from the concept of 'case' was inspirational.

Biodiversity should be celebrated.

Adrian


--
Adrian Williamson