I haven't tested it fully yet, but here is a few preliminary notes...
This one is probably not related to the previous problem, but here is
a strange log:
Jan 16, 2008 12:11:08 AM org.apache.cayenne.access.QueryLogger log
INFO: \--> [ALTER TABLE PAINTING ADD COLUMN NEWCOL1 VARCHAR(10)]
Also there is a caveat with auto-increments: even if a column is
marked as "generated", user can still provide it explicitly and/or
adapter may not support auto generation, so Cayenne will provide the
value explicitly.
Andrus
On Jan 15, 2008, at 10:47 PM, Michael Gentry wrote:
Andrus, I had to modify the QueryLogger API just a tad to pass down if
the query was an INSERT query (which are the ones you want to skip
showing the PK value). My test cases worked. Please let me know if
you encounter other issues.
Thanks,
/dev/mrg
On Jan 13, 2008 10:47 AM, Michael Gentry <[EMAIL PROTECTED]> wrote:
Ah, I didn't look that closely at it. That would indeed be an issue.
On Jan 13, 2008 10:37 AM, Andrus Adamchik <[EMAIL PROTECTED]>
wrote:
HSQLDB 1.8 has an identity column. To quote the documentation,
"Identity columns are autoincrement columns."
As an aside, last time I checked, the generated values were not
passed
back via the java.sql.Statement.getGeneratedKeys() - the method we
rely upon to init ObjectId's on commit. Maybe worth checkin again...
Andrus
On Jan 13, 2008, at 5:30 PM, Michael Gentry wrote:
HSQLDB 1.8 has an identity column. To quote the documentation,
"Identity columns are autoincrement columns."
I'll look at Derby. Thanks!
On Jan 13, 2008 10:28 AM, Andrus Adamchik <[EMAIL PROTECTED]>
wrote:
MySQL. But you can try Derby - it supports autoincremented PK
(while
HSQLDB does not).
Andrus
On Jan 13, 2008, at 5:13 PM, Michael Gentry wrote:
What database are you using? I'm guessing MySQL? I was trying
to
set
up HSQLDB, but the modeler doesn't generate the SQL to use
identity
columns so I'm not certain if Cayenne supports identity columns
with
it? I suppose I can install MySQL later (I normally use
PostgreSQL).
I had MySQL running before upgrading to Leopard, which broke it.
Thanks,
/dev/mrg
On Jan 2, 2008 11:41 AM, Andrus Adamchik <[EMAIL PROTECTED]>
wrote:
Found one issue with the new logging today - insert queries with
auto
incremented PK incorrectly shift the labels to include the PK.
E.g.,
assuming auto-increment PK column name is "C2":
INSERT INTO mytable (C1, C3, C4) VALUES (?, ?, ?)
[bind: 1->C1:V1, 2->C2:V3, 3->C3:V4]
At position #2, the PK label that is not used in the statement
SQL is
shown, and from there, all the values are off by one position.
Andrus
On Dec 28, 2007, at 5:07 PM, Michael Gentry wrote:
That's what I was hoping. Thanks for the update!
/dev/mrg
On Dec 28, 2007 9:53 AM, Andrus Adamchik <[EMAIL PROTECTED]
>
wrote:
I deployed this in production some time back. I am really
happy
with
the new format - makes log analysis much easier.
Cheers,
Andrus
On Nov 30, 2007, at 4:39 AM, Michael Gentry wrote:
Changes are checked in. Please let me know if you have
problems,
etc.
Thanks,
/dev/mrg
