That's right we *can't change the spec*.
But hold on guys, I can put a try catch block for this and prevent
the SQLException from being thrown in the implementation,
wherever it is being called in from *other classes*.
And hence we can go on.
But yes I can't change the spec class from doing that.
It's(return -1 for variable length data) is used by various JDBC drivers
that way.
Does that help ?
Also, Can you give me the stack trace of the exception ?
I just want to confirm the try catch block I put is at the right place
(i.e. exception will be caught in CachedRowSetImpl.java),
and works with you guys.
Further I am going to file a bug with our bug tracking system
and get some feedback internally, so that I keep things black and white.
You can view it over the web later as well.
thanks,
Amit Handa,
Sun Microsystems Inc.
PS: Kevin seems to have vanished ;)
I was trying to help him with this and some other problems but never
heard back,
though I am also slow at times to get back to him because of other
commitments.
Kris Jurka wrote:
>On Wed, 9 Feb 2005, Nebojsa Vasiljevic wrote:
>
>
>
>>[CachedRowSet doesn't like -1 precision values for numeric]
>>
>>
>
>This has been discussed, partially offlist, with Sun in the context of
>getColumnDisplaySize returning -1 for text and varchar with no length
>attribute fields. I've been waiting for a suggestion from Sun on what
>value to use, or if they will alter their code to remove this check.
>
>Your suggestion of using 0 passes their check, but still isn't legal:
>
>jurka=# create table nt (a numeric(0,0));
>ERROR: NUMERIC precision 0 must be between 1 and 1000
>
>The -1 value is intended to indicate we don't know or unlimited or
>similar and I don't particularly like it, but I've yet to hear a better
>suggestion.
>
>Kris Jurka
>
>
>