Re: Possible Type Conversion Defect

We wanted to be very explicit with our data coercion story and as far as we
have been able we are not going to allow getting a type from the server that
would require a downcast to the client and possible loss of data. This
strategy has the advantage of high predictability with limited chance of
data loss, but it is very restrictive.

Quite frankly I was expecting to see a lot more people commenting on these.
In your case 108 is of type NUMERIC, a 38bit precission decimal and you are
trying to shove it into an INTEGER. Type 38 is an INTEGER which does not fit
on a SMALLINT.

We only have two choices here that don't involve data loss (something we are
definitelly not going to allow),
1) We can NEVER allow a conversion from a type if _the type you are trying
to convert_ does not fit into the type that you are trying to coerce it
into. This is the behavior that we have opted for in the 2005 JDBC driver.

2) We can allow a conversion from a type that does not fit into the coreced
type _only_ when the current value that you are asking for can be coerced
into the type that you are asking for. This is the behavior of the 2000 JDBC
driver.

Let's say that you have a NUMERIC column that has a value of 5, when you
call getInt on this we will throw an exception if following (1) but the
coercion will work on a driver that supports (2) since 5 does fit into an
INTEGER type. When you have a driver that provides the (1) functionality you
will realize the first time you run your code that a NUMERIC column will not
always fit into an int and change your code accordingly. When working with a
driver of the (2) type you will test and deploy your application with
getInt. When the value of the NUMERIC column goes over what an INTEGER can
handle you will get a runtime exception and you will have to go service your
deployed application.

We realize that it can be inconvenient to have this kind of issues surfaced
early, but we feel it is better to let you know up front about possible data
coercion issues, if you really wanted to get an integer from the server you
would have defined your table accordingly, or you could have requested an
integer in your query with the CONVERT function.

I think that this is going to be a common question, I am going to convert
this post into a blog and post it into the http://blogs.msdn.com/dataaccess/
with a complete data coercion table to help make this design clearer, of
course comments/suggestions are welcome.
--
Angel Saenz-Badillos [MS] DataWorks
This posting is provided "AS IS", with no warranties, and confers no
rights.Please do not send email directly to this alias.
This alias is for newsgroup purposes only.
I am now blogging: http://weblogs.asp.net/angelsb/




"Eric Molitor" <EricMolitor@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:237EAB6D-63BA-4DB2-AB0C-7EB8D98B3E2D@xxxxxxxxxxxxxxxx
> Seem to be getting these consistantly in certain portions of our
> application.
> Works fine with the older JDBC driver (2000) but under the 2005 driver we
> see....
>
> com.microsoft.sqlserver.jdbc.SQLServerException: Unsupported conversion
> from
> 108 to INTEGER
> at com.microsoft.sqlserver.jdbc.SQLServerStatement.getRowsetField(Unknown
> Source)
> at com.microsoft.sqlserver.jdbc.SQLServerResultSet.getInt(Unknown Source)
> at com.microsoft.sqlserver.jdbc.SQLServerResultSet.getInt(Unknown Source)
>
> and
>
> com.microsoft.sqlserver.jdbc.SQLServerException: Unsupported conversion
> from
> 38 to SMALLINT
> at com.microsoft.sqlserver.jdbc.SQLServerStatement.getRowsetField(Unknown
> Source)
> at com.microsoft.sqlserver.jdbc.SQLServerResultSet.getShort(Unknown
> Source)
> at com.microsoft.sqlserver.jdbc.SQLServerResultSet.getShort(Unknown
> Source)
>
> I'm looking at the SQL and the code but I'd think 38 is a valid SMALLINT
> and
> that 108 is a valid INTEGER.
>
>


.

Relevant Pages

  • Re: Possible Type Conversion Defect
    ... There is still time to integrate customer feedback into this data coercion ... >> to convert_ does not fit into the type that you are trying to coerce it ... >> driver. ...
    (microsoft.public.sqlserver.jdbcdriver)
  • Re: bindings 4 a rookie...
    ... the Driver X's because of the narrow fit of the Malamutes didn't fit ... the Burton P1 Carbon bindings and switched boots as well to the Burton Ions. ...
    (rec.skiing.snowboard)
  • Re: Driver Signing without Windows Logo
    ... In short these kinds of devices DON'T fit into the WHQL scheme. ... > My company has a USB medical device and we have developed a USB driver. ...
    (microsoft.public.development.device.drivers)
  • Re: Saving slides to pdf
    ... either truncated the image to fit, or scale it to fit in the allowable image ... You need to look for a printer or a driver at least, ... you may only get true borderless printing on three ... but no luck. ...
    (microsoft.public.mac.office.powerpoint)