Hi Kathey,
It seems to me that Suresh's response on the "paging Satheesh" thread
applies here too. The BOOLEAN type was only disabled at the language
level and has continued to be supported at the Store level. Several
system tables, in fact, have BOOLEAN columns: sysconglomerates,
sysstatements, sysaliases, systriggers, and sysstatistics. So there
shouldn't be any on-disk downgrade issues if we re-enable BOOLEAN.
Cheers,
-Rick
Kathey Marsden wrote:
Myrna van Lunteren wrote:
On 8/9/05, Jeffrey Lichtman <[EMAIL PROTECTED]> wrote:
This looks to me like an edge case. Rather than disable an entire
datatype, I'd recommend logging bugs on these edge cases. Unless
code archeology at IBM discloses some serious problem like data
corruption, I would recommend re-enabling the BOOLEAN datatype.
-Rick
I agree.
I have looked at the parser, datatype code and JDBC implementation,
and I don't see where anything is missing. Cloudscape used to have a
full implementation of boolean before IBM disabled a bunch of
features. This included the use of "?" parameters, boolean
expressions, et al. It looks to me like all IBM did was put one line
of code in the parser to prevent the use of the boolean type except
for internal purposes. Unless I'm mistaken, it should be possible to
completely re-enable the feature by eliminating that one line of code.
The line of code is:
checkInternalFeature(TypeId.BOOLEAN_NAME);
in the dataTypeCommon() rule.
- Jeff Lichtman
[EMAIL PROTECTED]
Check out Swazoo Koolak's Web Jukebox at
http://swazoo.com/
Yes, at IBM the Boolean type was disabled. I haven't done the
archeology, but disabling was not done because of some horrendous known bug
in this area.
However, before someone undisables the lines identified by Jeff, I will
mention that more work than meets the eye went into it - at least into
adjusting the tests. Many of the tests are intended to test functionality
with every single datatype. When the Boolean datatype was disabled, all
these tests had to be modified and usage of Boolean is thus *nowhere* tested
anymore, and if reinstated, testing should be rewritten.
Also, the derbyclient probably has no support for Boolean datatype.
Myrna
I am not too up on what is or is not needed for BOOLEAN, but if
someone puts it in, pretty please don't forget that soft upgrade check #:)
Thanks
Kathey
